Biosystems & Biorobotics
José González-Vargas Jaime Ibáñez Jose L. Contreras-Vidal Herman van der Kooij José Luis Pons Editors
Wearable Robotics: Challenges and Trends Proceedings of the 2nd International Symposium on Wearable Robotics, WeRob2016, October 18–21, 2016, Segovia, Spain
Biosystems & Biorobotics Volume 16
Series editor Eugenio Guglielmelli, Laboratory of Biomedical Robotics, Università Campus Bio-Medico, Rome, Italy e-mail:
[email protected] Editorial Board Dino Accoto, Campus Bio-Medico University of Rome, Rome, Italy Sunil Agrawal, University of Delaware, Newark, DE, USA Fabio Babiloni, Sapienza University of Rome, Rome, Italy Jose M. Carmena, University of California, Berkeley, CA, USA Maria Chiara Carrozza, Scuola Superiore Sant’Anna, Pisa, Italy Paolo Dario, Scuola Superiore Sant’Anna, Pisa, Italy Arturo Forner-Cordero, University of Sao Paolo, São Paulo, Brazil Masakatsu G. Fujie, Waseda University, Tokyo, Japan Nicolas Garcia, Miguel Hernández University of Elche, Elche, Spain Neville Hogan, Massachusetts Institute of Technology, Cambridge, MA, USA Hermano Igo Krebs, Massachusetts Institute of Technology, Cambridge, MA, USA Dirk Lefeber, Universiteit Brussel, Brussels, Belgium Rui Loureiro, Middlesex University, London, UK Marko Munih, University of Ljubljana, Ljubljana, Slovenia Paolo M. Rossini, University Cattolica del Sacro Cuore, Rome, Italy Atsuo Takanishi, Waseda University, Tokyo, Japan Russell H. Taylor, The Johns Hopkins University, Baltimore, MA, USA David A. Weitz, Harvard University, Cambridge, MA, USA Loredana Zollo, Campus Bio-Medico University of Rome, Rome, Italy
Aims & Scope Biosystems & Biorobotics publishes the latest research developments in three main areas: 1) understanding biological systems from a bioengineering point of view, i.e. the study of biosystems by exploiting engineering methods and tools to unveil their functioning principles and unrivalled performance; 2) design and development of biologically inspired machines and systems to be used for different purposes and in a variety of application contexts. The series welcomes contributions on novel design approaches, methods and tools as well as case studies on specific bioinspired systems; 3) design and developments of nano-, micro-, macrodevices and systems for biomedical applications, i.e. technologies that can improve modern healthcare and welfare by enabling novel solutions for prevention, diagnosis, surgery, prosthetics, rehabilitation and independent living. On one side, the series focuses on recent methods and technologies which allow multiscale, multi-physics, high-resolution analysis and modeling of biological systems. A special emphasis on this side is given to the use of mechatronic and robotic systems as a tool for basic research in biology. On the other side, the series authoritatively reports on current theoretical and experimental challenges and developments related to the “biomechatronic” design of novel biorobotic machines. A special emphasis on this side is given to human-machine interaction and interfacing, and also to the ethical and social implications of this emerging research area, as key challenges for the acceptability and sustainability of biorobotics technology. The main target of the series are engineers interested in biology and medicine, and specifically bioengineers and bioroboticists. Volume published in the series comprise monographs, edited volumes, lecture notes, as well as selected conference proceedings and PhD theses. The series also publishes books purposely devoted to support education in bioengineering, biomedical engineering, biomechatronics and biorobotics at graduate and post-graduate levels.
About the Cover The cover of the book series Biosystems & Biorobotics features a robotic hand prosthesis. This looks like a natural hand and is ready to be implanted on a human amputee to help them recover their physical capabilities. This picture was chosen to represent a variety of concepts and disciplines: from the understanding of biological systems to biomechatronics, bioinspiration and biomimetics; and from the concept of human-robot and human-machine interaction to the use of robots and, more generally, of engineering techniques for biological research and in healthcare. The picture also points to the social impact of bioengineering research and to its potential for improving human health and the quality of life of all individuals, including those with special needs. The picture was taken during the LIFEHAND experimental trials run at Università Campus Bio-Medico of Rome (Italy) in 2008. The LIFEHAND project tested the ability of an amputee patient to control the Cyberhand, a robotic prosthesis developed at Scuola Superiore Sant’Anna in Pisa (Italy), using the tf-LIFE electrodes developed at the Fraunhofer Institute for Biomedical Engineering (IBMT, Germany), which were implanted in the patient’s arm. The implanted tf-LIFE electrodes were shown to enable bidirectional communication (from brain to hand and vice versa) between the brain and the Cyberhand. As a result, the patient was able to control complex movements of the prosthesis, while receiving sensory feedback in the form of direct neurostimulation. For more information please visit http://www.biorobotics.it or contact the Series Editor.
More information about this series at http://www.springer.com/series/10421
José González-Vargas Jaime Ibáñez Jose L. Contreras-Vidal Herman van der Kooij José Luis Pons •
•
Editors
Wearable Robotics: Challenges and Trends Proceedings of the 2nd International Symposium on Wearable Robotics, WeRob2016, October 18–21, 2016, Segovia, Spain
123
Editors José González-Vargas Spanish National Research Council Cajal Institute Madrid Spain
Herman van der Kooij University of Twente Enschede The Netherlands José Luis Pons Spanish National Research Council Cajal Institute Madrid Spain
Jaime Ibáñez Spanish National Research Council Cajal Institute Madrid Spain Jose L. Contreras-Vidal Cullen College of Engineering University of Houston Houston, TX USA
ISSN 2195-3562 Biosystems & Biorobotics ISBN 978-3-319-46531-9 DOI 10.1007/978-3-319-46532-6
ISSN 2195-3570
(electronic)
ISBN 978-3-319-46532-6
(eBook)
Library of Congress Control Number: 2016951712 © Springer International Publishing AG 2017 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. Printed on acid-free paper This Springer imprint is published by Springer Nature The registered company is Springer International Publishing AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Contents
Part I Clinical Focus on Rehabilitation and Assistive WRs Clinical Evaluation of a Socket-Ready Naturally Controlled Multichannel Upper Limb Prosthetic System . . . . . . . . . . . . . . . . . . . . . . Ivan Vujaklija, Sebastian Amsuess, Aidan D. Roche, Dario Farina and Oskar C. Aszmann Evaluation of a Robotic Exoskeleton for Gait Training in Acute Stroke: A Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ghaith J. Androwis and Karen J. Nolan Wearable Exoskeleton Assisted Rehabilitation in Multiple Sclerosis: Feasibility and Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shuo-Hsiu Chang, Marcie Kern, Taimoor Afzal, Shih-Chiao Tseng, John Lincoln and Gerard Francisco Exoskeletons for Rehabilitation and Personal Mobility: Creating Clinical Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. T. Arun Jayaraman and William Z. Rymer Lower Limb Wearable Systems for Mobility and Rehabilitation Challenges: Clinical Focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Federica Tamburella, Marcella Masciullo, Iolanda Pisotta, Nevio Luigi Tagliamonte and Marco Molinari
3
9
15
21
25
Part II Emerging Technologies in WRs Impedance Control of Series Elastic Actuators Using Acceleration Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrea Calanca, Riccardo Muradore and Paolo Fiorini
33
v
vi
Contents
Kinetic Energy Recovery in Human Joints: The Flywheel-Infinitely Variable Transmission Actuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Roberta Alò, Francesco Bottiglione and Giacomo Mantriota A Compliant Lightweight and Adaptable Active Ankle Foot Orthosis for Robotic Rehabilitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marta Moltedo, Tomislav Bacek, Kevin Langlois, Karen Junius, Bram Vanderborght and Dirk Lefeber A Novel Shoulder Mechanism with a Double Parallelogram Linkage for Upper-Body Exoskeletons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Simon Christensen and Shaoping Bai A Soft Robotic Extra-Finger and Arm Support to Recover Grasp Capabilities in Chronic Stroke Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . Irfan Hussain, Gionata Salvietti, Giovanni Spagnoletti, David Cioncoloni, Simone Rossi and Domenico Prattichizzo A Quasi-Passive Knee Exoskeleton to Assist During Descent . . . . . . . . . Emily Rogers, Panagiotis Polygerinos, Stephen Allen, Fausto A. Panizzolo, Conor J. Walsh and Dónal P. Holland Wearable Sensory Apparatus for Multi-segment System Orientation Estimation with Long-Term Drift and Magnetic Disturbance Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sebastjan Šlajpah, Roman Kamnik and Marko Munih A Portable Active Pelvis Orthosis for Ambulatory Movement Assistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrea Parri, Tingfang Yan, Francesco Giovacchini, Mario Cortese, Marco Muscolo, Matteo Fantozzi, Raffaele Molino Lova and Nicola Vitiello
39
45
51
57
63
69
75
Part III Soft Wearable Robotics XoSoft - A Vision for a Soft Modular Lower Limb Exoskeleton . . . . . . . Jesus Ortiz, Eduardo Rocon, Valerie Power, Adam de Eyto, Leonard O’Sullivan, Markus Wirz, Christoph Bauer, Samuel Schülein, Konrad S. Stadler, Barbara Mazzolai, Wouter B. Teeuw, Chris Baten, Corien Nikamp, Jaap Buurke, Freygardur Thorsteinsson and Jeanette Müller On the Efficacy of Isolating Shoulder and Elbow Movements with a Soft, Portable, and Wearable Robotic Device . . . . . . . . . . . . . . . . Zahra Kadivar, Christopher E. Beck, Roger N. Rovekamp, Marcia K. O’Malley and Charles A. Joyce
83
89
Contents
vii
Design Improvement of a Polymer-Based Tendon-Driven Wearable Robotic Hand (Exo-Glove Poly) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Haemin Lee, Brian Byunghyun Kang, Hyunki In and Kyu-Jin Cho
95
Affective Touch and Low Power Artificial Muscles for Rehabilitative and Assistive Wearable Soft Robotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Jonathan Rossiter, Espen Knoop and Yuichi Nakamura Evaluation of Force Tracking Controller with Soft Exosuit for Hip Extension Assistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Hao Su, Ye Ding, Ignacio Galiana, Jozefien Speeckaert, Nikos Karavas, Philippe Malcolm, Christopher Siviy and Conor J. Walsh Part IV Neural Interfacing of WRs Endogenous Control of Powered Lower-Limb Exoskeleton . . . . . . . . . . . 115 Kyuhwa Lee, Dong Liu, Laetitia Perroud, Ricardo Chavarriaga and José del R. Millán Natural User-Controlled Ambulation of Lower Extremity Exoskeletons for Individuals with Spinal Cord Injury . . . . . . . . . . . . . . . 121 Kiran Karunakaran, Ghaith Androwis and Richard Foulds Real-Time Modeling for Lower Limb Exoskeletons . . . . . . . . . . . . . . . . . 127 Guillaume Durandau, Massimo Sartori, Magdo Bortole, Juan C. Moreno, José L. Pons and Dario Farina Towards Everyday Shared Control of Lower Limb Exoskeletons . . . . . . 133 Tom Carlson Part V Biomechanics and Neurophysiological studies with WRs Joint-Level Responses to Counteract Perturbations Scale with Perturbation Magnitude and Direction . . . . . . . . . . . . . . . . . . . . . . . 139 Mark Vlutters, Edwin H. F. van Asseldonk and Herman van der Kooij Metabolic Energy Consumption in a Box-Lifting Task: A Parametric Study on the Assistive Torque . . . . . . . . . . . . . . . . . . . . . . 143 Mohammad S. Shourijeh, Moonki Jung and Michael Damsgaard Analysis of the Movement Variability in Dance Activities Using Wearable Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Miguel Xochicale, Chris Baber and Mourad Oussalah
viii
Contents
Part VI New developments in Wearable Rehabilitation Robotics Real Time Computation of Centroidal Momentum for the Use as a Stability Index Applicable to Human Walking with Exoskeleton . . . . . . 157 Je Hyung Jung, Lidwine Van Opheusden, Pierre Barralon and Jan F. Veneman A Versatile Neuromuscular Exoskeleton Controller for Gait Assistance: A Preliminary Study on Spinal Cord Injury Patients . . . . . . 163 Amy R. Wu, Florin Dzeladini, Tycho J. H. Brug, Federica Tamburella, Nevio L. Tagliamonte, Edwin van Asseldonk, Herman van der Kooij and Auke J. Ijspeert Introducing a Modular, Personalized Exoskeleton for Ankle and Knee Support of Individuals with a Spinal Cord Injury . . . . . . . . . 169 Cory Meijneke, Shiqian Wang, Victor Sluiter and Herman van der Kooij Towards Exoskeletons with Balance Capacities . . . . . . . . . . . . . . . . . . . . 175 Herman van der Kooij, Edwin H. F. van Asseldonk and Mark Vlutters EMG-Based Detection of User’s Intentions for Human-Machine Shared Control of an Assistive Upper-Limb Exoskeleton . . . . . . . . . . . . 181 A. Accogli, L. Grazi, S. Crea, A. Panarese, J. Carpaneto, N. Vitiello and S. Micera Part VII Legal Framework, standardization and ethical issues in WRs Safety Standardization of Wearable Robots—The Need for Testing Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 Jan F. Veneman The Potential and Acceptance of Exoskeletons in Industry . . . . . . . . . . . 195 Michiel P. de Looze, Frank Krause and Leonard W. O’Sullivan Wearable Robots: A Legal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Andrea Bertolini A Verification Method for Testing Abrasion in the Use of Restraint Type Personal Care Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Yoji Yamada, Mao Xuewei, Kengo Yoshida, Yasuhiro Akiyama and Shogo Okamoto Part VIII Benchmarking in WRs and related communities Kinematic Comparison of Gait Rehabilitation with Exoskeleton and End-Effector Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Byung-Woo Ko and Won-Kyung Song
Contents
ix
Evaluating the Gait of Lower Limb Prosthesis Users . . . . . . . . . . . . . . . . 219 Stephanie L. Carey, Kyle B. Reed, Amanda Martori, Tyagi Ramakrishnan and Rajiv Dubey Some Considerations on Benchmarking of Wearable Robots for Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 Jan F. Veneman Benchmarking Data for Human Walking in Different Scenarios. . . . . . . 231 Katja Mombaur, Debora Clever and Alexander Schubert Clinical Gait Assessment in Relation to Benchmarking Robot Locomotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 J. H. Buurke, J. F. Veneman and D. Torricelli Part IX Symbiotic Control of WRs Attention Level Measurement During Exoskeleton Rehabilitation Through a BMI System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 Álvaro Costa, Guillermo Asín-Prieto, José González-Vargas, Eduardo Iáñez, Juan C. Moreno, Antonio J. Del-Ama, Ángel Gil-Agudo and José M. Azorín Detection of Subject’s Intention to Trigger Transitions Between Sit, Stand and Walk with a Lower Limb Exoskeleton . . . . . . . . . . . . . . . . . . 249 Fernando Trincado-Alonso, Antonio J. del Ama-Espinosa, Guillermo Asín-Prieto, Elisa Piñuela-Martín, Soraya Pérez-Nombela, Ángel Gil-Agudo, José Luis Pons and Juan C. Moreno The New Generation of Compliant Actuators for Use in Controllable Bio-Inspired Wearable Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255 T. Bacek, M. Moltedo, J. Gonzalez-Vargas, G. Asin Prieto, M. C. Sanchez-Villamañan, J. C. Moreno and D. Lefeber An EMG-informed Model to Evaluate Assistance of the Biomot Compliant Ankle Actuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 Elena Ceseracciu, Luca Tagliapietra, Juan C. Moreno, Guillermo Asin, Antonio J. del-Ama, Soraya Pérez, Elisa Piñuela, Ángel Gil and Monica Reggiani Tacit Adaptability of a Mechanically Adjustable Compliance and Controllable Equilibrium Position Actuator, a Preliminary Study . . . . . 267 Guillermo Asín-Prieto, Shingo Shimoda, José González, M. Carmen Sánchez-Villamañán, José L. Pons and Juan C. Moreno
x
Contents
Part X Emerging Applications Domains of WRs, Emerging Technologies in WRs Design and Kinematic Analysis of the Hanyang Exoskeleton Assistive Robot (HEXAR) for Human Synchronized Motion . . . . . . . . . . . . . . . . . 275 Wansoo Kim, Hojun Kim, Donghwan Lim, Hyungi Moon and Changsoo Han Design and Experimental Evaluation of a Low-Cost Robotic Orthosis for Gait Assistance in Subjects with Spinal Cord Injury . . . . . . . . . . . . . 281 Josep M. Font-Llagunes, Daniel Clos, Urbano Lugrís, F. Javier Alonso and Javier Cuadrado A Powered Low-Back Exoskeleton for Industrial Handling: Considerations on Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 Stefano Toxiri, Jesús Ortiz, Jawad Masood, Jorge Fernández, Luis A. Mateos and Darwin G. Caldwell Efficient Lower Limb Exoskeleton for Human Motion Assistance . . . . . 293 Nazim Mir-Nasiri Active Safety Functions for Industrial Lower Body Exoskeletons: Concept and Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 Jawad Masood, Luis A. Mateos, Jesús Ortiz, Stefano Toxiri, Leonard O’Sullivan and Darwin Caldwell SOLEUS: Ankle Foot Orthosis for Space Countermeasure with Immersive Virtual Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305 Pierre Letier, Guillaume Fau, Uwe Mittag, Jochen Zange, Joern Rittweger, Moonki Jung, Jo McIntyre and Arnaud Runge SPEXOR: Spinal Exoskeletal Robot for Low Back Pain Prevention and Vocational Reintegration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 Jan Babič, Katja Mombaur, Dirk Lefeber, Jaap van Dieën, Bernhard Graimann, Michael Russold, Nejc Šarabon and Han Houdijk Part XI Posters HeSA, Hip Exoskeleton for Superior Assistance . . . . . . . . . . . . . . . . . . . . 319 Thomas G. Sugar, Eduardo Fernandez, Darren Kinney, Kevin W. Hollander and Sangram Redkar SPEXOR: Towards a Passive Spinal Exoskeleton . . . . . . . . . . . . . . . . . . 325 L. De Rijcke, M. Näf, C. Rodriguez-Guerrero, B. Graimann, H. Houdijk, J. van Dieën, K. Mombaur, M. Russold, N. Sarabon, J. Babič and D. Lefeber
Contents
xi
Autonomous Soft Exosuit for Hip Extension Assistance. . . . . . . . . . . . . . 331 Nikos Karavas, Jinsoo Kim, Ignacio Galiana, Ye Ding, Adam Couture, Diana Wagner, Asa Eckert-Erdheim and Conor Walsh Comparison of Ankle Moment Inspired and Ankle Positive Power Inspired Controllers for a Multi-Articular Soft Exosuit for Walking Assistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 Martin Grimmer, Brendan T. Quinlivan, Sangjun Lee, Philippe Malcolm, Denise M. Rossi, Christopher Siviy and Conor J. Walsh Biomechanical Analysis and Inertial Sensing of Ankle Joint While Stepping on an Unanticipated Bump . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Taira Miyatake, Sangjun Lee, Ignacio Galiana, Denise M. Rossi, Christopher Siviy, Fausto A. Panizzolo and Conor J. Walsh A Novel Approach to Increase Upper Extremity Active Range of Motion for Individuals with Duchenne Muscular Dystrophy Using Admittance Control: A Preliminary Study . . . . . . . . . . . . . . . . . . . 349 Madeline C. Corrigan and Richard A. Foulds Modulation of Knee Range of Motion and Time to Rest in Cerebral Palsy Using Two Forms of Mechanical Stimulation . . . . . . . . . . . . . . . . . 355 Peter A. Michael, Ghaith J. Androwis and Richard A. Foulds Training Response to Longitudinal Powered Exoskeleton Training for SCI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361 A. Ramanujam, A. Spungen, P. Asselin, E. Garbarini, J. Augustine, S. Canton, P. Barrance and G. F. Forrest Adaptive Classification of Arbitrary Activities Through Hidden Markov Modeling with Automated Optimal Initialization . . . . . . . . . . . . 367 Chris T. M. Baten, Thijs Tromper and Leonie Zeune Design and Motion Analysis of a Wearable and Portable Hand Exoskeleton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373 Matteo Bianchi, Francesco Fanelli, Roberto Conti, Lapo Governi, Enrico Meli, Alessandro Ridolfi, Andrea Rindi, Federica Vannetti and Benedetto Allotta Nitiglove: Nitinol-Driven Robotic Glove Used to Assist Therapy for Hand Mobility Recovery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 C. Jiménez, R. Mora, J. Pérez and K. Quirós 3D Printed Arm Exoskeleton for Teleoperation and Manipulation Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385 Pierre Letier, Gonzalo Rodriguez, Guillaume Fau, Shashank Govindaraj, Torsten Siedel, Jeremi Gancet and Michel Ilzkovitz
xii
Contents
Musculoskeletal Simulation of SOLEUS Ankle Exoskeleton for Countermeasure Exercise in Space . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 Moonki Jung, Guillaume Fau, Pierre Letier, Uwe Mittag, Jochen Zange, Jörn Rittweger and Arnaud Runge Human Gait Feature Detection Using Inertial Sensors Wavelets . . . . . . . 397 S. Glowinski, A. Blazejewski and T. Krzyzynski On the Importance of a Motor Model for the Optimization of SEA-driven Prosthetic Ankles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403 Tom Verstraten, Glenn Mathijssen, Joost Geeroms, Louis Flynn, Bram Vanderborght and Dirk Lefeber Assessment of a 7-DOF Hand Exoskeleton for Neurorehabilitation . . . . 409 Kevin Abbruzzese and Richard Foulds Improving the Standing Balance of People with Spinal Cord Injury Through the Use of a Powered Ankle-Foot Orthosis . . . . . . . . . . . . . . . . 415 Amber Emmens, Iolanda Pisotta, Marcella Masciullo, Edwin van Asseldonk and Herman van der Kooij Transparent Mode for Lower Limb Exoskeleton . . . . . . . . . . . . . . . . . . . 421 R. Mendoza-Crespo, R. Soto and J. L. Pons Human-Robot Mutual Force Borrowing and Seamless Leader-Follower Role Switching by Learning and Coordination of Interactive Impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427 Ker-Jiun Wang, Mingui Sun and Zhi-Hong Mao Upper Limb Exoskeleton Control for Isotropic Sensitivity of Human Arm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433 Rok Goljat, Tadej Petrič and Jan Babič AUTONOMYO: Design Challenges of Lower Limb Assistive Device for Elderly People, Multiple Sclerosis and Neuromuscular Diseases . . . . 439 A. Ortlieb, M. Bouri and H. Bleuler Passive Lower Back Moment Support in a Wearable Lifting Aid: Counterweight Versus Springs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445 Ralph J. L. Macke, Arvid Q. L. Keemink and Arno H. A. Stienen
Part I
Clinical Focus on Rehabilitation and Assistive WRs
Clinical Evaluation of a Socket-Ready Naturally Controlled Multichannel Upper Limb Prosthetic System Ivan Vujaklija, Sebastian Amsuess, Aidan D. Roche, Dario Farina and Oskar C. Aszmann
Abstract Research conducted over the last decades indicates a necessity of having larger number of EMG sensors in order to extract sufficient information needed for natural control of upper limb prosthetics. Various studies have addressed this issue, though clinical transition and evaluation of such systems on a larger pool of patients is still missing. We propose a specifically designed system which allows users to perform clinically relevant tests in an unobstructed way while handling dexterous prosthesis. Eight electrodes were embedded into customized sockets along with the controllers driving an algorithm recently tested in laboratory conditions that allows simultaneous manipulation of four out of seven prosthetic functions. The fully self-contained system was evaluated on seven amputees conducting the Southampton Hand Assessment Procedure. The scores achieved were compared to those obtained using their own commercial devices. The study shows the necessary steps to validate novel control algorithms in a clinically meaningful context.
This work was supported by the Christian Doppler Research Foundation of the Austrian Federal Ministry of Science, Research and Economy and by the European Research Council Advanced Grant DEMOVE (contract #267888). I. Vujaklija (&) D. Farina Institute of Neurorehabilitation Systems, University Medical Center Göttingen, Göttingen, Germany e-mail:
[email protected] D. Farina e-mail:
[email protected] S. Amsuess Ottobock Healthcare Products GmbH, Vienna, Austria e-mail:
[email protected] A.D. Roche O.C. Aszmann Christian Doppler Laboratory for Restoration of Extremity Function, Medical University of Vienna, Vienna, Austria e-mail:
[email protected] O.C. Aszmann e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_1
3
4
I. Vujaklija et al.
1 Introduction Various research conducted in the last decades indicated a necessity for introducing additional EMG sensors in order to provide upper limb prosthetic users with more versatile and natural control [1, 2]. Namely, the currently available solutions are almost strictly depending on just a few electrodes and a cumbersome control paradigm [3, 4]. On the other hand, more advanced systems have seemed to fail in reaching the clinical testing stage. Reasons for this are numerous, but they mostly boil down to the issue of robustness. Transferring myoelectric solutions into systems that can be tested in an acceptable manner through different clinical scenarios requires production of sockets which are capable of hosting the majority of the hardware. Even once the transfer is made, the system is facing challenges such as extensive sweat production, electrode misplacement, uneven loading of the socket and stump volume changes [5]. Here, we present a socket-ready implementation of a previously laboratory tested system for simultaneous and proportional control of the dexterous prosthetic hand [6]. The outcomes of a clinical evaluation are presented and discussed along with the challenges that were encountered during the testing and fitting process.
2 Methods 2.1
Subjects
Total of seven transradial amputees (6 male, 1 female, aged 35.14 ± 10.11) agreed to participate in the study after reading and signing consent forms approved by local ethics board of Medical University of Vienna (Ethics Commission number: 1044/2015). All participants are active myoelectric users, though have no previous experience with advanced control systems.
2.2
Hardware and Control Algorithm
All seven subjects were fitted with custom sockets by a certified prosthetist. Each socket was able to host eight Otto Bock raw signal electrodes 13E200 = 50AC, through a predefined gap positioned over the most muscular circumference of the forearm. Individual electrodes were placed into the 3D printed casings which enabled their quick equidistant placement using a single rubber band. Tip of the thermoplast based liner provided the support for lamination rings needed for prosthetic attachment and was enclosed in cellacast, the fiberglass free plaster.
Clinical Evaluation of a Socket-Ready Naturally Controlled Multichannel…
5
The prosthesis that was used throughout the experiment was Ottobock Michelangelo hand with wrist flexor and rotation units. If the length of the stump prevented battery pack and the Axon®bus system to be tucked into the socket, these components were simply attached on the outside of the shaft. Communication with the computer which was running the control algorithm was done via Bluetooth. In this way full flexibility and control over the implemented algorithm which allowed proportional activation of the wrist flexor and rotation units (linear regression) combined with sequential, proportional hand function (linear discriminant analysis classifier) [6] was available without compromising comfort of the user during testing.
2.3
Clinical Testing
In order to make a clinically relevant evaluation of patient performance using the implemented system Southampton Hand Assessment Procedure (SHAP) [7] was conducted. This test includes a variety of tasks ranging from more abstract ones to those replicating the activities of daily living. The overall end score can range from 0 representing no useful hand function at all, to 100 which is equivalent to a fully functional able bodied hand. Subjects were given two opportunities in performing each task and the execution time out was set to 100 s.
2.4
Experiment Protocol
Each subject was invited to come three times. During the first visit prosthetists conducted all the necessary measurements for building up the socket, and participants were once again briefed on what is expected from them throughout the experiment. In the second session the new socket was tested and modifications were made if needed. Additionally, subjects were asked to perform the SHAP test using their own myoelectric device in order to acquire the “baseline” score. In the final session, the training data needed for the control of the prosthetic device was collected in a standard myoelectric pattern recognition based manner, as in previously conducted study [8]. After a short brake the participant was invited to test the device and get familiar and comfortable with its capabilities and a final SHAP test was conducted.
6 Table 1 SHAP scores
I. Vujaklija et al. Subject ID
Baseline session
Advanced session
S1 S2 S3 S4 S5 S6 S7
77 65 72 62 59 46 53
56 50 36 47 39 35 32
3 Results All subjects were successfully fitted with the prosthesis and were able to follow the experiment. No participant reported any problems with the fitting during the third, advanced session. The only notable complaints were regarding the overall weight of the system and its poor weight distribution in cases of users with longer stumps. The results of SHAP testing are presented in Table 1. Baseline score across all subjects was 62.00 ± 10.65 and in the final session patients achieved the average score of 42.14 ± 8.93. Closest score using the advanced prosthesis to the baseline one was achieved by subject S6 with 11 points difference, while the greatest discrepancy was observed in S3, 36 points. Average SHAP score difference between the two systems was 19.86 ± 8.05.
4 Conclusions and Discussion In this study we have implemented an eight channels based advanced control system, which is able of delivering simultaneous and proportional control over 3.5 degrees of freedom, into a custom fitted wearable socket. The purpose was to conduct a clinically viable evaluation of the solution in order to identify its potential for real world applications. Based on the feedback received from the participants the terminal load of the prosthetic fitting was regarded problematic and a more balanced weight distribution should be aimed at, especially in patients with longer stumps. The fact that all participants managed to conduct the entire experiment using the proposed system, even though somewhat lower scores were achieved, indicates the potential of this approach of becoming a candidate for a proper clinical study. Though, it also points out a necessity for conducting a longitudinal evaluation which would eliminate the effects of learning and accommodation of the patient to an entirely new system and control paradigm. In addition, inclusion of motion analysis could indicate possible changes in compensatory movements [9].
Clinical Evaluation of a Socket-Ready Naturally Controlled Multichannel…
7
Acknowledgments The authors would like to extend their gratitude to Mr. Hans Opel and his team of prosthetists at Otto Bock Healthcare Products GmbH for manufacturing and fitting all the patients with custom made sockets.
References 1. Hudgins, B., Parker, P., Scott, R.N.: A new strategy for multifunction myoelectric control. IEEE Trans. Biomed. Eng. 40(1), 82–94 (1993) 2. Jiang, N., Vujaklija, I., Rehbaum, H., Graimann, B., Farina, D.: Is accurate mapping of EMG signals on kinematics needed for precise online myoelectric control? IEEE Trans. Neural Syst. Rehabil. Eng. 22(3), 549–558 (2014) 3. Scott, R.N., Parker, P.A.: Myoelectric prostheses: state of the art. J. Med. Eng. Technol. 12(4), 143–151 (1988) 4. Oskoei, M.A., Hu, H.: Myoelectric control systems-a survey. Biomed. Signal Process. Control 2(4), 275–294 (2007) 5. Hargrove, L., Englehart, K., Hudgins, B.: A training strategy to reduce classification degradation due to electrode displacements in pattern recognition based myoelectric control. Biomed. Signal Process. Control 3(2), 175–180 (2008) 6. Amsuess, S., Vujaklija, I., Gobel, P., Roche, A., Graimann, B., Aszmann, O., Farina, D.: Context-dependent upper limb prosthesis control for natural and robust use. IEEE Trans. Neural Syst. Rehabil. Eng., 1 (2015) 7. Light, C.M., Chappell, P.H., Kyberd, P.J.: Establishing a standardized clinical assessment tool of pathologic and prosthetic hand function: normative data, reliability, and validity. Arch. Phys. Med. Rehabil. 83(6), 776–783 (2002) 8. Roche, A.D., Vujaklija, I., Amsuess, S., Sturma, A., Göbel, P., Farina, D., Aszmann, O.C.: Structured rehabilitation training for improved multifunctional prosthetic control : a case study. J. Vis. Exp. (2015) 9. Montagnani, F., Controzzi, M., Cipriani, C.: Exploiting arm posture synergies in activities of daily living to control the wrist rotation in upper limb prostheses: a feasibility study. In: Proceedings of Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, vol. 2015, pp. 2462–2465, November 2015
Evaluation of a Robotic Exoskeleton for Gait Training in Acute Stroke: A Case Study Ghaith J. Androwis and Karen J. Nolan
Abstract The application of robotic exoskeletons in gait therapy during stroke rehabilitation has grown rapidly over the past decade. The purpose of this investigation is to determine the impact of a robotic exoskeleton (RE) on temporal spatial gait parameters as compared to traditional standard of care gait training in a single inpatient after acute stroke. Outcome measures included temporal spatial gait parameters while walking with and without an RE during a single gait training session two weeks post stroke. During gait training in the RE, walking speed, and stride length increased while step width decreased on the affected and unaffected side. Total double support time decreased, and single support and swing time increased on the affected and unaffected limb during gait training in the RE. Gait training in the RE had a positive effect on the patients overall gait which included improved temporal spatial parameters and gait speed. Positive changes in temporal spatial parameters were evident on the affected and unaffected limb. Preliminary rehabilitative improvements with the RE device may include a gait training environment that is more symmetrical and may improve weight transfer and overall gait speed. Further research with a larger sample with various level of impairment utilizing an RE for gait training is needed to determine the potential utility of REs as an alternative to traditional gait training.
G.J. Androwis K.J. Nolan (&) Kessler Foundation Human Performance and Engineering Research, West Orange, NJ, USA e-mail:
[email protected] K.J. Nolan Department of Physical Medicine and Rehabilitation, Rutgers – New Jersey Medical School, Newark, NJ, USA G.J. Androwis K.J. Nolan Children Specialized Hospital, 150 New Providence Road, Mountainside, NJ, USA G.J. Androwis Department of Biomedical Engineering, New Jersey Institute of Technology, Newark, NJ, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_2
9
10
G.J. Androwis and K.J. Nolan
1 Introduction Stroke is considered one of the leading causes of long-term disability in the United States and every year more than 795,000 individuals are affected by stroke [1]. Individuals post stroke experience mobility limitations including gait impairments and a decrease in functional ambulation [1]. One of the primary goals in stroke rehabilitation is to regain walking function, which can directly impact community participation and quality of life. [2]. Early stage recovery, during the acute phase of rehabilitation, is dependent on the treatment effectiveness and the time frame at which treatments are introduced to patients [3]. The effectiveness of stroke rehabilitation may be based on the role of neuroplasticity and the ability to relearn lost function [3, 4]. The integration of state-of-the-art robotic technologies, such as robotic exoskeletons (RE), for gait rehabilitation of individuals with neuromuscular disabilities is essential during acute stroke rehabilitation, when plasticity level is high, to maximize functional benefits and long term outcomes. The utility of RE devices has been rapidly expanding over the past decade [5]. The demand for utilization of RE for robotic therapy has also been growing side by side with the development of new and more advanced RE devices [6, 7]. Wearable RE for gait training are driven by high torque motors anatomically aligned at the hip and knee joints. The RE combines the power needed during gait therapy for ambulation with human input (through motion sensors and force/loading sensors) to determine RE performance and the way that the wearer interacts with the device [8, 9]. The purpose of this investigation is to determine the impact of a robotic exoskeleton (RE) on temporal spatial gait parameters as compared to traditional standard of care gait training in a single inpatient after acute stroke.
2 Materials and Methods 2.1
Participants
One female participant with acute stroke and left unilateral hemiparesis (Age 30; Height 1.63 m; Weight 87.75 kg) was recruited for RE gait training in conjunction with standard of care therapy in an inpatient rehabilitation hospital. Data was collected during a single gait training session while participants ambulated with and without the RE over level surfaces with the assistance of a physical therapist. Participant inclusion requirements: medical clearance, upright standing tolerance ( 15 min), intact skin, and physically fit into the device. All procedures performed in this investigation were approved by the Human Subjects Review Board and informed consent was obtained prior to study participation.
Evaluation of a Robotic Exoskeleton for Gait Training in Acute Stroke
11
Fig. 1 Participant using the RE device, Ekso-GT, for gait rehabilitation therapy
2.2
Robotic Exoskeleton (RE) Device
Robotic gait therapy was administered via a commercially available robotic device (Ekso-GT, Ekso Bionics, Inc. Richmond, CA, USA), Fig. 1. Patient’s lower extremities were driven in a predefined repetitive trajectory to facilitate over ground walking. The RE used in this investigation includes two motorized joints (hip and knee) and passively adjustable stiffness (sagittal plane) at the ankle joint. The RE was operated in combination with a single point straight cane.
2.3
Experimental Procedure and Data Analysis
Temporal Spatial parameters were measured using gait analysis, all kinematic data was collected at 60 Hz using a 12 camera 3-D motion analysis system (Motion Analysis Corporation, Santa Rosa, CA, USA). The participant was directed to ambulate at self-selected speed over level ground with and without the RE with the assistance of a physical therapist. All data was exported to Matlab (MATLAB R2015B, MathWorks Inc., Natick, MA, USA) for custom analysis. Kinematic data were filtered using a Butterworth filter (zero-lag, 4th order; cut-off frequencies of 6 Hz). Kinematic data were normalized to 100 % of a gait cycle (GC) defined as foot strike to the ipsilateral foot strike; each GC was then sectioned into four phases: Initial double support (IDS), single support (SS), terminal double support (TDS) and swing (SW). Timing of temporal and spatial parameters were then calculated based on the phasic segmentation of GC.
12
G.J. Androwis and K.J. Nolan
3 Results Temporal spatial outcome variables are summarized in Table 1. During gait training in the RE, walking speed increased by 0.05 m/s, and stride length increased bilaterally, while step width decreased on the affected and unaffected side, Table 1. Step length was still variable between affected and unaffected side but was more symmetrical during gait training in the RE. During gait training in the RE, total double support time (IDS + TDS seconds) decreased, and single support and swing time increased on the affected and unaffected limb, Fig. 2. Total stance time (IDS + SS + TDS) remained relatively unchanged in the RE.
Table 1 Temporal spatial parameters of participant’s affected and unaffected side with and without the RE Average
Affected with RE
Affected without RE
Unaffected with RE
Unaffected without RE
Step length (mm) Step width (mm) Stride length (mm) Gait speed (m/s)
391.41
158.37
277.06
326.43
162.98
205.05
165.42
200.53
660.63
480.14
660.49
460.24
0.29
0.26
0.30
0.25
Fig. 2 Timing of affected and unaffected limbs with and without RE
Evaluation of a Robotic Exoskeleton for Gait Training in Acute Stroke
13
4 Discussion The objective of the RE is to provide a therapeutic environment with high step dosing, while providing quality gait retraining. The presented temporal spatial data begins to quantify the temporal spatial differences between gait training in an RE and physical therapist guided gait retraining. In this investigation gait training in the RE had a positive effect on the patients overall gait which included improved temporal spatial parameters and gait speed. Positive changes in temporal spatial parameters were evident on the affected and unaffected limb. Preliminary rehabilitative improvements with the RE device may include a gait training environment that is more symmetrical and may improve weight transfer and overall gait speed. Improvements in gait quality when utilizing the RE were not limited to the affected side. There was a decreased dependency of the unaffected limb during gait training in the RE. Reduced stride width was demonstrated with the RE, which may have an additional rehabilitative affect by reducing the lateral acceleration and deceleration and allowing for a more energy efficient gait pattern (narrower step). Further research with a larger sample with various level of impairment utilizing an RE for gait training is needed to determine the potential utility of REs as an alternative to traditional gait training.
References 1. United States Department of Health and Human Services, “Underlying Cause of Death 1999– 2010” on CDC WONDER Online Database, released 2012, ed. 2. Patterson, S.L., Rodgers, M.M., Macko, R.F., Forrester, L.W.: Effect of treadmill exercise training on spatial and temporal gait parameters in subjects with chronic stroke: a preliminary report. J. Rehabil. Res. Dev. 45, 221 (2008) 3. Jongbloed, L.: Prediction of function after stroke: a critical review. Stroke 17, 765–776 (1986) 4. Levin, M.F., Kleim, J.A., Wolf, S.L.: What do motor “recovery” and “compensation” mean in patients following stroke? Neurorehabilitation Neural Repair (2008) 5. Chen, B., Ma, H., Qin, L.-Y., Gao, F., Chan, K.-M., Law, S.-W., et al.: Recent developments and challenges of lower extremity exoskeletons. J. Orthop. Transl. 5, 26–37 (2016) 6. Agrawal, A., Banala, S.K., Agrawal, S.K., Binder-Macleod, S.A.: Design of a two degree-of-freedom ankle-foot orthosis for robotic rehabilitation. In: 9th International Conference on Rehabilitation Robotics, ICORR 2005, pp. 41–44 (2005) 7. Zoss, A.B., Kazerooni, H., Chu, A.: Biomechanical design of the Berkeley lower extremity exoskeleton (BLEEX). IEEE/ASME Trans. Mechatron. 11, 128–138 (2006) 8. Guizzo, E., Deyle, T.: Robotics trends for 2012. IEEE Robot. Autom. Mag. 19, 119–123 (2012) 9. Tsukahara, A., Kawanishi, R., Hasegawa, Y., Sankai, Y.: Sit-to-stand and stand-to-sit transfer support for complete paraplegic patients with robot suit HAL. Adv. Robot. 24, 1615–1638 (2010)
Wearable Exoskeleton Assisted Rehabilitation in Multiple Sclerosis: Feasibility and Experience Shuo-Hsiu Chang, Marcie Kern, Taimoor Afzal, Shih-Chiao Tseng, John Lincoln and Gerard Francisco
Abstract Impaired ability to walk independently is a significant consequence of multiple sclerosis (MS) resulting in substantial limitation in mobility and performance of daily activities, thus restricting full participation and home and community re-integration. Gait training/restoration in MS is necessary but often limited due to the progress and severity of the disease and limitations of traditional strengthening exercises Much attention has been emphasized in developing pharmaceutical treatment to stop or delay the progress of the disease, but not in developing rehabilitation strategies to improve quality of life and facilitate community re-integration for this population. Recently, wearable lower extremity robotic exoskeletons have been developed to restore ambulation in paralyzed or weak individuals with spinal cord injuries. Utilizing exoskeleton in MS rehabilitation may provident difference users experience and improve rehabilitation outcomes. The purpose of this paper is to provide preliminary results and early experience in our MS exoskeleton assisted rehabilitation program.
The study is partly supported by National Multiple Sclerosis Society (PI: Chang), TIRR Foundation, Memorial Hermann Foundation, and the University of Texas Health Science Center at Houston (UTHealth) Physical Medicine and Rehabilitation Departmental Funds. S.-H. Chang M. Kern T. Afzal G. Francisco NeuroRecovery Research Center at TIRR Memorial Hermann, Houston, TX, USA S.-H. Chang (&) T. Afzal G. Francisco Department of Physical Medicine and Rehabilitation, UTHealth, Houston, TX, USA e-mail:
[email protected] S.-C. Tseng School of Physical Therapy, Texas Woman’s University Houston, Houston, TX, USA J. Lincoln Department of Neurology, UTHealth, Houston, TX, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_3
15
16
S.-H. Chang et al.
1 Introduction Multiple Sclerosis is an autoimmune disease of the central nervous system resulting in progressive deterioration of health in young adults particularly aged 20−40. Significant gait impairment occurs as a result of motor weakness, spasticity, ataxia and sensory disturbance [1]. Many MS patients acquire a disturbing gait pattern and the capacity to walk significantly diminishes over time with one-third patients could not retain walking 20 years after diagnosis. Traditionally, persons with MS undergo various forms of rehabilitation including standard exercises as a therapeutic intervention [2] and treadmill training [3] for maintain and improving the overall health, however, the outcomes are inconclusive. Wearable exoskeletons have been developed as assistive walking device for paralyzed or weak individuals suffered from SCI and stroke. It has become acceptable that wearable exoskeleton may provide opportunity for gait training following task specific training principle. In this study, we explore the feasibility of wearable exoskeleton assisted rehabilitation in MS.
2 Materials and Methods 2.1
Subjects
Four subjects with MS participated wearable exoskeleton assisted training during walking (5 session per week for 3 weeks). Subject characteristics are listed in Table 1. Subjects were recruited if they have diagnosed MS with Kurtzke Expanded Disability Status Scale (EDSS) score between 6−7.5. Subjects with severe and untreated spasticity, joint contracture, unhealed pelvic and limb fracture were excluded.
2.2
Exoskeleton Assisted Training
After screening for eligibility, subject will receive a total of 15 sessions of exoskeleton assisted training. The exoskeleton, Ekso ® (Richmond, CA) was used Table 1 Subject characteristics
ID
Gender
Age
1 Female 32 2 Female 45 3 Female 70 4 Male 52 RRMS: Relapsing-Remitting MS SPMS: Secondary Progressive MS PPMS: Primary Progressive MS
Type
EDSS
RRMS SPMS RRMS PPMS
6 7 6.5 7.5
Wearable Exoskeleton Assisted Rehabilitation in Multiple Sclerosis: Feasibility and Experience17
Fig. 1 Exoskeleton assisted training and assessment sessions. One subject walked with exoskeleton with body weight support system (Vector Gait and Safety System®, Bioness, left). Metabolic and EMG data collection during exoskeleton assisted walking (center and right)
for this study (Fig. 1, left). During exoskeleton assisted walking, exoskeleton can adjust the level/amount of assistance provide to initiate and complete each step. At the beginning of the training, the exoskeleton was fitted according to the subject’s pelvic width and thigh and shank length per manufacture’s setting During training, the subject was trained to walk with Ekso and walker in various tasks: standing balance, slow and fast walking, turning, and sit-to-stand and vice versa. Subject’s perceived exertion was assessed by self-report using the Borg scale of perceived exertion during each training session.
2.3
Outcome Measures
Outcome measures were performed prior to training (pre-training: without exoskeleton) and right after training (post-training: with and without exoskeleton). Functional assessments included six minute walk test (6MWT) and 25 Feet Walk Test (25FWT, 7.62 m) at self-selected speed. The metabolic costs and muscle activation profile during 6MWT and 25FWT were assessed by measuring maximum rate of oxygen consumption (VO2 max, Cosmed k4b2) and EMG (Motion Lab Systems, Baton Rouge, LA), respectively (Fig. 1, center and right). EMG signals were measured and recorded from the right Rectus Femoris, Vastus Medialis, Bicep Femoris, Semitendinosus, Tibialis Anterior, Soleus and Gastrocnemius muscles.
18
2.4
S.-H. Chang et al.
Data Analysis
Values of VO2 max during 6MWT and 25FWT were calculated by averaging the breath-by-breath data. Mean EMG amplitude of each muscle was calculated in one gait cycle (heel contact and next heel contact of the right foot).
3 Results The results were presented in Tables 2 and 3. Subjects walked with exoskeleton during 6MWT has less net oxygen consumption (comparison between walking with and without exoskeleton at post) regardless the total distance. Improved walking distance in 6MWT and walking speed (less time in 25FWT) after training (comparison walking without exoskeleton between pre and post) were also observed (except subject 3).
Table 2 6MWT ID
Net VO2max (ml/kg/min)
W
Pre W/O
Post W/O
W
Mean EMG (µv.) Post W/O W
151.2 129.9 92.0 89.8
5.2 8.3 6.0 5.8
5.3 4.0 5.0 8.6
3.9 3.2 4.3 6.4
16.8 10.2 36.1 59.9
Distance (m) Pre W/O
Post W/O
1 221.3 243.2 2 71.0 75.0 3 99.5 93.9 4 53.0 85.1 W/O: without exoskeleton W: with exoskeleton
14.0 7.9 28.6 49.0
Table 3 25FWT ID
Time (s) Pre W/O
Net VO2max (ml/kg/min) Post W/O
1 11.4 8.1 2 45.2 36.3 3 26.0 31.9 4 68.3 25.9 W/O: without exoskeleton W: with exoskeleton
W
Pre W/O
Post W/O
W
Mean EMG (µv.) Post W/O W
16.1 22.3 32.9 30.4
2.6 3.9 1.9 4.4
1.2 1.1 2.2 3.7
0.8 1.9 2.4 1.8
16.0 8.0 28.4 25.9
16.0 8.0 28.4 25.9
Wearable Exoskeleton Assisted Rehabilitation in Multiple Sclerosis: Feasibility and Experience19
4 Discussion The preliminary results suggest that wearable exoskeleton assisted training is safe and feasible for persons with MS. Safety during wearable exoskeleton assisted training should be emphasized. According to our knowledge, there is no fall prevention strategy developed during exoskeleton assisted walking. We successfully used body weight support system (Fig. 1) for balance and fall prevention during training, in addition to physical assistance from a second staff member for one subject (subject 4) who could not use assistive devices for balance. Our training protocol was designed to train subjects to utilize exoskeleton as assistive device. Therefore, the observed improvement in walking distance and speed were unexpected and could be resulted from increased amount of physical activity. Based on training principles, training protocol that aims to improve walking function should be designed to reach maximal therapeutic effects. Moreover, the success of the training protocol relies the experience of the trainer (therapist) and control software capability. Furthermore, walking with wearable exoskeleton requires learning. The capability of motor control and learning plays critical role in efficient and effective exoskeleton assisted rehabilitation. Therefore, the training protocol should consider the variance among different patient populations, as MS patients suffer fatigue easily and may have impaired skill acquisition and retention capability.
5 Conclusion Wearable exoskeleton assisted training is safe and feasible for persons with MS for gait training and mobility. Individualized training protocols that emphasize either gait training or mobility training should apply. Refining current technologies to design more user friendly device and flexible software control for this population should be considered. Acknowledgment We want to thank participants for their time and efforts.
References 1. Shakespeare, D.T., Boggild, M., Young, C.: Anti-spasticity agents for multiple sclerosis. Cochrane Database Syst. Rev. 4, CD001332 (2003) 2. Briken, S., et al.: Effects of exercise on fitness and cognition in progressive MS: a randomized, controlled pilot trial. Mult. Scler. 20, 382–390 (2014) 3. van den Berg, M., et al.: Treadmill training for individuals with multiple sclerosis: a pilot randomized trial. J. Neurol. Neurosurg. Psychiatry 77, 531–533 (2006)
Exoskeletons for Rehabilitation and Personal Mobility: Creating Clinical Evidence P.T. Arun Jayaraman and William Z. Rymer
Abstract Exoskeletons are devices which have recently gained substantial attention in the field of rehabilitation, especially for use in individuals with neurological disorders (ND). In individuals with ND, mobility limitations and subsequent reduced activity levels contribute to significant reductions in quality of life. Wearable robots or exoskeletons hold much promise to fulfill this unmet need of over-ground mobility and unimpaired gait. However, variable research evidence and clinical efficacy are deterring the impact of these eloquent devices from being introduced into everyday rehabilitation practice. This research summary will highlight some research studies conducted at the Rehabilitation Institute of Chicago (RIC) with various exoskeletons, focusing on their clinical efficacy and personal mobility outcomes.
1 Introduction Each year in the United States, over a million people are affected by strokes, spinal cord injuries and brain injuries [1, 2]. With advances in medical care, most individuals with these neurological injuries or disorders survive their initial injuries and have long-life spans. However, many of these individuals are still affected by the inability to stand and walk, which not only affects their everyday mobility and
Research is supported NIDLRR RERC grants no. 90RE5010-01-01 & 90RE5014-02-00. P.T. Arun Jayaraman (&) W.Z. Rymer (&) Departments of Physical Medicine and Rehabilitation, Northwestern University, Chicago, IL, USA e-mail:
[email protected] W.Z. Rymer e-mail:
[email protected] P.T. Arun Jayaraman W.Z. Rymer Research Program, The Rehabilitation Institute of Chicago, Chicago, IL, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_4
21
22
P.T. Arun Jayaraman and W.Z. Rymer
independence, but also has significant effects on secondary health consequences including, pressure ulcers, spasticity/muscle spasms, muscle atrophy, bone loss, and impaired functioning of their respiratory, digestive, renal, and cardiovascular system. Traditional therapy is successful in improving mobility to a subset of the impaired population, but is still limited in its ability to impact the recovering individual across the continuum of care. Current, insurance reimbursements policies and patient-care models in the United States at least, limit the numbers of therapy sessions a participant can get for his/her condition and thus are left to fend for themselves quite early in their recovery process. Wearable robotic systems or exoskeletons have the potential to provide high-intensity and dosage of gait or walking training across the continuum of care, from acute rehabilitation, to outpatient clinics, to achieving personal mobility in the home and the community. Early studies on exoskeletons have been shown them to be safe and efficacious to achieve over-ground walking in individuals with different neurological disorders struggling with varying levels of walking disability [3–5]. Even though hundreds of exoskeletons currently exist in the research labs, hospitals, and homes, no rigorous studies have been performed in many clinical populations looking at their therapeutic benefit compared to traditional care or quantifying their impact at home and in the community. This research summary will highlight some of the research studies being conducted at the RIC using various exoskeletons, focusing on their clinical efficacy and personal mobility outcomes.
2 Material and Methods 2.1
Patient Populations
The populations we tested in our clinical studies at RIC include, individuals with motor complete and incomplete spinal cord injury, individuals following a stroke, individuals with MS, and children with cerebral palsy. Most study participants suffered chronic neurological impairments and had also received RIC’s traditional physical rehabilitation care.
2.2
Exoskeletons
The exoskeletons can be split to full lower body exoskeletons or modular single joint exoskeletons. The devices which will discussed in this research summary include, powered lower body exoskeletons (powered hips, knees, and passive ankles): Ekso (Ekso Bionics), ReWalk (ReWalk Robotics), Indego (Parker Hannifin); modular robots: SMA-powered hips (Honda), Keeogo-powered knees (B-Temia).
Exoskeletons for Rehabilitation and Personal Mobility
2.3
23
Clinical Studies
The studies discussed in this research summary will include different levels of evidence ranging from case studies to randomized controlled trials. We will discuss the clinical utility of exoskeletons compared to gold standard physical therapy in individuals with spinal cord injury and stroke. We will also look at the impact of wearable exoskeletons as personal mobility devices and their impact on home and community mobility. We will discuss on how to plan clinical research on exoskeletons including, identifying appropriate patient populations, plans timing and dosing strategies, choosing appropriate clinical outcome measures and discuss future implications on modifying existing controller and hardware limitations and safety concerns including falls.
3 Results Large multi-site safety and efficacy clinical trials indicate that exoskeletons are safe for individuals from C7-L5 spinal cord level injuries to use for mobility without any major adverse events. However, most exoskeleton walking speeds currently available are still much slower than average non-injured human’s community walking speeds. Currently, exoskeletons require more than thirty sessions of training for performing advance skills sets such stairs and ramp climbing. Additionally, regular use of exoskeletons in the community has the potential to reduce long term secondary health complications associated with spinal cord injury, thus making a good health-care economics case for reimbursement. On the therapeutic benefits side, full lower body exoskeletons with variable assistance mode are better in improving mobility outcomes in severe stroke participants compared to traditional therapeutic strategies, while modular single joint exoskeletons are able to provide the similar therapeutic benefits as traditional care in mild-to-moderate stroke participants, thus suggesting the scenario that these devices if taken home can provide the benefits of a physical therapy through the continuum care from hospital to the home. Rigid robots still not inclusive of the full physical and cognitive range of individuals with disabilities.
4 Discussion Wearable exoskeletons have come a long way in the last five years following commercialization. They show great promise in moving the field of physical rehabilitation forward by providing the potential for over-ground mobility to a wide- range of NDs. However, limitations of the controller and hardware of current generation exoskeletons results in them being used in a small subsets of the NDs to
24
P.T. Arun Jayaraman and W.Z. Rymer
achieve meaningful therapeutic and mobility benefits. Furthermore, no safety or fall preventions/recovery strategies currently exist in these devices making them one event from major clinical and legal catastrophes. With over forty exoskeletons about to be commercialized over the next five years, these devices are going to become a staple of human performance augmentation in both in non-disabled and disabled populations.
5 Conclusion Wearable exoskeletons when used appropriately can provide superior therapeutic benefits in certain sub-sets of individuals with certain NDs and can also act as eloquent personal mobility devices for large subset of individuals with NDs.
References 1. Post, M.W., et al.: Life satisfaction of persons with spinal cord injury compared to a population group. Scand. J. Rehabil. Med. 30, 23–30 (1998) 2. Stretton, C.M., et al.: Interventions to improve real-world walking after strong: a systematic review and meta-analysis. Clin. Rehabil.,7 April 2016 3. Hartigan, C., et al.: Mobility outcomes following five training sessions with a powered exoskeleton. Top Spinal Cord Inj. Rehabil. 21(2), 93–99 (2015). Spring 4. Spungen, A.M., et al.: Assessment of in-hospital walking velocity and level of assistance in powered exoskeleton in person with a spinal cord injury. Top Spinal Cord Inj. Rehabil. 21(2), 100–109 (2015). Spring 5. Jayaraman, A., et al.: Effects of wearable exoskeleton stride management assist on spatiotemporal gait characteristics in individuals after stroke: a randomized control trial. J. Neural Eng. Rehabil. 12, 69 (2015)
Lower Limb Wearable Systems for Mobility and Rehabilitation Challenges: Clinical Focus Federica Tamburella, Marcella Masciullo, Iolanda Pisotta, Nevio Luigi Tagliamonte and Marco Molinari
Abstract Lower limb exoskeletons (EXOs) may be devised as an ambulation device, as a rehabilitation tool or may be aimed at allowing both objectives. Regarding rehabilitation, it is worth noticing that EXOs also provide a perfect environment for precise assessing of rehabilitation protocols and effects. Different is the case of EXO for mobility, in this area the old wheelchair is still largely winning the challenge. In all functional areas: velocity, safety, portability, acceptance as well as autonomy in the ADL none of today EXOs can compete with the performances of an average wheelchair. EXO usage requires learning, and brain changes associated with tool usage introduce the human in the loop concept, a key aspect of clinical relevance for EXO usage. At present, interesting data on the biological mechanisms and rehabilitation relevance of embodiment are providing hints for guiding rehabilitation. These challenges will be presented from a clinical rehabilitation perspective and expectations and critics discussed.
1 Introduction The idea of using technology systems to replace the function of limbs lost by trauma or diseases is not a prerogative of the modern era: archaeological data have shown that already in ancient Greece and Roman time several hip replacements have been built [1]. Over the past twenty years, the idea of human machine interactions has evolved into one of the fastest-growing areas of neuroscience research [2]. Nowadays, this technology, is well out of science fiction [3], studies, on the possibility of creating EXOs, come from the early ʼ60s. Due to recent developments in robotic technologies, powered EXOs have emerged as both practical devices for assisting individuals with lower limb paralysis F. Tamburella (&) M. Masciullo I. Pisotta N.L. Tagliamonte M. Molinari NeuroRobot Lab and Neurological and Spinal Cord Injury Rehabilitation Department A IRCCS, S. Lucia Foundation, Via Ardeatina 306, 00179 Rome, Italy e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_5
25
26
F. Tamburella et al.
or weakened limbs and as devices to assist and support the work of physical therapists. We can broadly distinguish two main categories: machines that are aimed at rehabilitation, usually used in combination with treadmills and some form of external body weight support, and mobile EXOs, aimed at providing locomotion for people unable to recover legs functionality. Mobile EXOs do not use external body weight support system and usually rely on some external support (crutches, rollator) for balance. As regarding target patient population stroke and trauma patient either with spinal or brain lesion represent the vast majority of the potential users for robotic rehabilitation devices. Unfortunately, while EXO technology continues to advance and psychological as well as physiological benefits of legged mobility are likely, the benefits across multiple physiological systems remain largely unproven.
2 Gait Rehabilitation Within the category of rehabilitation robots, Gait Trainer and the Lokomat, are the most popular and diffused body weight support devices to assist task-oriented, intensive practice of walking, even if based on two different approaches: Lokomat is an EXO actively moving thigh and shank by motors and passively the foot by a spring; Gait Trainer (GT) is and end-effector device [4]. Both devices are used and commercially proposed in several neurological conditions with different functional walking status. In spite of this commercial success, available data on their efficacy are not striking [5]. Effectiveness studies fail to demonstrate the superiority of robotic vs traditional therapy. Many hypotheses have been advanced none of them being conclusive and very few tested. Correctly, selecting patient for the right robotic/EXO is for user a possible approach [6]. A correct approach would be somehow the reverse. To pinpoint the rehabilitation needs of the subject and develop an EXO capable of matching them. Beside the general rules and techniques applied in the so called “traditional rehabilitation” at the base of a successful treatment there is the professional ability of the physiotherapist of modulating the intervention according to the ever-changing needs of the patient. None of the available robotic/EXO devices is at present capable of adapting to the specific conditions of a given subject. Examples of more adaptable and therapy oriented EXOs can be found in the Lopes prototype by Twente university [6], in the BioMot project leaded by CSIC and in the upgrading for the Lokomat proposed in the EU founded BETTER project. BETTER approach combined multiple levels of neural information with the resulting motion data focusing on top-down approaches to develop innovative gait protocols for EXO [7].
Lower Limb Wearable Systems for Mobility and Rehabilitation Challenges
27
3 Gait Substitution In many cases the rehabilitative intervention is directed to the development of autonomy in daily life using the wheelchair, which, nowadays, is the main tool to regain mobility and independence, even if sometimes independence might be hinted by barriers, both physical and psychological [8]. Besides the complexity of the interrelationships between patient and orthosis, aiming at restoring the complete loop including incoming movement related sensory information [1]. Commercial EXOs produced by Ekso Bionics of Richmond, California, Rex Bionics, based in Auckland, New Zealand, and ReWalk EXO Argo Medical Technologies in Israel have already reached the market. On the other hand different labs have developed functioning prototypes. Indego®, by Parker Hannifin, (USA) is not currently available for sale, present a modular design is small and light, with no backpack components. The main limitation of current wearable EXOs developed is that they do not provide stability to the user, as a consequence patients have to rely on external stabilizers such as crutches. Furthermore, existing wearable EXOs are not designed to exploit the remaining, highly patient specific, physiological functions and capacities. Nevertheless the use of EXOs among persons with gait impairments is still extremely limited and almost exclusive confined to supervised environments. Available EXOs allow very little interactions with the user. In general stepping is produced in a stereotyped way through a binary command either from a hand controlled tool or by weight shifting, as for the EXO created within the framework of the European project the MINDWALKER. To allow a broader diffusion, EXOs have to overcome many of the present limitations. Bulkiness, autonomy, velocity, adaptability to all terrains, balance, portability are among the critical aspects that must be answered in the near future, even if development of human machine interaction modules capable of collecting biological signal to drive EXO movement is the real challenge. A balance shared control module will represent an extremely important step for the development of gait shared control modules. These are the main goals of the recently founded EU project Symbitron (http://www.symbitron. eu/). It proposed a patient- instead of a technology-driven methodology in which the design and control of the wearable EXO and the interface between the EXO and patient are customized to the unique features (not only impairments but also remaining capacities) of each individual patient.
4 Clinical Aspects for Gait Rehabilitation It is generally agreed that locomotor training using robot-assisted therapies depends on patient engagement, reward and the desire of the patient to conquer the disability. Thus, wearable robots that engage the patient with gait intent detection can help motivate the patient, provide short-term reward via immediate feedback of
28
F. Tamburella et al.
movement, and empower the patient with direct control of his/her own health and quality of life [9]. Towards this extent, gait intention detection systems for EXO control, such as by monitoring shifts in the center of mass (e.g. tilting the trunk or the pelvis as for the MINDWALKER EXO), may improve the efficacy of rehabilitation by encouraging the user to actively participate rather than relying solely on the assistance provided by the EXO [10]. EXO should be selected according to patients’ features and residual capabilities. Furthermore, due to the continuative use of the EXO it is possible for patient to partially improve some residual activities and to modify brain activities. EXO usage requires learning, and learning is always associated with brain changes at least at synaptic level. Although use of EXOs is focused on ambulation and lower limb motor improvement, there may be secondary benefits across multiple physiological systems, such as improved bowel/bladder function, decreased chronic pain, reduced spasticity, and increased bone marrow density [10].
5 Conclusions For a better acceptance in clinical rehabilitation framework, EXOs have to improve their capacity to adapt to rehabilitation needs allowing interactions with rehab professionals enabling customable evaluation and training sessions. To develop close loop real time control mechanisms based on biological signals capable of adapting EXO performances to the ever-changing needs of patients appears at present the right line for future developments in this field. Acknowledgment This work was supported by EU FP7 Program under contract #611626 (SYMBITRON) and Italian Ministry of Health grant.
References 1. Hernigou, P.: Ambroise pare IV: the early history of artificial limbs (from robotic to prostheses). Int. Orthop. 37(6), 1195–1197 (2013) 2. Shih, J.J., Krusienski, D.J., Wolpaw, J.R.: Brain-computer interfaces in medicine. Mayo Clin. Proc. 87(3), 268–279 (2012) 3. Daly, J.J., Wolpaw, J.R.: Brain-computer interfaces in neurological rehabilitation. Lancet Neurol. 7(11), 1032–1043 (2008) 4. Pohl, M., Werner, C., Holzgraefe, M., Kroczek, G., Mehrholz, J., Wingendorf, I., Hoolig, G., Koch, R., Hesse, S.: Repetitive locomotor training and physiotherapy improve walking and basic activities of daily living after stroke: a single-blind, randomized multicentre trial (DEutsche GAngtrainerStudie, DEGAS). Clin. Rehabil. 21(1), 17–27 (2007) 5. Chang, W.H., Kim, Y.H.: Robot-assisted therapy in stroke rehabilitation. J. Stroke 15(3), 174–181 (2013)
Lower Limb Wearable Systems for Mobility and Rehabilitation Challenges
29
6. Morone, G., Iosa, M., Bragoni, M., De Angelis, D., Venturiero, V., Coiro, P., Riso, R., Pratesi, L., Paolucci, S.: Who may have durable benefit from robotic gait training? A 2-year follow-up randomized controlled trial in patients with subacute stroke. Stroke 43(4), 1140–1142 (2012) 7. Belda-Lois, J.M., Mena-del Horno, S., Bermejo-Bosch, I., Moreno, J.C., Pons, J.L., Farina, D., Iosa, M., Molinari, M., Tamburella, F., Ramos, A.: Rehabilitation of gait after stroke: a review towards a top-down approach. J. Neuroeng. Rehabil. 8, 66 (2011) 8. Simpson, L.A., Eng, J.J., Hsieh, J.T.C., Wolfe and the Spinal Cord Injury Rehabilitation Evidence (SCIRE) Research Team, D.: The health and life priorities of individuals with spinal cord injury: a systematic review. J. Neurotrauma 29(8), 1548–1555 (2012) 9. Contreras-Vidal, J.L., Bhagat N, A., Brantley, J., Cruz-Garza, J.G., He, Y., Manley, Q., Nakagome, S., Nathan, K., Tan, S.H., Zhu, F., Pons, J.L.: Powered EXOs for bipedal locomotion after spinal cord injury. J. Neural Eng. 13(3), 031001 (2016) 10. Zeilig, G., Weingarden, H., Zwecker, M., Dudkiewicz, I., Bloch, A., Esquenazi, A.: Safety and tolerance of the ReWalk exoskeleton suit for ambulation by people with complete spinal cord injury: a pilot study. J. Spinal Cord Med. 35, 96–101 (2012)
Part II
Emerging Technologies in WRs
Impedance Control of Series Elastic Actuators Using Acceleration Feedback Andrea Calanca, Riccardo Muradore and Paolo Fiorini
1 Introduction More and more robots are designed to help or substitute humans both in daily activities and dangerous scenarios. These robots should be able to cope with humans and with other robots and to move in houses, factories, hospitals and uncertain outdoor terrains. To accomplish these targets robots must be able to move safely in unstructured or uncertain environments and to display relatively high forces with high dynamical precision. Currently a widespread solution is that of using impedance controlled elastic joint robots [7]. While impedance control allows to safely deal with unknown kinematic and dynamic environmental constraints [4], series elastic joints allows impedance controllers to be implemented on high force and high power density motors. This is because accurate impedance control is often implemented relying on a explicit inner force loop [2] and series compliance can dramatically improve explicit force control robustness [1, 8]. The physical interaction with humans and surrounding environments is historically considered an hard challenge in robotics. Solutions are often based on passivitybased (PB) control which is an established control paradigm to provide a high level of stability robustness [3]. In particular in the passive interaction framework passivity of the robot impedance is a sufficient condition to ensure a stable interaction with any passive environment [12], and the human has historically been considered as passive [5]. Several passivity-based force and impedance controllers has been introduced for SEA’s. In the seminal works of Pratt and Williamson a passive force controller was introduced, based on positive load acceleration feedback [8]. An interesting aspect of this algorithm is that it not only shows robust stability (passivity) but also robust performance, thanks to the acceleration feedback [1]. It happened that most of subsequent research focused on robust stability while not accounting for robust performance. Vallery et. al. investigated the passivity of SEA control considering a velocity controlled motor. They considered an impedance control schema and A. Calanca (B) · R. Muradore · P. Fiorini University of Verona, Strada Le Grazie 15, Verona, Italy e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_6
33
34
A. Calanca et al.
found that “SEA cannot display a higher pure stiffness than the spring stiffness if passivity is desired”. Interestingly this result has been confirmed also with different control architecture. In [9] the impedance is controlled by shaping the admittance at the motor level, aiming at reducing the force control effort for high impedance rendering. Despite not reported in the literature it can be proved that this algorithm leads to the same passivity result i.e. it cannot passively render a stiffness higher than the physical spring stiffness. A similar outcome emerged for the passive impedance control of the the DLR lightweight arm where a parallel force-position architecture has been adopted. Authors implicitly showed that the passivity constraint doesn’t allow to overcame the physical transmission stiffness. Further related results state that the physical stiffness limit is even hard to reach. In [10] it is shown that only by using high force control gains the physical stiffness can be approached while in [6] it has been shown that in the case of non collocated force sensor the stiffness limit can be significantly lower than the physical spring stiffness. Also in [11] it has been shown that an impedance controlled SEA cannot passively display a virtual Vogit model dynamics, i.e. the dynamics of a spring in parallel to a damper. In conclusion given the current state of the art there exists no passive algorithm that allows to overcame the physical spring stiffness. Moreover existing passive algorithms cannot display any virtual damping in parallel to the virtual spring. In this paper we present the first passive algorithm that can overcame such stiffness and damping limitations. The idea is inspired by the seminal work of Pratt and Williamson [8] where they used the above mentioned load acceleration feedback to control the SEA output force. Such acceleration feedback allows the motor to have the same acceleration of the environment thus compensating for load motion and leading to robust performance, i.e. predictable error dynamics. Quite surprising such a performance robustness was not highlighted neither in the original work nor in subsequent publications which focused more on the robust stability issue rather than on robust performance. In fact the effect of the acceleration feedback is to cancel out the load dynamics and we will formally show that leveraging on such load dynamics cancellation we can passively render any passive impedance. Fig. 1 A model of SEA interacting with an environment
Human τm
k
τs
Jm θ
q
Impedance Control of Series Elastic Actuators … Fig. 2 Impedance control of a SEA. It uses an inner force control loop and an outer position loop to shape the impedance
qr
sI(s)
35
1 Js2
C(s) ua
θ
k
τe E q
2 Impedance Control of Series Elastic Actuators SEA’s are usually modeled using the following equations τs = k(θ − q) Jm θ¨ = τm − τs
(1) (2)
where, considering the scenario in Fig. 1, τs is the spring force (or torque in the case of a rotary joint), θ is the motor position, q is the environment position and τm is the motor input force (or torque). The actuator parameters are k and Jm which represent the spring stiffness and the motor inertia respectively. Let us highlight that τm is the actuator input while τs is actuator output i.e. the force (or torque) exerted on the environment. An impedance control architecture applied to a SEA is shown in Fig. 2, where u a = 0. The force controller F(s) uses the spring deflection feedback and is fed by an outer loop that measures the load position and computes the force reference needed to obtain the desired impedance. As example if we desire a first order impedance, the outer controller should be implemented as s I (s) = dd s + kd
(3)
where dd and kd are desired damping and stiffness.
3 Impedance Control Using Acceleration Feedback In this section we analyze the impedance shaping of a SEA considering the control structure in Fig. 2 where C(s) is a generic stable linear controller, I (s) implements a desired passive impedance and u a is an auxiliary control output or motor input. This impedance relation can be computed using standard block algebra as −τs =
C(s)I (s) + J s 2 − u a q. C(s) + kJ s 2 + 1
If we initially neglect u a it turns out that the desired impedance relation is approximately satisfied only if C(s) J s 2 and C(s) kJ s 2 + 1. If this is not the case the actual impedance can be quite far from the desired impedance I (s) an the impedance
36
A. Calanca et al.
error is influenced by the environment motion. Interestingly, if we consider the following expression for u a u a (t) =
J τ¨s (t) + τs (t) + J q(t) ¨ k
which includes a positive acceleration feedback, the actual impedance results exactly equal to the desired impedance: −τs = I (s)q. This means accurate and passive rendering of any passive impedance.
4 Conclusions The proposed acceleration-based law suggests that embedding the load/environment acceleration in the control law can help to design accurate impedance controllers. Unfortunately the acceleration signal is usually a noisy measure or an approximate estimation. Moreover acceleration overestimation can lead to feedback inversion and instability [8]. For these reasons an acceleration measure or estimation needs to be handled carefully. Also common acceleration processing, such as filtering and amplification, may alter control passivity. In our preliminary experimental results we achieved stable rendering of 3k, where k is the physical spring stiffness. In our future work we will analyze the passivity of existing impedance controllers and compare them to the acceleration-based approach.
References 1. Calanca, A., Fiorini, P.: On the role of compliance in force control. In: Menegatti, E., Michael, N., Berns, K., Yamaguchi, H. (eds.) International Conference on Intelligent Autonomous Systems IAS-13, Padova, Italy. Springer International Publishing (2014) 2. Calanca, A., Muradore, R., Fiorini, P.: A review of algorithms for compliant control of stiff and fixed compliance robots. IEEE Trans. Mechatron. 21(2), 613–624 (2016) 3. Colgate, E., Hogan, N.: An analysis of contact instability in terms of passive physical equivalents. In: IEEE International Conference on Robotics and Automation, vol. 1, pp. 404–409 (1989) 4. Hogan, N.: Impedance control: an approach to manipulation: Part I, II, III. J. Dyn. Syst. Meas. Contr. 107, 1–24 (1985) 5. Hogan, N.: Controlling impedance at the man/machine interface. In: International Conference on Robotics and Automation, Scottsdale, AZ, pp. 1626–1631. IEEE Computer Society Press (1989) 6. Oblak, J., Matja, Z.: On stability and passivity of haptic devices characterized by a series elastic actuation and considerable end-point mass. In: IEEE International Conference on Rehabilitation Robotics, vol. 1, pp. 1–5. ETH Zurich Science City (2011)
Impedance Control of Series Elastic Actuators …
37
7. Paine, N., Mehling, J.S., Holley, J., Radford, N.A., Johnson, G., Fok, C.-L., Sentis, L.: Actuator control for the NASA-JSC Valkyrie humanoid robot: a decoupled dynamics approach for torque control of series elastic robots. J. Field Robot. 32, 378–396 (2015) 8. Pratt, G.A., Williamson, M.M..: Series elastic actuators. In: International Conference on Intelligent Robots and Systems, vol. 1, pp. 399–406. IEEE (1995) 9. Pratt, G.A., Willisson, P., Bolton, C., Hofman, A.: Late motor processing in low-impedance robots: impedance control of series-elastic actuators. In: American Control Conference, pp. 3245–3251 (2004) 10. Tagliamonte, N.L., Accoto, D.: Passivity constraints for the impedance control of series elastic actuators. J. Syst. Control Eng. 228(3), 138–153 (2013) 11. Tagliamonte, N.L., Sergi, F., Accoto, D., Carpino, G., Guglielmelli, E.: Double actuation architectures for rendering variable impedance incompliant robots: a review. Mechatronics 22(8), 1187–1203 (2012) 12. Van Der Schaft, A.J.: L2-Gain and Passivity Techniques in Nonlinear Control. Lecture Notes in Control and Information Sciences, vol. 218. Springer (1996)
Kinetic Energy Recovery in Human Joints: The Flywheel-Infinitely Variable Transmission Actuator Roberta Alò, Francesco Bottiglione and Giacomo Mantriota
Abstract In this extended abstract we present a novel architecture of an electric-mechanical actuator suitable for application to systems with periodic motion and here presented as an alternative to traditional actuators employed in wearable robots for lower limbs: the Flywheel - Infinitely Variable Transmission (F-IVT).
1 Introduction Wearable robots for lower limbs aspire to become portable devices that will help disabled people in every day life [1]. Unfortunately, a lot of work is still to be done in order to achieve this goal and to fulfill the related requirements of portability and reliability. Most of the wearable robots presently under development are electrically powered and their less portable components are the actuators and the batteries. The characteristics of lower limbs movements in terms of velocities and forces/torques are not the best ones for electric drives. Indeed, the motor size must be selected in function of the peak torque and power requested that are much larger than the average torque and power, thus leading also to a loss of efficiency. This leads to oversize and to overweight the batteries and the actuators with respect to their actual usage. Here we present a novel architecture named F-IVT (Flywheel-Infinitely Variable Transmission) actuator that in principle permits to exploit the characteristics of the
R. Alò F. Bottiglione (&) G. Mantriota Department of Mechanics Mathematics and Management, Politecnico Di Bari, Bari, Italy e-mail:
[email protected] R. Alò e-mail:
[email protected] G. Mantriota e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_7
39
40
R. Alò et al.
lower limb movement optimizing the operating conditions of the electric motor and overcoming the aforementioned problems [2, 3].
2 The F-IVT Actuator: Working Principle and Performance Calculation 2.1
Working Principle of F-IVT
The idea at the basis of the F-IVT actuator comes from the evidence that, in walking, the lower limbs joints (the knee in particular) sometimes need power for propulsion, sometimes need to be braked, in a periodic fashion. For this reason the average power per gait is much smaller than the peak power (positive or negative) requested, being even negative in the case of the knee joint in normal walking: in principle it is possible to harvest energy from the knee reducing the metabolic cost associated with braking the joint [4]. Unfortunately, the back drivability of commonly adopted mechanical drives is hard to achieve and the electric energy recovery has been shown to be inefficient [6]. Moreover, the electric machine should be sized in function of the peak value of the torque and power requested in a gait. The working principle of the F-IVT is the following: a mechanical energy accumulator (the flywheel) is placed between the actuator and the transmission and coupled directly with the motor. The motor angular velocity is close to its optimal value and it is almost constant thanks to the flywheel. The transmission must adapt the velocity of the flywheel to the changeable angular velocity requested by the knee. In order to do so, it should be characterized by a small and variable speed ratio. In practical applications, it would be made of two devices in series: a fixed ratio drive and a continuously variable transmission. In particular, the transmission will be of infinitely variable transmission type (IVT) in order to permit also the change of the rotation wise of the output shaft. The fixed ratio drive is a harmonic drive (HD) or a ball screw (BS), depending on the attachment to the joint/leg. In the former case, the F-IVT is of rotating type, in the latter case it is of linear type (Fig. 1). For one given walking regime and in ideal conditions, (negligible mechanical power losses and flywheel with infinitely large inertia), the variable power request of the knee would be filtered by the flywheel, and the motor should only feed the system with the average power needed to walk, applying also an almost constant torque to the flywheel, thus working on a fixed operating point at optimal electric efficiency. Thus, since average power and torque are much smaller than their peak values the motor can be greatly undersized.
Kinetic Energy Recovery in Human Joints
41
Fig. 1 Schematic picture of the F-IVT architectures
2.2
Performance Calculation
The actual performance of F-IVT actuator is different from the ideal because of the finite size of the flywheel and the power loss in the transmission. In order to estimate the potential benefits in terms of size, weight and energy efficiency of the F-IVT with respect to existing actuators for similar applications, these aspects must be taken into account. We followed a back-dynamic approach starting from the torque and angular velocity requirements of the knee joint, available in the literature [5, 6]. We used kinematics and efficiency models of mechanical components (constant ratio drive, infinitely variable transmissions, motor) [2, 3] to develop a simulation tool for the calculation of the energy required per gait by the actuator. The IVT is characterized by a variable efficiency with respect to speed ratio, torque and power flow direction, and thus accurate models of IVT efficiency are adopted [7]. Also the harmonic drive is modeled in detail for similar reasons. The selection of the motor, the flywheel inertia, the IVT, the fixed ratio drive sizes, and of the location of the attachment points to the leg (for linear F-IVT) has been done by minimizing the energy required per gait (objective function) with some additional constraints mostly related to the strength of the mechanical devices as declared by manufacturers. After the selection of the components has been done, the simulation tool is used to calculate the achievable performance of the actuator. Following the outlined approach, we also calculated the performance of a traditional direct drive actuator (DD) for comparison purposes.
42
R. Alò et al.
3 Results and Discussion Figure 2 shows the instant values of electric power needed to achieve the torque/velocity request of a gait cycle (biological walking at 1.1 m/s) with rotating F-IVT, linear F-IVT and a Direct Drive (DD) actuators, optimized for this working condition. It can be noted that both flywheel based actuators permit the motor to deliver constant power whereas a direct drive actuator follows point-by-point the power requested by the joint. In our calculations we consider that negative power is dissipated (electric non-regenerative braking) as it (almost) always happens in these applications. Moreover, it can be noted that the power consumption of linear F-IVT is less than the one achieved by the rotating F-IVT. The reason is that the efficiency of the HD is by far less then the efficiency of the Ball Screw (BS), and the HD is hardly back drivable, thus limiting the possibility to recover energy in the braking phases. In Table 1 it is shown that linear F-IVT in principle permits the gait cycle to be completed without any energy consumption (as mentioned in the introduction)
Fig. 2 Instant values of electric power in normal walking at 1.1 m/s powered by rotating F-IVT, linear F-IVT and Direct Drive (DD) actuators
Table 1 Simulation results: knee joint v [m/s] slope [°] 1.1 0° [-] 30°
E [J/kg] PP [W/kg] PT [Nm/kg] E [J/kg] PP [W/kg] PT [Nm/kg]
Rotating F-IVT
Linear F-IVT
Direct Drive
0.24 0.32 3.4 10−4 3.02 2.43 2.2 10−3
0 0 6.9 10−6 1.25 0.93 7.3 10−4
0.74 5.69 7.2 10−3 2.10 5.03 1.9 10−3
Kinetic Energy Recovery in Human Joints
43
whereas the rotating F-IVT can not. Data shown in Table 1 also suggest that the motor size of flywheel-based drives is by far smaller than the Direct Drive (the peak of torque is at least one order of magnitude smaller than for the Direct Drive actuator) even in case of stairs ascend, where the energy requirements of the actuators increases so as to make the DD more convenient than the rotating F-IVT.
4 Conclusion We presented the architecture of the F-IVT actuators and some model based estimations of the achievable performance in wearable robots for lower limbs. It has been shown that the linear F-IVT actuator may lead to improvements of actuators portability because of the reduced energy consumption with respect to traditional devices. In terms of weight and size, the F-IVT actuators permits to undersize the motor, but they also need an additional micro-IVT transmission that is still under development. Feasibility of efficient micro-IVT transmissions will be actually crucial to determine whether the F-IVT has a future in bio-robotics.
References 1. Pons, J.L.: Wearable Robots: Biomechatronic Exoskeletons. John Wiley & Sons Ltd, Chichester (2008) 2. Alò, R., Bottiglione, F., Mantriota, G.: An innovative design of artificial knee joint actuator with energy recovery capabilities. ASME J. Mech. Robot. 8(1), 011009 (2016) 3. Alò, R., Bottiglione, F., Mantriota, G.: Artificial knee joints actuators with energy recovery capabilities: a comparison of performance. J. Robot. 2016 (2016) 4. Donelan, J.M., Li, Q., Naing, V., Hoffer, J.A., Weber, D.J., Kuo, A.D.: Biomechanical energy harvesting: generating electricity during walking with minimal user effort. Science 319(5864), 807–810 (2008) 5. Grimmer, M., Eslamy, M., Seyfarth, A.: Energetic and peak power advantages of series elastic actuators in an actuated prosthetic leg for walking and running. Actuators 3(1), 1–19 (2014) 6. Riener, R., Rabuffetti, M., Frigo, C.: Stair ascent and descent at different inclinations. Gait & Posture 15(1), 32–44 (2012) 7. Bottiglione, F., Mantriota, G.: Reversibility of power-split transmissions. ASME J. Mech. Des. 133(8), 084503 (2011)
A Compliant Lightweight and Adaptable Active Ankle Foot Orthosis for Robotic Rehabilitation Marta Moltedo, Tomislav Bacek, Kevin Langlois, Karen Junius, Bram Vanderborght and Dirk Lefeber
Abstract This paper presents the design of a compliant, lightweight and adaptable active ankle foot orthosis (AAFO) and preliminary test of its ankle actuator. The ankle actuator is designed to keep its weight as low as possible. The adaptability of the AAFO allows adjusting the device to different patients, without the need of customized versions.
1 Introduction During the push-off phase of walking, the ankle joint alone generates nearly all the positive work of the stance leg. In this phase it provides support to the body, ensures forward progression and initiates the swing of the leg. However, the capability of the muscles acting at the ankle joint can be negative altered by several neuromuscular This work is supported by the European Commission’s 7th Framework Program as part of the project BioMot (Grant Agreement number IFP7-ICT-2013-10-611695), by the Flemish agency for Innovation by Science and Technology as part of the project MIRAD (IWT-SBO 120057) and by the FWO grant (no. G026214N). M. Moltedo (B) · T. Bacek · K. Langlois · K. Junius · B. Vanderborght · D. Lefeber Department of Mechanical Engineering, Robotics & Multibody Mechanics Research Group, Vrije Universiteit Brussel, 1040 Brussel, Belgium e-mail:
[email protected] T. Bacek e-mail:
[email protected] K. Langlois e-mail:
[email protected] K. Junius e-mail:
[email protected] B. Vanderborght e-mail:
[email protected] D. Lefeber e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_8
45
46
M. Moltedo et al.
disorders. Affected patients present impaired walking patterns, often represented by lack of propulsion, foot drop and toe drag [1]. In these cases, the use of AAFOs as rehabilitation devices can greatly enhance the recovery of patients’ motor capabilities. To be suitable for rehabilitation purposes, an AAFO has to be safe for the user, wearable (meaning lightweight, compact and autonomous) and robust to provide the necessary torque to the user. In addition, modularity and adaptability to different subjects are preferable characteristics, to avoid the need of building customized versions of the same prototype. Several AAFOs to assist the human ankle joint during walking have been built [2], but none of them satisfies all the aforementioned characteristics. In this paper, the design of a compliant, lightweight and adaptable AAFO and the results of a preliminary characterization test of its ankle actuator are presented. A more detailed description of the design of the AAFO can be found in [3]. The results of the characterization tests performed on the presented actuator will be used as feedback to design a second prototype providing a higher level of assistance.
2 Mechanical Design of the AAFO 2.1 Ankle Actuator The AAFO presented in this paper is shown in Fig. 1. The ankle actuator implemented in the AAFO can provide a peak torque of 25 Nm and it is based on the so-called spindle-driven MACCEPA. The MACCEPA is a torque-controlled, variable stiffness actuator that basically works as a torsion spring and allows independent control of its equilibrium position and joint stiffness. The working principle of the MACCEPA and the design of its spindle-driven version to actuate a joint of an exoskeleton can be found respectively in [4, 5]. As said before, wearability is one of the desired characteristics for an AAFO. For this reason, the design of the presented actuator is made to be as lightweight and compact as possible. A reduction of size and weight of the actuator is obtained by inserting the MACCEPA spring into the lever arm. This realization allows reducing the dimensions of the foot link (Fig. 1). A further reduction of weight is achieved by removing unnecessary material for the structural integrity from the actuator’s components. This process was based on the analysis of the distribution of the stresses and deformations obtained using a Finite Element Method (FEM) analysis. Thanks to this approach, the total weight of the ankle actuator (including electronics and excluding batteries) is only 1.1 kg.
A Compliant Lightweight and Adaptable Active Ankle Foot Orthosis ...
47
Fig. 1 The presented AAFO is lightweight (the total weight of the AAFO structure and the ankle actuator is 1.7 kg, including electronics and excluding batteries), adaptable to different patients and modular (it can be used as the ankle joint of an exoskeleton). The ankle actuator is a compliant actuator based on the MACCEPA [4] providing a peak torque of 25 Nm
2.2 Connections to the User Adaptability of the AAFO to different subjects is achieved by means of the design of the connections between the ankle actuator and the user. The foot of the user is attached to two carbon fibers plates, respectively placed below the heel and the toes of the foot. Each of them is connected to the ankle actuator by means of an aluminum link (heel link and toe link), whose relative position can be changed in the sagittal plane. Another advantage of having movable connecting links is the possibility of aligning the robot’s and user’s ankle joint axes. This alignment is important to minimize unwanted residual forces on the user when the actuator is active. To allow the biological rotation of the foot around the toe joint while walking, the toe plate is connected to the toe link by means of a revolute joint. The shank of the user is attached by means of two cuffs to a carbon fibers shank frame, which is attached to the shank link of the actuator through a hinged joint. The combination of a carbon fiber frame and a hinge joint, allows the shank frame to bend in the frontal plane, to follow the curved shape of the user’s leg, while the actuator is kept perpendicular to the ground. The vertical positions of the shank frame and of the two cuffs can be varied with respect to the ankle joint axis, to adapt the AAFO to different shank lengths.
48
M. Moltedo et al.
The shank frame and foot plates are made more rigid in the sagittal plane and more compliant in the frontal plane. In this way, in the sagittal plane, their necessary stiffness to transfer the actuator torque is not compromised, but they are still able to bend to follow the movement of the user in the frontal and transversal planes, allowing a more comfortable walking pattern.
3 Ankle Actuator Characterization A static characterization test of the actuator was performed. In this test, both links of the actuator were rigidly fixed to the experimental setup. The actuator was fixed to the setup only at its internal side, meaning the side that is attached to the user’s leg, to simulate the connection to the AAFO frame. The fixed link of the actuator (Fig. 1) was directly fixed to the experimental setup. The output link (Fig. 1) was connected to the setup via a force sensor, used to measure the output torque provided by the actuator. During the experiment, the lever arm was controlled by means of a PI controller which used the motor velocity signal as control signal. The input to the controller was an error signal computed as the difference between the desired and actual torque. The actual torque was computed multiplying the measured force in the load cell and the length of the output link. The desired output torque was defined to be a sinusoidal curve having 25 Nm as amplitude.
Fig. 2 Desired predefined output torque (in red) and actual output torque (in blue) obtained during the static characterization experiment, in which both links of the actuator were rigidly fixed to the experimental setup
A Compliant Lightweight and Adaptable Active Ankle Foot Orthosis ...
49
Figure 2 shows the predefined reference signal (in red) and the actual torque obtained during the experiment (in blue). As it can be seen, the actuator successfully followed the reference trajectory. Small errors at the lowest amplitude, small oscillations and a constant delay in the actuator trajectory come from the gains of the PI controller, which were not set to be optimal for the given scenario.
4 Conclusion This paper presented the design of a compliant, lightweight and adaptable AAFO. The design of the connections between the ankle actuator and the foot and shank of the user is made to be adaptable to different subjects. The results of a preliminary characterization test of the ankle actuator showed that the ankle actuator is able to provide 25 Nm as peak torque. Starting from the design of the presented actuator, a second prototype will be designed and built. In the second prototype, the conceptual design and working principle of the actuator will remain the same, but the actuator will be upscaled to provide 50 Nm as peak torque, while maintaining the weight and size of the AAFO as low as possible. Acknowledgments The authors would like to thank MOBILAB @ Thomas More for the development of the carbon fiber foot plates.
References 1. Park, Y.-L., Chen, B.-R., Young, D., Stirling, L., Wood, R.J., Goldfield, E., Nagpal, R.: Bioinspired active soft orthotic device for ankle foot pathologies. In: IEEE/RSJ International Conference on Intelligent Robots and Systems, San Francisco, CA, USA, pp. 4488–4495 (2011) 2. Duerinck, S., Swinnen, E., Beyl, P., Hagman, F., Jonkers, I., Vaes, P., Van Roy, P.: The added value of an actuated ankle-foot orthosis to restore normal gait function in patients with spinal cord injury: a systematic review. J. Rehabil. Med. 44(4), 299–309 (2012) 3. Moltedo, M., Bacek, T., Junius, K., Vanderborght, B., Lefeber, D.: Mechanical design of a lightweight compliant and adaptable active ankle foot orthosis. In: 6th IEEE RAS & EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob 2016) (submitted for publication) 4. Van Ham, R., Vanderborght, B., Van Damme, M., Verrelst, B., Lefeber, D.: MACCEPA, the mechanically adjustable compliance and controllable equilibrium position actuator: design and implementation in a biped robot. Rob. Auton. Syst. 55(10), 761–768 (2007) 5. Bacek, T., Unal, R., Moltedo, M., Junius, K., Cuypers, H., Vanderborght, B., Lefeber, D.: Conceptual design of a novel variable stiffness actuator for use in lower limb exoskeletons. In: IEEE International Conference on Rehabilitation Robotics, Singapore, pp. 583–588 (2015)
A Novel Shoulder Mechanism with a Double Parallelogram Linkage for Upper-Body Exoskeletons Simon Christensen and Shaoping Bai
Abstract The design of an innovative spherical mechanism with three degrees of freedom for a shoulder joint exoskeleton is presented in this paper. The spherical mechanism is designed with a double parallelogram linkage, which connects two revolute joints to implement the motion as a spherical joint, while maintaining the remote center of rotation. The design has several new features compared to the current state-of-the-art: (1) a relative large range of motion free of singularity, (2) high overall stiffness, (3) lightweight and (4) compact, which make it suitable for assistive exoskeletons.
1 Introduction Traditional designs of shoulder exoskeletons use an serial linkage system with 3revolute (3R) joints [1, 2] to generate the spherical motion of the human shoulder joint. A problem with a serial structure is its workspace limit. The user of the exoskeleton can only raise the upper arm a small angle in the frontal plane before the shoulder mechanism collides with his/hers shoulder, neck or head. To avoid this problem, some alternative designs have been proposed. The designs in [1, 2] minimized the effect of these problems by designing their exoskeletons so that the singular configurations and collision problem of the 3R mechanism occur at postures that are less likely for the user to reach. In another approach, reported in [3], one of the links in the 3R mechanism is replaced with a circular guide to further avoid collision with the user. In this paper a novel spherical mechanism using double parallelogram linkages (DPM in short) is presented. The mechanism is featured with a compact The work reported here is supported by the EU AAL Programme and Innovation Fund Denmark S. Christensen (B) · S. Bai Department of Mechanical and Manufacturing Engineering, Aalborg University, 9220 Aalborg, Denmark e-mail:
[email protected] S. Bai e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_9
51
52
S. Christensen and S. Bai
structure, lightweight yet rigid design, and a large range of motion free of singularity. The proposed mechanism is to be used as a novel glenohumeral joint mechanism for an upper-body exoskeleton.
2 Conceptual Design of Novel Shoulder Mechanism for an Upper-Body Exoskeleton The proposed design consists of two revolute joints that are connected together via four links, as shown in Fig. 1. The four links form a double parallelogram mechanism (DPM), which under the given configuration form a remote center of motion mechanism [4]. The revolute joints produce extension/flexion and abduction/adduction respectively, while the DPM produces the internal/external rotation. The working principle is similar to the 3R mechanisms, meaning it can rotate about three independent axes that all coincide in one point, namely, the Remote Center (RC) (see Fig. 1). The proposed design is constructed as a hybrid mechanism. The benefit of the proposed design is that the chance of collision with the user is minimized compared to the classical 3R mechanism having two links in series. Moreover, the structure is more compact, lighter and less complicated compared to the 3R mechanism using a circular guide. The design parameters of the DPM are displayed in Fig. 2, which include four link lengths and two offset angles. The links L 1 and L 2 are the lengths of the first parallelogram and L 3 and L 4 are the lengths of the second parallelogram. For the proposed design, the offset angles are introduced mainly to account for the footprint of the actuators, where φ2 accounts for the actuator at θ1 and φ1 for the actuator at θ3 . The range of motion of the DPM, i.e. θ2 , is limited by the footprint of the two revolute joints and the collision with the human shoulder. Axis of joint 1 is aligned with L1 , while axis of joint 3 is aligned with L2 .
Fig. 1 Design concept of the double parallelogram mechanism
Link 2 Link 1 Joint A Joint B
Joint C Joint F Joint D
Link 3 Link 4 Joint G Joint E
Actuator 2
Actuator 1 Axis 2
Axis 1 Motor Mount 1
Remote Center
Axis 3
Motor Mount 2
A Novel Shoulder Mechanism with a Double Parallelogram Linkage … Table 1 Denavit-Hartenberg parameters of the proposed mechanism
αi−1
53
Link, i
ai−1
di
θi
1 2
0 0 L 3 sin φ2 −90◦
0 0
3
L2
0
0
4
L3
0
0
5
L 2 sin φ1 90◦
θ1 90◦ + φ 1 − θ2 180◦ − φ1 − φ 2 + θ2 90◦ + φ 2 − θ2 θ3
0
3 Kinematic Analysis of the Mechanism The kinematics of the proposed mechanism is formulated based on DenavitHartenbergs convention. Cartesian coordinate frames are attached to each link of a manipulator, as shown in Fig. 2. The corresponding DH parameters can be obtained as listed in Table 1. Using these parameters the rotation matrix is obtained as: ⎡
⎤ cθ1 cθ2 cθ3 − sθ1 sθ3 −cθ3 sθ1 − cθ1 cθ2 sθ3 −cθ1 sθ2 R = ⎣ cθ1 sθ3 + cθ2 cθ3 sθ1 cθ1 cθ3 − cθ2 sθ1 sθ3 −sθ1 sθ2 ⎦ cθ3 sθ2 −sθ2 sθ3 cθ2
(1)
where θ1 , θ2 and θ3 are the joint angles. Also, c and s stands for cos and sin respectively.
Fig. 2 Coordinate frames attached to the links of the shoulder mechanism
F
L3
2
G x3 C
y3
G'
L4
z3 D
y4 E y5 x5 E' z4 x4 z5 θ3
L2 L1 A 1
A'
x2 y2
z2 B x0,x1 z0,z1 B' y0,y1
1 2
θ1
θ2 RC
54
S. Christensen and S. Bai
The inverse kinematics problem is solved for the three joint angles: θ2 = arccos(r33 ) θ1 = arctan 2 (−r23 /sθ2 , −r13 /sθ2 ) θ3 = arctan 2 (−r32 /sθ2 , r31 /sθ2 ) where ri, j stands for the (i,j)th element of the matrix R. It should be noted that there are two possible solutions for θ2 , but given the allowable range of motion only the solution between 0 and 180◦ is used. The velocity and singularity analysis of the DPM can be performed by deriving the Jacobian for the angular velocities: θ˙ = Jω −1 ωe
(2)
T where θ˙ = θ˙1 θ˙2 θ˙3 is a vector with the joint angular velocities, Jω is the Jacobian T and ωe = ωx ω y ωz is the end-effector angular velocities. The Jacobian is found as: ⎡ ⎤ 0 sθ1 −cθ1 sθ2 Jω = ⎣ 0 −cθ1 −sθ1 sθ2 ⎦ (3) 1 0 cθ2 A common measure of the evaluating the performance of a mechanism is the manipulability index μ, which is defined as: μ(J) =
JJT = |sθ2 |
(4)
From Eq. (4) it is clear that the kinematic performance of the DPM only depends on the angle of the double parallelogram. The manipulability index over the range of motion of the parallelogram is shown in Fig. 3, where it is seen that the DPM is at a singular configuration in the case that the joint axes constitute a common plane (θ2 = 0, π ). Due to the two offset angles, the proposed design has a minimum angle of θ2,min = φ1 + φ2 . Thus, the singular configuration of θ2 = 0, is not obtainable. According to [2], a range of motion for shoulder internal/external rotation of 135◦ is sufficient for most of our activities of daily living. Hence, the mechanism is free of singularities and covers the required range of motion if the maximum angle satisfies the following condition θ2,max = θ2,min + 135◦ < 180◦ .
4 Application of the Novel Spherical Shoulder Mechanism The application of the novel shoulder mechanism in an exoskeleton design is illustrated in Fig. 4. The exoskeleton is intended as a part of a portable exoskeleton for elderly to assist them in their daily activities. For the proposed exoskeleton design,
A Novel Shoulder Mechanism with a Double Parallelogram Linkage … Fig. 3 The manipulability index of the DPM within its operation range
55
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
20
40
60
80
100 120 140 160 180
Fig. 4 The novel spherical shoulder mechanism utilized in an upper-body exoskeleton
the two revolute joints are actuated by a Flat DC motor (EC60 from maxon motors) and a Harmonic Drive Gear (CSD-25-2A from Harmonic Drive) each, while the DPM is left passive. As seen from the figure, the novel shoulder mechanism has a compact design without compromising the range of motion.
5 Conclusion In this paper, a novel spherical mechanism using a double parallelogram linkage and two revolute joints for upper-body exoskeletons has been presented. The design has several new features compared to the current state-of-the-art, such as a relative large range of motion free of singularity, high overall stiffness, lightweight and compact. As a result, the design is well suited for portable exoskeletons. A prototype of the novel design is being developed for design validation and experiments.
References 1. Naidu, D., Stopforth, R., Bright, G., Davrajh, S.: A 7 DOF exoskeleton arm: shoulder, elbow, wrist and hand mechanism for assistance to upper limb disabled individuals. In: IEEE AFRICON Conference, pp. 13–15, Sept. 2011
56
S. Christensen and S. Bai
2. Perry, J.C., Rosen, J., Burns, S.: Upper-limb powered exoskeleton design. IEEE/ASME Trans. Mechatron. 12(4), 408–417 (2007) 3. Jung, Y., Bae, J.: Kinematic analysis of a 5 DOF upper-limb exoskeleton with a tilted and vertically translating shoulder joint. In: 2013 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), pp. 1643–1648, July 2013 4. Bai, G., Li, D., Wei, S., Liao, Q.: Kinematics and synthesis of a type of mechanisms with multiple remote centers of motion. J. Mech. Eng. Sci. 228(18), 3430–3440 (2014)
A Soft Robotic Extra-Finger and Arm Support to Recover Grasp Capabilities in Chronic Stroke Patients Irfan Hussain, Gionata Salvietti, Giovanni Spagnoletti, David Cioncoloni, Simone Rossi and Domenico Prattichizzo
Abstract In this paper, we present the combination of the Soft-SixthFinger, a wearable robotic extra-finger designed to be used by chronic stroke patients to compensate for the missing hand function, with a robotic arm that is used as an assistive device to support the patient arm. The extra-finger is a tendon-driven modular structure worn at the paretic forearm. The robotic extra-finger is used jointly with the paretic hand/arm to grasp an object similarly to the two parts of a robotic gripper. The flexion/extension of the robotic finger is controlled by the patient using an Electromyography (EMG) interface embedded in a cap. The robotic arm is controlled to partially compensate for the weight of the paretic arm, while not interfering with the user arm motion. The system has been designed as a tool that can be used by chronic stroke patients to compensate for grasping in many Activities of Daily Living (ADL). We performed a pilot test to demonstrate that the proposed system can significantly improve the performance and the autonomy in ADL.
This work was supported in part by the EU Horizon 2020, project no. 688857 SOFTPRO. I. Hussain · G. Salvietti (B) · G. Spagnoletti · D. Prattichizzo Dipartimento di Ingegneria dell’Informazione, Università degli Studi di Siena, Via Roma 56, 53100 Siena, Italy e-mail:
[email protected] I. Hussain e-mail:
[email protected] G. Salvietti · D. Prattichizzo Department of Advanced Robotics, Istituto Italiano di Tecnologia, Via Morego 30, 16163 Genoa, Italy e-mail:
[email protected] D. Cioncoloni · S. Rossi Dipartimento di Scienze Neurologiche e Neurosensoriali, Azienda Ospedaliera Universitaria Senese, Siena, Italy e-mail:
[email protected] S. Rossi e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_10
57
58
I. Hussain et al.
1 Introduction Long-term disabilities of the upper limb affects millions of stroke survivors [1]. More than 80 % of individuals who experience severe hemiparesis after stroke cannot completely recover hand and arm functionality [2]. The majority of robotic devices have been developed to provide intensive, repetitive and task-specific rehabilitation procedure to patients with mild to severe motor impairments after neurologic injury [3, 4]. However, at the moment, robot-assisted therapy did not significantly improve motor function in patients with long-term upper-limb deficits after stroke, as compared with usual care or intensive therapy [3]. Thus, together with the current efforts in improving rehabilitation technologies, we believe that there is a need of active tools able to recover basic hand and arm functions in chronic stroke patients with a paretic upper limb. In this work, we present a possible solution to recover hand grasp capability based on the combination of an extra-robotic finger, called the Soft-SixthFinger [5], and a robotic arm. The aim of the Soft-SixthFinger is not to assist the paretic hand motion of the patient, but rather to add just what is needed to grasp: an extra thumb [6]. The robotic extra-finger is worn on the user forearm and can accomplish a given task in cooperation with the paretic limb, see Fig. 1. The robotic extra-finger has been designed to guarantee high wearability and portability, with kinematics and actuation inspired by recent works on underactuated compliant robotic hands [7]. In particular, the robotic extra-finger is passively compliant due to its flexible joints. Only one motor is used to control the device flexion/extension with a tendon-driven actuation. The patient can control the motion of the robotic extra-finger through an EMG based interface [8]. Such interface can recognize, through the acquisition of the EMG signal read at the frontalis muscle of the patient, when the patient voluntary
Robotic arm support EMG interface Soft-SixthFinger
Fig. 1 The experimental setup. The Soft-SixthFinger works with the paretic limb to compensate for hand grasp functionality. The patient can control the device thanks to a wearable EMG interface embedded in a cap. The weight of the arm is partially supported by the robotic arm.
A Soft Robotic Extra-Finger and Arm Support to Recover Grasp Capabilities ...
59
Fig. 2 Tasks executed using the Soft-SixthFinger and the robotic arm support. a Blocks and boxes test. b Pouring water on a glass. c Drinking from a small bottle
moves his or her eyebrows upwards. Frontalis muscle contractions generate events that switches the states of a Finite State Machine (FSM) which regulates the finger flexion/extension. The whole system is embedded in a cap. Electrodes can be easily placed on the patient’s forehead just wearing the interface. To test the proposed devices for grasp compensation, we set up a pilot experiment where a subject was asked to perform three different tasks without using the hand: blocks and box test, pouring water from a bottle and drinking from a small bottle, see Fig. 2.
2 The Soft-SixthFinger and Robotic Arm System The Soft-SixthFinger has been designed to be wearable, robust and capable to adapt to different object shapes. The flexible finger has a modular flexible structure attached to a support base that can be worn at the forearm. Each module has a 3D printed ABS polymeric part that acts as a rigid link and a 3D printed thermoplastic polyurethane part that realizes the flexible joint. The CAD of the module as well as details of the extra-finger parts are reported in Fig. 3. The modules are connected by sliding the thermoplastic polyurethane part in the ABS part, see Fig. 3. The Soft-SixthFinger has one servomotor that account for the motion of the whole finger through a tendon. Due to flexible parts in the modules, the device can passively adapt to objects with different shapes. The support base of the finger is realized in ABS. It contains the actuator and an elastic band that allows the user to wear the device on the forearm. The structure of the support base is symmetrical, feature that enables the robotic finger to be worn on both the left or right arm of the patients without any modification in the device. The extra-finger flexion and extension are commanded through the EMG interface described in [5]. The robotic arm used was a Kinova Mico. This arm has a payload of 1.9 kg a workspace of around 900 mm and six degrees of freedom.
60
I. Hussain et al.
Fig. 3 The Soft-SixthFinger exploded view and part details
3 Pilot Study As a pilot study, we tested the system in three different tasks: (a) blocks and boxes, (b) pouring water from a bottle to a glass and (c) drinking from a small bottle. The subject was able to grasp the different objects using the Soft-SixthFinger. The robotic arm was used to partially support the weight of the arm. The simulated paretic limb of the subject was linked to the robotic arm using the two-finger gripper available with the Kinova Mico arm. The robotic arm was controlled in gravity compensation so to assist the arm motion without interfering the subject motion.
4 Conclusion In this extended abstract, we presented our preliminary setup where the SoftSixthFinger is combined with a robotic support for the paretic arm. In our previous works on active tools for manipulation compensation, we focused mostly on the grasping part developing a robotic supernumerary finger that can adapt to different object shapes. However, we noticed that most of the patients testing our devices were still not able to fulfill basic ADL due to the poor mobility of the arm. The system proposed in this work is a first step toward the realization of an assisting platform for chronic stroke patients. We are currently testing the devices with patients also using passive assistive device for the arm support.
A Soft Robotic Extra-Finger and Arm Support to Recover Grasp Capabilities ...
61
References 1. Go, A.S., Mozaffarian, D., Roger, V.L., Benjamin, E.J., Berry, J.D., Blaha, M.J., Dai, S., Ford, E.S., Fox, C.S., Franco, S., et al.: Heart disease and stroke statistics-2014 update: a report from the american heart association. Circulation 129(3), e28 (2014) 2. Nakayama, H., Jorgensen, H.S., Raaschou, H.O., Olsen, T.S.: Compensation in recovery of upper extremity function after stroke: the Copenhagen Stroke Study. Arch. Phys. Med. Rehabil. 75(8), 852–857 (1994) 3. Lo, A.C., Guarino, P.D., Richards, L.G., Haselkorn, J.K., Wittenberg, G.F., Federman, D.G., Ringer, R.J., Wagner, T.H., Krebs, H.I., Volpe, B.T., et al.: Robot-assisted therapy for long-term upper-limb impairment after stroke. N. Engl. J. Med. 362(19), 1772–1783 (2010) 4. Kwakkel, G., Kollen, B.J., Krebs, H.I.: Effects of robot-assisted therapy on upper limb recovery after stroke: a systematic review. Neurorehabil. Neural Repair (2007) 5. Hussan, I., Salvietti, G., Spagnoletti, G., Prattichizzo, D.: The Soft-SixthFinger: a wearable EMG controlled robotic extra-finger for grasp compensation in chronic stroke patients. IEEE Robot. Autom. Lett. (2016) 6. Prattichizzo, D., Malvezzi, M., Hussain, I., Salvietti, G.: The sixth-finger: a modular extra-finger to enhance human hand capabilities. In: Proc. IEEE Int. Symp. in Robot and Human Interactive Communication, Edinburgh, United Kingdom (2014) 7. Odhner, L.U., Jentoft, L.P., Claffee, M.R., Corson, N., Tenzer, Y., Ma, R.R., Buehler, M., Kohout, R., Howe, R.D., Dollar, A.M.: A compliant, underactuated hand for robust manipulation. Int. J. Robot. Res. 33(5), 736–752 (2014) 8. Salvietti, G., Hussain, I., Cioncoloni, D., Taddei, S., Rossi, S., Prattichizzo, D.: Compensating hand function in chronic stroke patients through the robotic sixth finger. Trans. Neural Syst. Rehabil. Eng. (2016)
A Quasi-Passive Knee Exoskeleton to Assist During Descent Emily Rogers, Panagiotis Polygerinos, Stephen Allen, Fausto A. Panizzolo, Conor J. Walsh and Dónal P. Holland
Abstract A pneumatic exoskeleton intended to decrease the muscle activity of the knee extensors during walking on a negative slope is presented. The device consists of an air spring that can be engaged and disengaged via a solenoid valve. When engaged, the air spring resists knee flexion. A preliminary evaluation of the device was conducted with a single healthy subject. During testing, the EMG activity of the rectus femoris decreased by 15 %, while the EMG activity of the vastus medialis increased by 8 %.
1 Introduction This paper presents the design and preliminary evaluation of a quasi-passive knee exoskeleton to assist during walking on negative inclines. During the stance phase of descent, when the knee is in flexion while supporting the weight of the body, the knee extensor muscles contract eccentrically to counteract large moments around the knee [1]. In other words, the knee effectively acts as a torque dampening mechanism during downhill walking. These large extensor forces can lead to muscle fatigue and high stresses on the knee joint in the short term and increased risk of degenerative joint diseases in the long term [2]. Quasi-passive exoskeletons
This work was supported by the Harvard John A. Paulson School of Engineering and Applied Sciences, the Harvard Biodesign Lab, and the Wyss Institute for Biologically Inspired Engineering. E. Rogers S. Allen F.A. Panizzolo C.J. Walsh Harvard John A. Paulson School of Engineering and Applied Sciences and the Wyss Institute for Biologically Inspired Engineering, Harvard University, Cambridge, USA P. Polygerinos Ira A. Fulton Schools of Engineering, Arizona State University, Tempe, USA D.P. Holland (&) School of Mechanical and Materials Engineering, University College Dublin, Dublin, Ireland e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_11
63
64
E. Rogers et al.
have previously been evaluated for assisting the knee during level walking [3], load-carrying [4] and running [5]. To the best of our knowledge, this paper presents one of the first evaluations focused on quasi-passive knee assistance during descent.
2 Materials and Methods The prototype presented here (Fig. 1) consists of three modules: two air springs that provide resistive torques at either knee, a fabric interface to attach the device to the body, and an integrated sensing and control system to detect the user’s gait and control the device accordingly. When heel strike is detected, the corresponding air spring is sealed and acts to resist flexion. When opposite heel strike is detected, the air spring is disengaged so that it does not resist flexion or extension during the late stance and swing phases. The air spring is the most critical component of the exoskeleton. By resisting compression when sealed on one end, the air spring generates a resistive torque about the knee to assist the extensor muscles during downhill walking. A motion capture study was performed to define the test subject’s knee kinematics during descent. The resulting data were combined with a simple model to calculate the air spring dimensions and attachment points required to produce the necessary knee extension moment. A low-friction glass cylinder with a graphite piston was used to create the air spring (Stock Piston and Cylinder Sets #2KS444-CP, Airpot Corp, Norwalk, CT). A protective PVC cylinder was added around the glass cylinder, custom end caps were designed and 3D printed, and a steel rod was connected to an alignment coupler, to ensure straight travel of the rod and piston within the cylinder and reduce undesired kinematic constraints on the user. Two valves at the proximal end of the air spring allow for control of air flow. The first valve is a one-way check valve, that allows air to enter the air spring during extension (swing phase), while preventing air from escaping during flexion. A high-flow check valve with a low cracking pressure (0.062 PSI) was used, allowing air to enter the air spring with minimal extension force from the knee (High flow check valve, #91030, Qosina, Edgewood, NY). This reduces the amount of resistance to the user during the swing phase. Flow through the second valve is controlled by a solenoid. This control valve is closed during stance to provide a resistive torque, and open during swing so as not to interfere with the natural gait. The sensing system consists of a foot switch in each shoe connected to a microcontroller, which regulates a solenoid valve that engages the air spring. When heel strike is detected by one of the foot switches, the signal is passed to the microcontroller and the valve is controlled appropriately. A preliminary evaluation of the prototype’s performance was carried out. A single healthy female subject walked on an instrumented treadmill (Bertec, Columbus, OH, USA; 2160 Hz) at a speed of 1.0 m/s and a slope of −5°. Surface electromyogram (sEMG) readings were taken from two knee extensors (rectus femoris, RF, and vastus medialis, VM) using a wired system (Delsys, Natick, MA,
A Quasi-Passive Knee Exoskeleton to Assist During Descent
65
Fig. 1 Left: (a) control system, (b) air spring, (c) fabric interface. Right: during early stance phase, a solenoid valve seals the air chamber and restricts flexion. When opposite heel strike is detected, the valve opens and the knee can flex freely.
USA; 2160 Hz). These muscles were chosen as they are the only muscles whose activity increases during descent [6]. Control data were recorded from the subject walking on an inclined treadmill while not wearing the exoskeleton, and experimental data consisted of the subject walking on the same incline while wearing the device. Ground reaction forces were measured to enable gait segmentation. The resulting data were used to estimate the change in muscle activation while wearing the exoskeleton prototype. The EMG and GRF data were analyzed in Visual 3D (C-Motion, Rockville, MD, USA). The EMG data were band-pass filtered (4th order Butterworth, cut-off 20–450 Hz), rectified and low-pass filtered (4th order Butterworth, cut-off 6 Hz) to obtain an EMG linear envelope. The last 10 strides of each 1 min trial were segmented using an automatic gait event detection algorithm based on GRF data (Visual 3D, C-Motion, Rockville, MD, USA) and averaged over the gait cycle.
3 Results and Discussion Figure 2 shows muscle activity during the entire gait cycle for the vastus medialis and rectus femoris, averaged over 10 strides. A 15 % decrease in EMG activity is seen for the RF, while an 8 % increase is seen for the VM. Thus, the device had the desired effect upon the RF but caused an undesirable increase for the VM. While an increase in extensor activity during the swing phase is to be expected, due to the mass added to the leg and the friction of the air spring, the increase in VM activity during the stance phase is surprising. The differences may be due to kinematic constraints at the knee that affect each muscle differently, or to effects of the device
66
E. Rogers et al.
Fig. 2 Left: EMG electrode placement during testing. Centre: EMG activity throughout the gait cycle, averaged over 10 strides. Right: The area under the curves in the EMG plots, giving an estimate of total EMG activity over the gait cycle.
on the hip; the RF is a biarticular muscle that crosses both hip and knee, while the VM crosses only the knee joint.
4 Conclusions This preliminary design has shown promising results for decreasing RF activity during descent using a quasi-passive knee exoskeleton. Further research is required in order to better understand the effect of the exoskeleton on muscle activity and joint kinetics. Future work will include testing with additional subjects and using inverse dynamics to estimate the device’s effect on knee moment during descent.
References 1. Riener, R., Rabuffetti, M., Frigo, C.: Stair ascent and descent at different inclinations. Gait Posture 15(1), 32–44 (2002) 2. Kuster, M., Wood, G.A., Sakurai, S., Blatter, G.: Stress on the femoropatellar joint in downhill walking: a biomechanical study. Z Unfallchir Versicherungsmed 86, 178–183 (1993) 3. Shamaei, K., et al.: Biomechanical effects of stiffness in parallel with the knee joint during walking. IEEE Trans. Biomed. Eng. 62(10), 2389–2401 (2015)
A Quasi-Passive Knee Exoskeleton to Assist During Descent
67
4. Walsh, C.J., Endo, K., Herr, H.: A quasi-passive leg exoskeleton for load-carrying augmentation. Int. J. Humanoid Rob. 4(3), 487–506 (2007) 5. Elliott, G., Marecki, A., Herr, H.: Design of a clutch–spring knee exoskeleton for running. J. Med. Devices 8(3) (2014) 6. Franz, J.R., Kram, R.: The effects of grade and speed on leg muscle activations during walking. Gait Posture 35(1), 143–147 (2012)
Wearable Sensory Apparatus for Multi-segment System Orientation Estimation with Long-Term Drift and Magnetic Disturbance Compensation Sebastjan Šlajpah, Roman Kamnik and Marko Munih
Abstract Orientation assessment based on wearable sensors is becoming crucial for providing feedback information in wearable robotics and sport monitoring. Gravitational acceleration and Earth’s magnetic field are commonly used as a reference vectors for orientation estimation. This paper presents a novel sensory fusion algorithm for assessing the orientations of human body segments in long-term human walking, and enhance performance in environment with magnetic disturbance. The proposed system was experimentally validated. The results show accurate joint angle measurements (error median below 5◦ ) with no expressed drift over time. The incorporated compensation of magnetic disturbances proved assessment with absolute median errors bellow 2.5◦ .
1 Introduction Wearable sensors are becoming essential in wearable robotics and sport applications. MEMS accelerometer, gyroscope, and magnetometer outputs fused within a Kalman filter are commonly used for assessment of kinematic parameters. Two reference vectors are used for orientation assessment: vector of gravitational acceleration for inclination and vector of Earth’s magnetic field for heading estimation. However, IMU accelerometer measures only difference between gravitational and dynamical acceleration. Henceforth, the assessment can be limited to This study was supported by the Slovenia Research Agency, and by EU FP7 project CYBERLEGs under grant FP7-ICT-2011-7-287894. S. Šlajpah (B) · R. Kamnik · M. Munih Laboratory of Robotics, Department of Measurements and Robotics, Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia e-mail:
[email protected] R. Kamnik e-mail:
[email protected] M. Munih e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_12
69
70
S. Šlajpah et al.
quasi-static motion [1, 2], or separated into slow motion relaying on gravity vector and fast dynamic motion relying on angular velocity integration [3, 4], what usually results in long-term drift. Magnetic disturbances were dealt with a normalisation of magnetic field vector in horizontal plane [3]. With this, magnetic disturbances influence only the estimated heading component of the orientation. Roetenberg et al. [5] used complementary Kalman filter where the standard deviation of modeled magnetic disturbance is increased relatively with the change of the total magnetic flux and magnetic dip angle. In this work we present a novel algorithm for orientation estimation with improved long-term stability based on kinematic model, and with enhanced robustness based on an advanced magnetic disturbance compensation approach.
2 Methods 2.1 Wearable Sensors Wearable sensory system comprises seven inertial and magnetic measurement units (IMUs) placed on segments of lower extremities and trunk as depicted in Fig. 1a. An IMU sensor (size 30 mm × 20 mm × 5 mm) consists of three-axis gyroscope (measuring range ±500 ◦ s−1 ), accelerometer (measurement range ±4G; G ≈ 9.81 m/s2 ), and a magnetometer (measurement range ±0.13 mT) [2]. Besides, two ParotecSystem pressure measurement shoe insoles are used for measuring reaction forces under the feet. The measuring system is operating with sampling frequency of 100 Hz.
Fig. 1 a Placement of the sensors; b right–left sequence of orientation calculations; c left–right sequence of orientation calculations
Wearable Sensory Apparatus for Multi-segment System Orientation Estimation …
71
2.2 Kinematic Relations The proposed sensory fusion for orientation estimation proposed in this paper is built upon an extended Kalman filter (EKF) algorithm [6]. The concept is based on a kinematic relation which states that on a rigid body the acceleration of any point can be determined if the angular velocity, angular acceleration and linear acceleration of other point on the body is known. The relation between measured acceleration from the IMU aS, j and the linear acceleration of the center point of joint connecting segment j with the previous one aC, j can be described as aS, j = aC, j + ωS, j × ωS, j × r j + ω˙ S, j × r j − gS, j ,
(1)
where ωS, j and ω˙ S, j denotes the measured angular velocity and angular acceleration, respectively. Vector r j is a position vector connecting aC, j and aS, j measurement points and gS, j gravitational acceleration.
2.3 Magnetic Model The global magnetic field BG,0 is used for orientation estimates at the following time steps. When the magnetic disturbance B D,k is introduced at the time step k, the orientational quaternion qk describes the relation between the measured magnetic field B I,k and the combination of Earth’s magnetic field BG,0 and magnetic disturbance: T T 0, BTI,k = qk∗ ⊗ 0, BTG,0 + BTD,k ⊗ qk .
(2)
We model the magnetic disturbance B D,k as B D,k = B D,k−1 + B˙ D,k t,
(3)
where B D,k−1 represents magnetic disturbance at previous time step k − 1 and t a length of a time step. The product B˙ D,k t describes the propagation of the magnetic disturbance and is defined as a difference between the global magnetic field assessed at previous time step and global magnetic field calculated from current measured magnetic field and change of orientation between two steps.
2.4 Model-Based Extended Kalman Filter Described models are implemented into EKF architecture for calculation of the orientation estimate by constantly fusing data from the gyroscope, accelerometer, and magnetometer [7]. For assessing the kinematic parameters in human walking a seven
72
S. Šlajpah et al.
segment model of human body was utilized. The foot that is in contact with the floor is considered to be the first segment in the kinematic chain. For the estimation of the individual segments orientation a recursive algorithm composed of seven EKFs is implemented. The direction of the calculation is determined with respect to the foot which is currently in the stance phase (see Fig. 1).
3 Results An experimental validation of human walking was accomplished in order to test the accuracy and long-term reliability of developed algorithm. Five subject were asked to walk continuously for 15 min on a polygon simulating every day activities: level ground walking, stair negotiation, and turning. Joint angles were assessed via wearable apparatus and compared to angles measured with reference optoelectronic measuring system Optotrak Certus. Absolute errors between the reference and assessed joint angles for ankles, knees, and hips during 1st, 5th, 10th, 15th min and during the whole trial for all trails are presented in Fig. 2a. For validation of magnetic disturbance compensation we used a 3 DOF pendulum equipped with IMU. The pendulum was manually swung in the vicinity of a big metal box. Assessed orientation was compared to the one obtained with reference measuring system. Results, describing the absolute angle errors around x, y, and z axes, are depicted in Fig. 2b.
Fig. 2 a Absolute error between assessed and reference joint angles for ankles, knees, and hips during different time slots accomplished during walking on a polygon; b absolute error for angles around x, y, and z axes (θ X , θY , θ Z ) assessed with EKF without magnetic compensation (EKF-NC) and EKF with compensation (EKF-C)
Wearable Sensory Apparatus for Multi-segment System Orientation Estimation …
73
4 Discussion and Conclusion The presented paper introduces the novel sensory system and algorithms for longterm orientation assessment with included magnetic disturbance compensation. Experimental testing demonstrate that the wearable sensory apparatus is appropriate for measuring kinematic parameters in long-term, uninterrupted human motion of different dynamics. The median achieved errors are 3.0◦ , 2.7◦ , 2.4◦ , 3.3◦ , 2.6◦ , and 3.1◦ for the left and the right ankle, knee, and hip joint angles, respectively. From results, no tendency for increasing of median error with time can be observed. The EKF with incorporated compensation of magnetic disturbances is capable of orientation estimation with moderate error even when the Earth’s magnetic field is disturbed. Compensation of magnetic disturbances yields significantly smaller error (θRMS < 2.5◦ ) than non-compensated (up to 220◦ ). Results prove that the proposed wearable sensory apparatus can be used for acquisition of human motion kinematic parameters or to provide feedback within the control of wearable robots in everyday life situations regardless of the duration of movements or surrounding environment.
References 1. Luinge, H.J., Veltink, P.H.: Measuring orientation of human body segments using miniature gyroscopes and accelerometers. Med. Biol. Eng. Comput. 43(2), 273–282 (2005) 2. Beravs, T., Podobnik, J., Munih, M.: Three-axial accelerometer calibration using Kalman filter covariance matrix for online estimation of optimal sensor orientation. IEEE Trans. Instrum. Meas. 61(9), 2501–2511 (2012) 3. Madgwick, S.O., Harrison, A.J., Vaidyanathan, R.: Estimation of IMU and MARG orientation using a gradient descent algorithm. In: 2011 IEEE International Conference on Re-habilitation Robotics (ICORR), p. 17. IEEE (2011) 4. O’Donovan, K.J., Kamnik, R., O’Keeffe, D.T., Lyons, G.M.: An inertial and magnetic sensor based technique for joint angle measurement. J. Biomech. 40(12), 2604–2611 (2007) 5. Roetenberg, D., Luinge, H.J., Baten, C.T., Veltink, P.H.: Compensation of magnetic disturbances improves inertial and magnetic sensing of human body segment orientation. IEEE Trans. Neural Syst. Rehabil. Eng. 13(3), 395–405 (2005) 6. Welch, G., Bishop, G.: An introduction to the Kalman filter. Department of Computer Science, University of North Carolina, Chapel Hill, NC (2006) 7. Šlajpah, S., Kamnik, R., Munih, M.: Kinematics based sensory fusion for wearable motion assessment in human walking. Comput. Methods Programs Biomed. 116(2), 131–144 (2014)
A Portable Active Pelvis Orthosis for Ambulatory Movement Assistance Andrea Parri, Tingfang Yan, Francesco Giovacchini, Mario Cortese, Marco Muscolo, Matteo Fantozzi, Raffaele Molino Lova and Nicola Vitiello
Abstract Aging of population and increased incidence of gait impairments are dominant trends undermining social welfare and healthcare system. Lower-limb wearable robots proved to be a viable solution for recovering mobility of people affected by gait disorders. This work presents the design of the mechatronic architecture of a fully self-contained active pelvis orthosis (APO) for assisting hip flexion/extension movements during daily living activities. The APO could act compliantly with the human biomechanics thanks to series-elastic actuation and to a novel kinematics chain endowed with passive degrees of freedom. The portability and autonomy of the control system have opened the horizon to explore different assistive tasks in out-of-lab scenarios.
A. Parri (&) T. Yan F. Giovacchini M. Cortese M. Muscolo M. Fantozzi N. Vitiello The BioRobotics Institute, Scuola Superiore Sant’Anna di Pisa, Pontedera, Pisa, Italy e-mail:
[email protected] T. Yan e-mail:
[email protected] F. Giovacchini e-mail:
[email protected] M. Cortese e-mail:
[email protected] M. Muscolo e-mail:
[email protected] M. Fantozzi e-mail:
[email protected] N. Vitiello e-mail:
[email protected] R.M. Lova N. Vitiello Fondazione Don Carlo Gnocchi, Florence, Italy e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_13
75
76
A. Parri et al.
1 Introduction Lower limb mobility is essential for the daily-life activities. Ambulatory disabilities could bring barriers to the daily life activities both physically and psychologically [1] undermining the industrialized society global welfare. Within this framework, researchers have never stopped exploring approaches to assist the locomotion or restore lost ambulatory functionalities by conceiving different assistive devices. In the recent decades, wearable robots, including assistive exoskeletons and rehabilitation orthoses [2, 3], are gaining considerable interests for coping with lower-limb disabilities for restoring walking capabilities or used as treadmill-based gait-training platforms in rehabilitation. In the state of the art, there are also exoskeletons conceived for assisting elderly or persons with mild gait impairments in daily living acitivities. For instance, the HONDA Stride Management Assist (Honda, Tokyo, Japan), an hip exoskeleton from Samsung (Samsung, South Korea) [4] and the Hyundai Wearable Robots for walking assistance (Hyundai Motor Company, South Korea) are portable light-weight devices specialized in assisting hip movements in the sagittal plane. Grounded on the experience matured in [5] we developed a portable version of an Active Pelvis Orthosis (APO) for assisting hip joints flexion-extension during locomotion-related activities. APO is endowed with batteries allowing sufficient power autonomy, still preserving low encumbrances and weights. The robot is designed to comply with the user’s biomechanics in order to prevent from rigid, uncomfortable interactions. In this framework, the so-called well-established Series Elastic Actuator (SEA) architecture [6] is exploited. Furthermore, a novel chain of passive degrees of freedom (DoFs) was designed for preserving the free motion of the limb around hip rotational axes different from the actuated one. A scene of assisted walking in ecological environment is given in Fig. 1.
2 Materials and Methods This section presents the mechatronic architecture of the APO. The main subsystems, i.e. mechanics, actuation units and control, are described in the following subsections.
2.1
Mechanics
The mechanical structure of the APO is constituted of three main parts: the pelvis frame, the transmission system and the thigh linkages. The pelvis frame structure is constituted of a main carbon fibre plate connecting the exoskeleton to the pelvis cuff and then to the user. Two carbon-fibre plates (one for each leg) can slide on it by
A Portable Active Pelvis Orthosis for Ambulatory Movement Assistance
77
Fig. 1 APO worn by a user in an out-of-lab scenario.
means of two linear guides. The linear guides allow an easy wearing of the actuation units of the exoskeleton. Two carbon-fibre lateral extensible arms are connected to the central element of the device through a novel chain of passive DoFs [7] that allowed to place the actuation units on the rear part of the device, i.e. on the back of the person for maintaining a more natural swing of the arms in the sagittal plane. As a result both axes of passive adduction abduction and active flexion extension are collocated with the anatomical joint axes. Furthermore, each lateral arm is endowed with passive regulation mechanisms to match the anthropometries and guarantee the alignment of robotic and human joint axes. A curved carbon fibre link was designed for connecting the APO to the thighs through orthotic shells.
2.2
Actuation Units
The APO is endowed with two SEA-based actuation units, one for each side of hip flexion-extension joint, located on the rear part of the lateral arms, Each SEA has a
78
A. Parri et al.
single axis configuration constituted by a 70 W DC motor (EC60, Maxon Motor®, Sachseln, Switzerland) coupled with a 100:1 Harmonic Drive (CPL-14A-100-2A, Harmonic Drive®, Limburg, Germany). The Harmonic Drive is connected to the torsional spring whose deformation is directly measured by an absolute 17-bit Rotary Electric Encoder™(DS-37Netzer Precision Motion Sensors Ltd., Misgav, Israel). The torque is transmitted to the hip joint through a pulley wrapped with steel cable. A 17-bit absolute Rotary Electric Encoder™ (DS-25Netzer Precision Motion Sensors Ltd, Misgav, Israel) placed on the hip axis measures the actual hip angle. Each actuation unit weights slightly higher than 1 kg.
2.3
Control
The control system runs on a real-time controller, a sbRIO-9632 (National Instruments, Austin, Texas, US), endowed with a FPGA processor. The high-level layer runs at 100 Hz. The low-level layer runs on the FPGA at 1 kHz. Both the experimenter and the user can turn off the apparatus by means of an emergency button. APO is powered by a 33-V LiPo battery pack ensuring an autonomy lasting about 2 h. Including electronics and battery, APO entirely weights 8 kg. Low-level closed-loop torque controllers are independent for each actuation unit and rely on a 2-pole-2-zero compensator designed to achieve the minimum joint output impedance and a relatively high closed-loop bandwidth.
3 System Validation and Results This section presents the experimental activity performed to assess the controller performance in terms of output mechanical impedance; results are then reported. The output joint mechanical impedance has been preliminarily validated through a characterization session carried out to assess the transparency of the APO to the intentional movement of the user. The test was carried out with the operator manually displacing the link with a 20° amplitude motion with a variable frequency in the range [0.5–3.2] Hz being the commanded torque set to 0 Nm. The transfer function from the joint angle displacement to actuator output torque is an estimate of the output impedance of torque control of APO, i.e. a measurement of the resistive action exerted by the robot responding to an external applied movement (Fig. 2 shows the Amplitude Bode plot of the transfer function). The interaction torque amplitude increases with the motion frequency as it can be seen that the joint output impedance increases across the spectrum, from about 1 Nm/rad, for a 0.5 Hz motion, up to about 5 Nm/rad at 3.0 Hz.
A Portable Active Pelvis Orthosis for Ambulatory Movement Assistance
79
Fig. 2 Amplitude Bode diagram of the transfer function from angular displacement to interaction torque.
4 Discussion and Conclusion In this work, we presented a novel robotic platform designed to assist the locomotion-related activities of daily living of people with mild lower-limb impairments. Compared with its first laboratory prototype version [5], the APO is totally portable, being control electronics and batteries integrated in a self-contained mechatronic structure. A novel kinematic chain and minimum output actuator impedance played a significant role in improving ergonomy and usability of the system. Portability and adaptive control policies based on smart cognitive human-robot interfaces guaranteed an effective assistance in activities of daily living [8]. Future works will focus on a mechatronics optimization steered toward a light-weight and compact APO with intuitive and reliable human-robot interfaces.
References 1. Ferrucci, L., Giallauria, F., Guralnik, J.M.: Epidemiology of aging. Radiol. Clin. North Am. 46(4), 643–652 (2008) 2. Yan, T., Cempini, M., Oddo, C.M., Vitiello, N.: Review of assistive strategies in powered lower-limb orthoses and exoskeletons. Robot. Auton. Syst. 64, 120–136 (2014) 3. Tucker, M.R., Olivier, J., Pagel, A., Bleuler, H., Bouri, M., Lambercy, O.: Control strategies for active lower extremity prosthetics and orthotics: a review control strategies for active lower extremity prosthetics and orthotics: a review. J. NeuroEng. Rehabil. 12 (1) (2015) 4. Jang, J., Kim, K., Lee, J., Lim, B., Shim, Y.: Online gait task recognition algorithm for hip exoskeleton. In: IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 5327–5332, Hamburg (2015) 5. Giovacchini, F., Vannetti, F., Fantozzi, M., Cempini, M., Cortese, M., Parri, A., Yan, T., Lefeber, D., Vitiello, N.: A light-weight active orthosis for hip movement assistance. Robot. Auton. Syst. September 2014 6. Pratt, G.A., Williamson, M.M.: Series elastic actuators. In: Proceedings 1995 IEEE/RSJ International Conference on Intelligent Robots and Systems. Human Robot Interaction and Cooperative Robots, vol. 1, pp. 399–406 (1995)
80
A. Parri et al.
7. Vitiello, N., Giovacchini, F., Cempini, M., Fantozzi, M., Moisè, M., Muscolo, M., Cortese, M.: Sistema di attuazione per ortesi di anca. Application No. FI2015A000025 (Italian Patent), Application Date: 9 February 2015. Pending for acceptance 8. Garate, V.R., Parri, A., Yan, T., Munih, M., Lova, R.M., Vitiello, N., Ronsse, R.: Walking assistance using artificial primitives. IEEE Robot. Autom. Mag. 23(1), 83–95 (2016)
Part III
Soft Wearable Robotics
XoSoft - A Vision for a Soft Modular Lower Limb Exoskeleton Jesus Ortiz, Eduardo Rocon, Valerie Power, Adam de Eyto, Leonard O’Sullivan, Markus Wirz, Christoph Bauer, Samuel Schülein, Konrad S. Stadler, Barbara Mazzolai, Wouter B. Teeuw, Chris Baten, Corien Nikamp, Jaap Buurke, Freygardur Thorsteinsson and Jeanette Müller Abstract XoSoft is an EU project that proposes the development of a modular soft lower-limb exoskeleton to assist people with mobility impairments. It aims to be user friendly and comfortable to wear, with a significant impact on the person’s mobility and health, on their independence and quality of life. Being a modular system, it comprises of ankle, knee and hip elements, which can be used individually or combined and used unilaterally or bilaterally. XoSoft follows a user centered design strategy achieved by involving primary, secondary and tertiary end users as participatory stakeholders in the design and
J. Ortiz (&) Department of Advanced Robotics, Istituto Italiano di Tecnologia, Via Morego, 30, 16163 Genoa, Italy e-mail:
[email protected] E. Rocon Neural and Cognitive Engineering Group, Spanish National Research Council, Madrid, Spain e-mail:
[email protected] V. Power A. de Eyto L. O’Sullivan Design Factors Group, University of Limerick, Limerick, Ireland e-mail:
[email protected] A. de Eyto e-mail:
[email protected] L. O’Sullivan e-mail:
[email protected] M. Wirz C. Bauer Institute of Physiotherapy, ZHAW Zurich University of Applied Sciences, Winterthur, Switzerland e-mail:
[email protected] C. Bauer e-mail:
[email protected] S. Schülein Geriatrics Center Erlangen, Erlangen, Germany e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_14
83
84
J. Ortiz et al.
development process. Preliminary findings of the interviews with the different users groups are presented in this paper. Advanced textiles and smart materials are being developed to create sensing, variable stiffness joints and flexible tactile sensors. Control will be through biomimetics to identify the user’s motion and intention and to determine and provide the appropriate level of assistance. Connected health connectivity and analysis will enable the wearer and their clinicians/therapist to review activity information. The concept will be tested extensively in the lab, and subject to trials in clinical settings and home environments.
1 Introduction Many elderly and patient groups experience varying degrees of mobility impairment. Assistive devices play a pivotal role in their lives and impact on their ability to live independently and perform basic tasks of daily living. But many assistive aids do not encourage or support activation of legs, e.g. powered wheel chairs. The proportion of the world’s elderly population is expected to increase significantly by 2050 [1]. Age-related decline in voluntary muscle strength results in important K.S. Stadler Institute of Mechatronic Systems, Zurich University of Applied Sciences, Winterthur, Switzerland e-mail:
[email protected] B. Mazzolai Center for Micro-BioRobotics of IIT@SSSA, Istituto Italiano di Tecnologia (IIT), Viale Rinaldo Piaggio 34, Pontedera, 56025 Pisa, Italy e-mail:
[email protected] W.B. Teeuw Saxion’s Research Group Ambient Intelligence, Enschede, Netherlands e-mail:
[email protected] C. Baten C. Nikamp J. Buurke Roessingh Research and Development, Enschede, Netherlands e-mail:
[email protected] C. Nikamp e-mail:
[email protected] J. Buurke e-mail:
[email protected] F. Thorsteinsson Össur, Reykjavik, Iceland e-mail:
[email protected] J. Müller Accelopment A.G., Zürich, Switzerland e-mail:
[email protected]
XoSoft - A Vision for a Soft Modular Lower Limb Exoskeleton
85
changes to body composition and function [2]. Remaining active and mobile during ageing is crucial to overall health and cognitive function [3]. Mobility assistance is also required by patients, such as stroke sufferers or patients with incomplete Spinal Cord Injuries (SCI). Globally circa 16 million people per year experience a stroke for the first time, of which 5 million experience varying degrees of mobility difficulty, which significantly impacts their ability to perform tasks of daily living [4]. Accidents are the most common cause of SCI lesions; of which about the 51 % are incomplete, i.e. the person is partially disabled [5]. Beside SCI caused by accidents, there are an in-creasing number of cases due to non-traumatic causes. A characteristic of the latter group is that patients are older and that the amount of incomplete SCI is even higher. Patients with an incomplete SCI do not suffer complete loss of sensory-motor function in the lower limbs but they may still require assistance to walk. XoSoft is an EU project that is currently developing a soft lower-limb exoskeleton to assist people with mobility restrictions due a partial loss of sensory or motor function. Typically, the existing exoskeletons have a rigid structure that is heavy weight, bulky and is a risk for lesions due to excessive pressure to the skin. In contrast, XoSoft will have a flexible and adaptable structure. Its design will be lightweight due to the basic structure and the fact that assistance requires low power consumption (e.g. batteries). XoSoft is not intended to substitute complete loss of function like already existing exoskeletons, but rather assist the user in a tailored manner. XoSoft is a user centered design lower limb exoskeleton using smart soft robotics, biomimetic controlled actuation and connected health data feedback and interface.
2 User Centered Design A core feature of XoSoft is that it follows a User Centered Design (UCD) approach [6]. UCD employs design ethnography and participatory stakeholder involvement as key drivers for the technology development to ensure user needs are at the forefront XoSoft’s development. This approach also uses an iterative product design methodology which re-evaluates and improves the user appropriateness of the system at each stage. Three versions of XoSoft are identifiable based on the module and subsystem developed, and persons tested on: 1. Alpha version: This version is being designed using existing technologies and its main purpose is to use it as a test bed for the technologies and a mechanism to ensure the design process remains user centered. 2. Beta version: Based on the alpha prototype, we will replace and incorporate new technologies to the system. The different Beta versions of the system will be tested during laboratory setups. The last Beta version will include the full sensing and actuation system but with limited autonomous functionality. This version will be used in the clinical validation.
86
J. Ortiz et al.
3. Gamma version: The final version of the Product Service System (PSS) will be fully autonomous as a person would use it (i.e. run on batteries and have an on-board computer) and it will be ready for the home trials.
3 User Groups In this section we provide some preliminary results of the analysis of the XoSoft stakeholders and their needs, including demographics of the primary users (PU), identification of the secondary users (SU) and the commercial analysis of the tertiary users (TU). We selected two distinct PU groups: neurological populations, specifically people with stroke and incomplete SCI, and older adults with mobility impairments. Apart from the assistive capabilities of the system, the PUs put special emphasis in the wearability and usability of the system, and in particular in the ability of using the system in their daily life, allowing then to use it under their normal clothes, and not impeding daily activities. SUs are defined as persons or organisations directly in contact with a PU. Professional SUs include healthcare professionals from multidisciplinary backgrounds, such as physicians, nurses, physiotherapists, occupational therapists, speech and language therapists, and public health nurses, as well as professional care assistants, home help service providers and other support staff. Non-professional SUs may include spouses, family members, friends, neighbours, and community and/or voluntary organisations. The main focus of the SU is centered in the functionality of the system and the positive immediate outcomes for the patients, rather than in aesthetics as the PUs. TUs are defined as bodies that have interests that are in some way affecting the potential for the device in question. They are often related to the framework that is in place on the regulatory side and/or to financial provisions for the devices that are supplied to patients or other users.
4 System Description XoSoft exoskeleton comprises an ankle-foot-knee module which can be worn on one or on both sides and a hip module. All modules will be designed to be highly customizable and to be operated separately or in combination. XoSoft is meant to be easy to wear, comfortable, serviceable and compatible with the daily life of the users. These aspects will ensure a high acceptability by the users, being used not only in clinical environments but also at home and everyday life. XoSoft uses non-traditional sensing and actuation systems based on smart materials and soft structures. Variable stiffness joints based on Electro-Rheological
XoSoft - A Vision for a Soft Modular Lower Limb Exoskeleton
87
Fluids (ERF) are being developed. Different solutions are being analyzed to develop flexible sensors to measure bending, stretching, force and pressure. In later stages, we will develop units fusing sensing and actuation capabilities based on the above technologies. The system control uses a biomimetic approach, which requires the measurement and identification of the movement of the user. The system aims to inertial motion sensors to measure and identify gait patterns and other variables, such as kinematics, intention of turning, transitions, and instabilities. This information, used in conjunction with the biomimetic approach, determines the appropriate assistance and strategies. The activity of the user will be monitored by providing connected online and offline feedback to the clinicians and users. The information will be segmented for easy analysis and understanding. The clinician or therapist can also adjust settings of the device remotely, for example to tune the device to provide more or less assistance. The system should be suitable for use in home environments. Therefore, it must operate autonomously for long periods of time with low energy requirements (lower battery size and weight, less heat dissipation, etc.). To reach this objective we will optimize the power requirements for sensors, actuators and computing units (ultra-low power processors). For the most demanding actuation units we will use soft mechanical solutions for energy accumulation.
5 Conclusions In this paper, we have presented the main characteristics of the soft exoskeleton, which will be developed in the XoSoft project, putting special emphasis in the user centered design process as well as preliminary findings from the analysis of the different user groups. Acknowledgment This work has received funding from the European Union’s Horizon 2020 framework programme for research and innovation under grant agreement No 688175.
References 1. World Health Organization, Facts about ageing (2014). http://www.who.int/ageing/about/ facts/en/ 2. Arnold, P., Bautman, I.: The influence of strength training on muscle activation in elderly persons, Exp. Gerontol. 58, 58–68 (2014) 3. Volkers, M., et al.: Lower limb muscle strength: why sedentary life should never start. Arch. Gerontol. Geriatr. 54, 399–414 (2012)
88
J. Ortiz et al.
4. Strong, K., Mathers, C., Bonita, R.: Preventing stroke: saving lives around the world. Lancet Neurol. 6(2), 182–187 (2007) 5. National Spinal Cord Injury Statistical Center, Annual Statistical Report – Complete Public Version, University of Alabama, Birmingham (2014) 6. Sanders, E.B.N.: From user-centered to participatory design approaches. Design and the Social Sciences: Making Connections, pp. 1–8 (2002)
On the Efficacy of Isolating Shoulder and Elbow Movements with a Soft, Portable, and Wearable Robotic Device Zahra Kadivar, Christopher E. Beck, Roger N. Rovekamp, Marcia K. O’Malley and Charles A. Joyce
Abstract Treatment intensity has a profound effect on motor recovery following neurological injury. The use of robotics has potential to automate these labor-intensive therapy procedures that are typically performed by physical therapists. Further, the use of wearable robotics offers an aspect of portability that may allow for rehabilitation outside the clinic. The authors have developed a soft, portable, lightweight upper extremity wearable robotic device to provide motor rehabilitation of patients with affected upper limbs due to traumatic brain injury (TBI). A key feature of the device demonstrated in this paper is the isolation of shoulder and elbow movements necessary for effective rehabilitation interventions. Herein is presented a feasibility study with one subject and demonstration of the device’s ability to provide safe, comfortable, and controlled upper extremity movements. Moreover, it is shown that by decoupling shoulder and elbow motions, desired isolated joint actuation can be achieved.
This work was funded by the DARPA Warrior Web program (BAA-13-43) and NASA Grant NNX14AK51G. Z. Kadivar M.K. O’Malley Rice University, Houston, USA e-mail:
[email protected] M.K. O’Malley e-mail:
[email protected] C.E. Beck (&) R.N. Rovekamp C.A. Joyce NASA JSC Wearable Robotics Laboratory, Houston, USA e-mail:
[email protected] R.N. Rovekamp e-mail:
[email protected] C.A. Joyce e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_15
89
90
Z. Kadivar et al.
1 Introduction Current rehabilitation research indicates that task-specific [1] and intensive [2] practice could significantly improve motor recovery and induce neuroplasticity after brain injury. The greater effectiveness of intensive task-specific practice relative to standard therapy techniques suggests that repetitive motor practice is a crucial rehabilitation component, and provides a key opportunity for the introduction of robotics in rehabilitation. Achieving voluntary isolated movements after neurological injuries such as TBI and stroke is a common therapy goal. Synergistic movement patterns that occur after neurological injuries have the potential to interfere with activities of daily living (ADLs) and limit functional independence of the individual. A common rehabilitation approach to address unwanted synergies includes manual fixation of undesired movements and facilitation of desired movements. The trend of applying robotics to address these rehabilitation challenges continues to grow. Existing robotic devices [3–5] that provide the capability to assist the affected proximal upper extremities for rehabilitative purposes, however, are ground-based. Soft, portable, wearable robotic devices offer additional potential advantages, including allowing more task-oriented therapy (i.e. performing functional tasks as opposed to merely prescribed motions that are somewhat decoupled from practical ADLs), and bringing therapy to new venues including the home. The authors developed a soft upper extremity wearable robotic device at the NASA JSC Wearable Robotics Laboratory for the purpose of providing effective upper extremity motor rehabilitation related to TBI, a condition that has left approximately 5.3 million Americans with a long-term disability [6]. In this paper, a feasibility study with a single subject is presented to demonstrate successful isolation of upper extremity movements with a soft wearable robotic device. Given the portable nature of the device, performance of these rehabilitative movements can occur in the modified settings and task-specific situations alluded to above, potentially offering a richer approach to the rehabilitation of upper extremity function.
2 Materials and Methods 2.1
Device Description
The novel soft, portable, wearable robotic device (see Fig. 1) actively controls the right shoulder and elbow, both positioning the limb in specific orientations and commanding the limb through desired motions. The device uses a minimal amount of rigid components and custom force-controllable tendon actuators developed at NASA JSC. While the system currently only actuates the right arm, it was designed to easily incorporate a left arm in the future. Careful design consideration was given to ease donning and doffing of the device, both to minimize valuable patient and physical therapist time and to
On the Efficacy of Isolating Shoulder and Elbow Movements
91
Fig. 1 An artist’s depiction of the soft upper extremity wearable robotic device developed at NASA JSC for motor rehabilitation related to TBI.
maximize patient comfort, especially in the presence of muscle tone. Moreover, comfort of the patient throughout the session was a priority in the design, resulting in a physical human-robot interface that effectively distributes loads around the torso, minimizing pressure points and interface migration. The device offers multiple control modes. In active assist mode, shoulder abduction and flexion, and elbow flexion, may be commanded either simultaneously via coordinated control or individually while holding position/orientation of the other joints. In passive assist mode, the user may freely move his or her limb while the system provides minimal torque to the shoulder and elbow degrees of freedom.
2.2
Subject Description
Evaluation of the device was conducted with a single subject, male, age 29, with moderate TBI, 1.5 years post-injury. Due to right hemiparesis the subject was unable to achieve full ranges of shoulder abduction and flexion against gravity. However, the subject was able to achieve full elbow flexion against gravity and minimal resistance. Compensatory movements were present during active shoulder and elbow movements leading to synergistic patterns (e.g. shoulder abduction and internal rotation during elbow flexion). Increased tone was present in the muscles controlling elbow flexion, finger flexion, and shoulder internal rotation of the affected limb.
2.3
Exercise Description
Due to the subject’s inability to isolate movements, a commanded isolated active elbow flexion/extension trajectory was chosen while holding a commanded position/ orientation of the shoulder. This allowed for the evaluation of the utility of the device
92
Z. Kadivar et al.
in supporting movement isolation. In the first trial, the subject was asked to follow a commanded elbow joint angle trajectory between 30° and 90° at a constant velocity of *13°/s. The subject was instructed to only move the elbow joint. During this trial, minimum shoulder and elbow joint torque was applied via the actuators such that the subject was responsible for performing the motion with no assistance from the device. In the second trial, the subject was asked to perform the same movement while the device maintained the shoulder in an abducted position. The subject was provided with real-time visual feedback of the desired and actual elbow angles during all trials.
3 Results The testing session lasted approximately one hour, without any adverse events or reports of discomfort from the subject. The actuated range of motion (intentionally constrained to avoid exceeding the patient’s passive range of motion) was demonstrated for shoulder abduction, flexion, and elbow flexion. In active assist mode, the device properly positioned the limb at the commanded orientations. The subject used compensatory shoulder abduction to achieve desired elbow range of motion (as seen in Fig. 2). With robot-assisted fixation of the shoulder, a similar range of elbow movement was achieved. However, compensation from shoulder abduction was greatly reduced. Fixation of the shoulder not only allowed
Fig. 2 Compensatory shoulder abduction is evident during elbow movement when shoulder is free (Trial 1; blue line). Shoulder fixation (Trial 2; black line) allows isolation of elbow and shoulder movements. (Color figure online)
On the Efficacy of Isolating Shoulder and Elbow Movements
93
for the ability to move the elbow joint in isolation, but it also improved smoothness of the movement, as joint velocity error decreased from the first trial, during which the shoulder was free.
4 Conclusion The results demonstrate the feasibility of isolating upper extremity movements with a soft, portable, wearable robotic device. The current study indicates the possibility of using the developed wearable robotic device for improving motor control and coordination of the upper extremity after traumatic brain injury. Although rehabilitation was not directly addressed, unique control features and the portability, wearability, and comfort of the device highlight its potential for upper limb rehabilitation. In light of this, such a device could pave the way to more task-based forms of therapy and provide a medium for increased dosage. Additional potential applications of the device are already under investigation, including assist-as-needed control of the upper extremity and assistance with ADLs. The desire to provide both beneficial rehabilitation and practical ADL assistance with the same hardware is well aligned with the capabilities of a soft, portable, wearable robotic device.
References 1. Hubbard, I.J., Parsons, M.W., Neilson, C., Carey, L.M.: Task-specific training: evidence for and translation to clinical practice. Occup. Ther. Int. 16(3–4), 175–189 (2009) 2. Sunderland, A., Tuke, A.: Neuroplasticity, learning and recovery after stroke: a critical evaluation of constraint-induced therapy. Neuropsychol. Rehabil. 15(2), 81–96 (2005) 3. Mihelj, M., Kiefer, G., Perndl, C., Müller, R., Riener, R.: ARMin – Exoskeleton for arm therapy in stroke patients. In: IEEE International Conference on Rehabilitation Robotics, Noordwijk, The Netherlands, pp. 68–74 (2007) 4. Krebs, H.I., Hogan, N., Aisen, M.L., Volpe, B.T.: Robot-aided neurorehabilitation. IEEE Trans. Rehabil. Eng. 6(1), 75–87 (1998) 5. Burgar, C.G., Lum, P.S., Shor, P.C., Van der Loos, H.F.M.: Development of robots for rehabilitation therapy: the Palo Alto VA/Stanford experience. J. Rehabil. Res. Dev. 37(6), 663–673 (2000) 6. Injury Prevention & Control: Traumatic Brain Injury & Concussion. (n.d.). Center for Disease Control and Prevention. http://www.cdc.gov/traumaticbraininjury/severe.html. Accessed 1 Apr 2016
Design Improvement of a Polymer-Based Tendon-Driven Wearable Robotic Hand (Exo-Glove Poly) Haemin Lee, Brian Byunghyun Kang, Hyunki In and Kyu-Jin Cho
Abstract This paper presents the design improvement of a polymer-based tendon-driven wearable robotic hand, Exo-Glove Poly. The wearability and adaptiveness are the key points to design the Exo-Glove Poly in considering the cases of practical use. Thus, magnets are embedded into the wearable part for easy donning and doffing. Also, the tendon length adjustment mechanism is designed to adapt different hand sizes by changing length of the tendons. Through these improvements, it is increased the change to practical use of the Exo-Glove Poly.
1 Introduction Recently, the number of patients suffering from the limited hand function is consistently increasing. Spinal cord injury (SCI) and stroke are the major reasons of the hand paralysis. Because the patients’ hand cannot be operated naturally, the patients need a physical assist by carers to perform activities of daily living (ADLs) such as washing, grooming, and feeding. Also, it is almost impossible for them to open a door knob or open a bottle by themselves. Therefore, hand disability deteriorates their quality of life due to not being able to live independently. To improve their quality of life by making them possible to perform ADLs without any help, many hand assistive devices have been developed. Assistive devices transfer external forces to the hand to generate proper movements of paralyzed hand. This study was supported by a grant (NRCTR-EX16001) of the Translational Research Center for Rehabilitation Robots, Korea National Rehabilitation Center, Ministry of Health & Welfare, Korea. H. Lee B.B. Kang H. In K.-J. Cho (&) Biorobotics Lab of Seoul National University, Seoul, Korea e-mail:
[email protected] H. Lee e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_16
95
96
H. Lee et al.
Fig. 1 Improved version of the Exo-Glove Poly
Conventional hand assistive devices are categorized into two groups, which are a rigid exoskeleton [1–4] and a soft exoskeleton [5–9]. Rigid exoskeletons usually applies forces to the finger in normal direction to the finger, which means easy force transmission and easy control. However, size of the structure is bulky due to the mechanisms to align exoskeleton joints to human finger joints. To overcome the size issue of rigid exoskeleton, soft exoskeletons made of soft materials have been proposed. Due to the softness and compliance of materials, wearing part of soft exoskeleton became compact and easily adaptable to the user’s hand without concerning joint alignment between robot and human. In a previous design, we developed a polymer based tendon-driven wearable robot hand, Exo-Glove Poly, to assist people with hand disability [10]. Since Exo-Glove Poly uses polymer as its base material of wearable part, it can be easily adapt to the user’s hand. Thus, the wearable part is more compact than rigid exoskeleton’s wearable part. Furthermore, by using a tendon-driven actuation method and minimized actuator units, portability has also been achieved. Thanks to these features, the Exo-Glove Ploy is the one of the most suitable assistive devices, which can assist people with hand disability. However, still several problems, such as wearability and adaptiveness, remain to be solved considering the cases of practical use. The Exo-Glove Poly cannot be easily worn to patients’ hand. Also, in
Design Improvement of a Polymer-Based Tendon-Driven...
97
order to maximize the performance of the Exo-Glove Poly among different users, not only the wearable part but also the length of the actuating tendons need to be adjusted (Fig. 1). In this paper, we present the design improvement to increase wearability and adaptiveness of Exo-Glove Poly. Magnets are embedded into the wearable part for easy donning and doffing. And its fabrication process was developed. Also, the tendon length adjustment mechanism is designed to adapt different hand sizes by changing length of the tendons.
2 Design Improvement 2.1
Magnet Embedment
In a previous design, the wearable part of the Exo-Glove Poly has palmar band which connects the palmar part and dorsal part of the wearable part in one piece. Because the palmar band permanently connects both palmar and dorsal part, it is hard for users to inset their all digits into the glove. It is very difficult to wear that glove in this way, because the people with hand disability cannot maintain their hand posture. To make the wearing process easy, magnetically attachable palmar band are designed, using a pair of small magnets. One magnet is imbedded at the dorsal part and the other is imbedded at the palmar band.
Fig. 2 Magnet is embedded at the palmar band and dorsal part to increase wearability of the Exo-Glove Poly
98
H. Lee et al.
The magnets are located at the proper depth from the surface of parts. Important point in magnet embedding design is to maximize contraction force between the magnets while polymer layers surrounding magnets are not thin enough to be torn (Fig. 2).
2.2
Tendon Length Adjustment Mechanism
In order to decrease the weight of the Exo-Glove Poly, it is important to minimize the number of actuators. As the same reason, the Exo-Glove Poly uses only one actuator to drive finger flexion and extension antagonistically. In the antagonistic actuation, the length of flexion tendon and extension tendon should be carefully designed. If lengths of tendons are too short, range of motion of finger will be limited. On the other hand, if the lengths of tendons are too long, mechanical backlash will be increased. Therefore, required tendon length varies with the hand size, especially the length of finger. As one of the solution to this current problem, the tendon length adjustment mechanism has been developed, as shown sin Fig. 3. This mechanism is designed to wind and to unwind for single tendon. Using a pair of this mechanism, each length of both tendons can be adjusted to desired length independently.
Fig. 3 Tendon length adjustment mechanism
Design Improvement of a Polymer-Based Tendon-Driven...
99
3 Conclusion In this paper, magnet imbedding technique and tendon length adjustment mechanism are presented to increase wearability and adaptiveness of the Exo-Glove Poly. Using a pair of magnets, it is possible for user to wear the glove easily. Also, since the lengths of tendons can be adjusted by the adjustment mechanism, the Exo-Glove Poly can be adapted to different hand size. Through these improvements, it is increased the change to practical use of the Exo-Glove Poly.
References 1. Fontana, M., Dettori, A., Salsedo, F., Bergamasco, M.: Mechanical design of a novel hand exoskeleton for accurate force displaying. In: IEEE International Conference on Robotics and Automation (ICRA), pp. 1704–1709 (2009) 2. Chiri, A., Vitiello, N., Giovacchini, F., Roccella, S., Vecchi, F., Carrozza, M.C.: Mechatronic design and characterization of the index finger module of a hand exoskeleton for post-stroke rehabilitation. IEEE/ASME Trans. Mechatron. 17(5), 884–894 (2012) 3. Hasegawa, Y., Mikami, Y., Watanabe, K., Firouzimehr, Z., Sankai, Y.: Wearable handling support system for paralyzed patient. In: IEEE/RSJ International Conference in Intelligent Robots and Systems (IROS), pp. 741–746 (2008) 4. Wege, A., Hommel, G.: Development and control of a hand exoskeleton for rehabilitation of hand injuries. In: IEEE/RSJ International Conference in Intelligent Robots and Systems (IROS), pp. 3046–3051 (2005) 5. Polygerinos, P., Galloway, K.C., Savage, E., Herman, M., O’Donnell, K., Walsh, C.J.: Soft robotic glove for hand rehabilitation and task specific training. In: International Conference on Robotics and Automation (ICRA) (2015) 6. Yap, H.K., Lim, J.H., Nasrallah, F., Goh, J.C.H., Yeow, R.C.H.: A soft exoskeleton for hand assistive and rehabilitation application using pneumatic actuators with variable stiffness. In: International Conference on Robotics and Automation (ICRA) (2015) 7. Lee, S., Landers, K., Park, H.S.: Development of a Biomimetic Hand Exotendon Device (BiomHED) for restoration of functional hand movement post-stroke. IEEE Trans. Neural Syst. Rehabil. Eng. 22(4), 886–898 (2014) 8. Idrogenet srl. (2014), Gloreha®. http://www.gloreha.com/ 9. In, H., Kang, B.B., Sin, M., Cho, K.J.: Exo-Glvoe: soft wearable robot for the hand using soft tendon routing system. IEEE Robot. Autom. Mag. 22(1), 97–105 (2015) 10. Kang, B.B., Lee, H., In, H., Jeong, U., Chung, J., Cho, K.J.: Development of a polymer-based tendon-driven wearable robotic hand. In: IEEE International Conference on Robotics and Automation (ICRA) (2016)
Affective Touch and Low Power Artificial Muscles for Rehabilitative and Assistive Wearable Soft Robotics Jonathan Rossiter, Espen Knoop and Yuichi Nakamura
Abstract The goal in wearable rehabilitation is to restore the lost functionality of the body by rebuilding the sensory-motor link. This may be achieved through a replication, in an artificial or robotic system, of the physiotherapy methods employed by human experts. These methods are typically focused on physical manipulation. We suggest that a lower reliance on manipulation, combined with affective touch stimulation, has the potential to provide effective rehabilitation in lower power and lighter wearable devices. Here we consider affective touch driven by soft actuation and how this may be combined with low power artificial muscle actuators for physical rehabilitation.
1 Introduction People who have suffered from injury, disease and stroke often have reduced mobility. In order to maintain their independence, quality of life and mental health these people require rehabilitation and assistance. Rehabilitation is a short term program which aims to restore as much physical capability as possible. After rehabilitation has reached a plateau the patient may transition to a long term assisted lifestyle. Wearable robotics may be helpful in providing enhanced and automatic rehabilitation and assistance. Conventional rigid robotics, such as exoskeletons, have set the bar for ‘wearable’ physical rehab and assist devices [1, 2].
J. Rossiter—Supported by EPSRC grants EP/M026388/1 and EP/M020460/1. E. Knoop—Supported by the Dyson Foundation. J. Rossiter (&) E. Knoop Department of Engineering Mathematics, University of Bristol, Bristol, UK e-mail:
[email protected] Y. Nakamura Faculty of Engineering, Academic Center for Computing and Media Studies, Graduate School of Electrical Engineering, Kyoto University, Kyoto, Japan © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_17
101
102
J. Rossiter et al.
Unfortunately the limitations of rigid robots and their underlying reliance on geared motors and rigid metals and plastics reduces the adaptability and comfort of these devices and elevates their cost and complexity. More recently soft robotic devices have come to the fore as an alternative to rigid robots [3,4]. Advantages to wearable soft robotics include inherent compliance (and associated increased safety), adaptability, lower cost and higher user acceptability. Although most approaches to soft robotic assist and rehab devices target the same high power range as their rigid robotic counterparts, there is great potential for lower power devices. A patient with movement deficit may not require a high power assist device that will restore 100 % of their prior mobility. Rather a restoration of only a few percent may be sufficient to make a step change in their capabilities and quality of life. In addition, the acceptability of the wearable device (be it orthotic or prosthetic) is a major stumbling block to adoption and use [5]. Acceptability includes subjective assessment of comfort, functionality and aesthetics. We suggest that affective touch, combined with low power soft artificial muscle actuators can provide natural, comfortable and low-cognitive-load soft wearable assistive and rehabilitative devices.
2 Affective Touch The affective sense of touch is concerned with natural sensations which cover the spectrum from pleasant to unpleasant [6]. For example, smoothness and softness are often linked to pleasantness, whereas stiffness, roughness and coarseness are linked to unpleasantness. Affective haptics seeks to overcome the limitation of simple mechanical tactile stimulators by targeting the richer conceptual and emotional channel of communication that is encompassed by the affective sense of touch. This is in contrast to conventional electromechanical tactile stimulation, such as vibrations, which are convenient to generate with simple technologies, but which are not related to the natural sensations encountered in evolutionary history. The disconnect between conventional stimulation and human cognitive processes suggests that
Fig. 1 Affective tactile stimulation (left) provides a natural context and channel for rehabilitation, in contrast with non-affective unnatural stimulation (right) such as vibration. Physical rehabilitation in red, muscles in orange Affective touch
Non-affective touch
Affective Touch and Low Power Artificial Muscles for Rehabilitative …
103
rehabilitation using the natural feelings afforded by affective tough may be more effective. Figure 1 illustrates how the low cognitive impedance of affective touch, in contrast to non-affective touch, can be combined with physical assistance to reinforce a rehabilitation activity such as bending the leg.
2.1
Affective Tactile Stimulation
Affective touch can be communicated via a range of affective tactile stimulators, providing the foundation technologies for inclusion in wearable rehabilitation and assistance devices. These include the Tickler, a wearable soft robotic tactile stimulator that stokes or tickles the skin [7] (Fig. 2a). The Tickler is fabricated from a 3d printed compliant frame with integrated 3d printed rigid bars which hinge at their mid points within a soft membrane (Objet260 Connex, Stratasys Ltd.) The bars are actuated by shape memory alloy filaments at the ends away from the skin, causing the bars to pivot about their hinges and the ends in contact with the skin to move laterally. The relative lateral motion of the bar ends across the skin stimulates sensations described by test subjects variously as ‘a subtle tickling sensation’, ‘like a massage’, ‘comfortable’, ‘feels good’. A smaller scale affective stimulator is achieved by replacing the bars of the Tickler with rounded pins. We can also reduce the size of the stimulation elements to achieve a higher resolution or lower profile (Fig. 3) [8]. To actuate the pin array stimulators we use a dielectric elastomer (DE) electroactive polymer actuator. This DE membrane is stretched across the base of the pins, and electrodes are applied to the regions between the pins. When these regions are electrically stimulated (electric field *1 MV/m) they expand, causing lateral motion at the tips of the pins in contact with the skin. Larger laterotactile stimulation can be achieved using more conventional servo motors to push apart, or bring together, two patches attached to the skin (Fig. 2b, c). These ‘skin stretchers’ interact with the skin only and do not directly drive movement in the limbs. They impart a larger sensory stimulation and a higher degree of urgency than the smaller scale stimulators, providing an enhanced link between stretcher movement and limb movement.
3 Artificial Muscle Rehabilitation Electroactive polymer (EAP) artificial muscles, such as the DE actuators used above in the laterotactile stimulators, have power densities of the same order as biological muscle. This suggests their use in soft wearable assist devices. Unfortunately fabrication challenges, mismatches in natural frequencies, and reliability issues mean that high power electroactive polymer assist devices are not currently achievable. A better application of EAPs is as lower power assist devices, for example to
104
J. Rossiter et al.
a.
b.
c.
Fig. 2 Affective touch stimulators: (a) The Tickler laterotactile stimulator, (b), (c) large skin stretchers using servos.
regulate or boost level walking, or to assist with standing stability in those who would benefit from a small level of assistance. The inclusion of the above presented affective tactile stimulators in these low power devices is expected to improve the acceptance, comfort and cognitive compatibility of these devices. Reducing the power output of the wearable rehabilitation device also means that it can be made smaller, lighter and with lower cost. These will improve user adoption if the wearable device can be worn discreetly and comfortably under conventional clothing. With the integration of affective touch, and its emphasis on natural tactile communication, the device will be minimally disruptive cognitively, with consequent benefits in rehabilitation and long-term assistance.
Affective Touch and Low Power Artificial Muscles for Rehabilitative …
105
Fig. 3 Electroactive polymer driven laterotactile stimulators: (a) two different tactile arrays, (b) sensitivity and discrimination experiments.
4 Conclusions We have presented affective touch as an important and overlooked channel of communication for integration in wearable soft robotic rehabilitation and assist devices. By integrating artificial muscle-driven affective touch, through natural laterotactile stimulation such as stoking and tickling, with low power EAP actuators we propose a new form of wearable device. These are expected to be comfortable, discrete and have wider acceptance and adoption in target patients undergoing neuro and physical rehabilitation or in need of long term assistance.
References 1. Delph II, M.A., Fischer, S.A., Gauthier, P.W., Martinez Luna, C.H., Clancy, E.A., Fischer, G. S.: A soft robotic exomusculature glove with integrated sEMG sensing for hand rehabilitation. In: Proceedings of the 13th International Conference on Rehabilitation Robotics (ICORR), Seattle, WA (2013) 2. Kawamoto, H., Kamibayashi, K., Nakata, Y., Yamawaki, K., Ariyasu, R., Sankai, Y., Sakane, M., Eguchi, K., Ochiai, N.: Pilot study of locomotion improvement using hybrid assistive limb in chronic stroke patients. BMC Neurol. 13, 141 (2013) 3. Polygerinos, P., Wang, Z., Galloway, K.C., Wood, R.J., Walsh, C.J.: Soft robotic glove for combined assistance and at-home rehabilitation. Robot. Auton. Syst. 73, 135–143 (2015) 4. Oguntosin, V., Harwin, W.S., Kawamura, S., Nasuto, S., Hayashi, Y.: Development of a wearable assistive soft robotic device for elbow rehabilitation. In: Proceedings of the 2015 ICORR 11th International Conference on Rehabilitation Robotics, Singapore (2015) 5. Roffman, C.E., Buchanan, J., Allison, G.T.: Predictors of non-use of prostheses by people with lower limb amputation after discharge from rehabilitation: development and validation of clinical prediction rules. J. Physiotherapy 60(4), 224–231 (2014)
106
J. Rossiter et al.
6. Essick, G.K., McGlone, F., Dancer, C., Fabricant, D., Ragin, Y., Phillips, N., Jones, T., Guest, S.: Quantitative assessment of pleasant touch. Neurosci. Biobehav. Rev. 34(2), 192–203 (2010) 7. Knoop, E., Rossiter, J.: The tickler: a compliant wearable tactile display for stroking and tickling. In: Proceedings of the CHI (2015) 8. Knoop, E., Rossiter, J.: A compliant soft-actuator laterotactile display. Smart Mater. Struct. 24, 045034 (2015)
Evaluation of Force Tracking Controller with Soft Exosuit for Hip Extension Assistance Hao Su, Ye Ding, Ignacio Galiana, Jozefien Speeckaert, Nikos Karavas, Philippe Malcolm, Christopher Siviy and Conor J. Walsh Abstract This abstract describes the design and experimental evaluation of a force tracking controller for hip extension assistance utilizing a soft exosuit connected to a tethered off-board actuation system. The new controller aims to improve the force profile tracking capability and demonstrate its advantages over our previously reported work. The controller was evaluated by one healthy participant walking on a treadmill at 1.35 m/s. Results showed that the system can deliver a predefined force profile robustly with a 200 N peak force. The measured peak force value using force controller was 198.7 ± 2.9 N, and the root-mean-squared (RMS) error was 3.4 N (1.7 % of desired peak force). These results indicate that the force control reduces peak force variability and improves force profile tracking capability.
1 Introduction In recent years we have seen exciting results demonstrating that exoskeleton devices can reduce energy cost during walking [1–4]. We have been developing soft wearable robots we call exosuits that are intended to provide a more conformal and compliant means to interface to the human body. An exosuit uses textiles and apparel, resulting in minimal resistance to natural human motion and does not add significant inertia to the lower extremities. In prior work, we have implemented a force-based position control (referred as position controller hereafter) [4–6] to indirectly regulate force to reach the desired peak values. This approach has proved H. Su (&) Y. Ding I. Galiana J. Speeckaert N. Karavas P. Malcolm C. Siviy C.J. Walsh John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, USA e-mail:
[email protected] C.J. Walsh e-mail:
[email protected] H. Su Y. Ding I. Galiana J. Speeckaert N. Karavas P. Malcolm C. Siviy C.J. Walsh Wyss Institute for Biologically Inspired Engineering, Harvard University, Cambridge, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_18
107
108
H. Su et al.
successful to date in our ankle and hip actuation systems where a desired force can be achieved by having the actuator follow a prescribed position profile. The force magnitude is then regulated by monitoring the force level on a step by step basis. However, from other work on wearable robotic systems, the advantage of direct force control is apparent, as it can potentially minimize the peak force variability and precisely track desired force profiles at a joint [7]. This can enable precise torque delivery to understand optimal assistive device design [8]. Further, force control is potentially more robust to variations in gait kinematics as it keeps minimizing force error regardless of joint position. To accurately regulate force delivery in exoskeletons, force control methods like iterative learning-based force control [7] have been studied in rigid devices. However soft exosuits impose particular challenges for force tracking because stiffness of the human-exosuit interface is nonlinear, and the system presents varying system characteristics both from human (e.g. gait variability) and the electromechanical system (e.g. suit stiffness changes and Bowden cable efficiency). To address this, we developed a force tracking controller for hip extension assistance and evaluated its efficacy during treadmill walking.
2 Material and Methods The hip extension soft exosuit system is composed of a mono-articular hip exosuit and a reconfigurable multi-joint actuation platform previously described [6] and the setup is shown in Fig. 1. The textile components of the hip exosuit include a waist belt, two thigh braces and two elastic straps on the lateral side of each leg to prevent migration of the thigh brace [6].
2.1
Sensing and Actuation
Sensors of this soft exosuit system included an Inertial Measurement Unit (IMU) and one load cell per leg. The thigh angle was measured with an IMU Fig. 1 Hip extension soft exosuit system. IMU measures thigh angles. Two sets of Bowden cables are connected between the multi-joint actuation platform and the exosuit to provide hip extension assistance during walking. The assistive force is transmitted from the multi-joint actuation platform to the wearer
Evaluation of Force Tracking Controller with Soft Exosuit
109
Fig. 2 Electromechanical sensing and control system. The timing of the motor force and position control profile is generated in real-time based on sensor input from thigh mounted IMU and load cell
(VN-100, VectorNav Technologies, USA) as shown in Fig. 1. The controller detects maximum hip flexion and the stride time is estimated as the time between two consecutive maximum hip flexion events [6]. The data acquisition card (PCI-6259, NI, USA) acquires the digital IMU and analog load cell signals. Then the controller processes the signal and sends a reference voltage to the Copley motor controller (ACP-090-36 Accelnet, USA) for the Maxon EC-4 pole brushless motor Fig. 2.
2.2
Force Tracking Controller Description
The objective of the force tracking controller is to minimize peak force variability and enable force profile tracking. This is motivated by the desire for precise and robust force manipulation to understand the relation between force delivery and human walking metabolics. The force controller has both position and force control loops to deliver sinusoidal force profile with 200 N peak. The bio-inspired force profile is a scaled version of the biological hip moment, designed to provide a joint torque equivalent to approximately 30 % of the biological hip extension moment.
3 Results The force tracking controller was evaluated by one healthy participant walking on a treadmill at 1.35 m/s. The objective is to evaluate if the controller delivers reliable and accurate force profile to the participant on a step-by-step basis. In a continuous walking session, the participant walked for 2 min using the position controller, and then walked for 2 min using the force controller. The measured force in the last 30 strides was used for analysis. The peak force using position controller was 196.3 ± 11.2 N, and the root-mean-squared (RMS) error was 9.5 N (4.8 % of desired peak force) as depicted in Fig. 3(a). The peak force using force controller was 198.7 ± 2.9 N, and the RMS error was 3.4 N (1.70 % of desired peak force) as depicted in Fig. 3(b). It is apparent that the force controller demonstrated better force tracking performance especially in the 10-20 % gait cycle region comparing Fig. 3(a) and (b).
110
H. Su et al.
(a)
(b)
Fig. 3 Force tracking results using the position controller a and the force controller b. The results were from a healthy participant at 1.35 m/s steady state walking for 30 strides
4 Conclusion We presented a force tracking controller for hip extension assistance. The experimental results demonstrate the effectiveness at steady-state walking in terms of reduced peak force variability and enhanced profile tracking capability. Further evaluation will include force profile tracking results for different subjects with gait change. Acknowledgments This material is based upon the work supported by the Defense Advanced Research Projects Agency (DARPA), Warrior Web Program (W911NF-14-C-0051). This work was also partially funded by the Wyss Institute for Biologically Inspired Engineering and the John A. Paulson School of Engineering and Applied Sciences at Harvard University.
References 1. Malcolm, P., Derave, W., Galle, S., De Clercq, D.: A simple exoskeleton that assists plantarflexion can reduce the metabolic cost of human walking. PLoS ONE 8(2), e56137 (2013) 2. Mooney, L.M., Rouse, E.J., Herr, H.M.: Autonomous exoskeleton reduces metabolic cost of human walking during load carriage. J. Neuroeng. Rehabil. 11, 80 (2014) 3. Collins, S.H., Wiggin, M.B., Sawicki, G.S.: Reducing the energy cost of human walking using an unpowered exoskeleton. Nature 522, 212–215 (2015) 4. Panizzolo, F., Galiana, I., Asbeck, A., Siviy, C., Schmidt, K., Holt, K., Walsh, C.: A biologically-inspired multi-joint soft exosuit that can reduce the energy cost of loaded walking. J. NeuroEng. Rehabi. 13(1), 1 (2016). In press 5. Ding, Y., Galiana, I., Asbeck, A., Santos, T., Araujo, V., Lee, S., Holt, K., Walsh, C.: Biomechanical and physiological evaluation of multi-joint assistance with soft exosuits. IEEE Trans. Neural Syst. Rehabil. Eng. 99, 1 (2016) 6. Ding, Y., Galiana, I., Malcolm, P., Walsh, C.: IMU-based iterative control for hip extension assistance with a soft exosuit. In: IEEE International Conference on Robotics and Automation (ICRA) (2016)
Evaluation of Force Tracking Controller with Soft Exosuit
111
7. Zhang, J., Cheah, C.C., Collins, S.H.: Experimental comparison of torque control methods on an ankle exoskeleton during human walking. In: IEEE International Conference on Robotics and Automation (ICRA 2015), Seattle, WA, pp. 5584–5589 (2015) 8. Jackson, R.W., Collins, S.H.: An experimental comparison of the relative benefits of work and torque assistance in ankle exoskeletons. J. Appl. Physiol. 119, 541–557 (2015)
Part IV
Neural Interfacing of WRs
Endogenous Control of Powered Lower-Limb Exoskeleton Kyuhwa Lee, Dong Liu, Laetitia Perroud, Ricardo Chavarriaga and José del R. Millán
Abstract We present an online decoding method for controlling a powered lower-limb exoskeleton using endogenously generated electroencephalogram (EEG) signals of human users. By performing a series of binary classifications, users control the exoskeleton in three directions: walk front, turn left and turn right. During the first classification phase, the user’s intention to either walk front or change direction is detected. If the user’s intention to change direction is detected, a subsequent classification for turning left or right is performed. Five subjects were able to successfully complete the 3-way navigation task while mounted in the exoskeleton. We report the improved accuracy of our cascaded protocol over a baseline method.
This work was supported by the Swiss National Centers of Competence in Research (NCCR) Robotics project. K. Lee (&) D. Liu L. Perroud R. Chavarriaga J.d.R. Millán Brain-Machine Interface Lab, School of Engineering, Center for Neuroprosthetics, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland e-mail:
[email protected] D. Liu e-mail:
[email protected] L. Perroud e-mail:
[email protected] R. Chavarriaga e-mail:
[email protected] J.d.R. Millán e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_19
115
116
K. Lee et al.
1 Introduction Recent developments in assistive devices enabled people with disability to benefit from more dexterous functionalities than wheelchairs [1, 2]. Among them, upper-limb and lower-limb exoskeletons have been given attention as the potential technology to assist paraplegic or tetraplegic population. Brain-machine interfaces (BMI) have shown the possibility of controlling exoskeletons and neuroprosthetic devices using non-invasive EEG (e.g. [3–5]). A BMI system can be controlled using either exogenous (evoked) or endogenous (spontaneous) signals. In exogenous BMI, evoked signals appear when a person senses external stimuli such as visual or auditory cues (e.g. [3]). The advantage of this approach includes minimal training time. In endogenous BMI, control signals are generated directly by the user without any external stimulation. It is useful for users who have sensory impairments while providing intuitive means of interactions (e.g. [6]). With the advances in materials, actuators, sensors and computer size, many wearable exoskeletons have become real-life products. Some examples of lower-limb exoskeletons include Rex [7], ReWalk [8], ALEX [9], H2 [10] and Indego [11]. We use Rex as our testing platform since it is currently the only available robotic platform that can move independently without the support of the user. It can perform basic movements such as walking front, back and sides, turning left and right, sitting down and standing up. In this preliminary study, we test with 5 healthy subjects to assess the feasibility of applying to end users who have difficulty in manual control.
2 Method 2.1
Hardware Setup
We use an active electrode system (g.GAMMAsys) with Ag/AgCl wet electrodes. g.Tec USBAmp was used to sample the signals in 512 Hz from 16 electrodes placed on the following positions on the scalp: Fz, FCz, Cz, CPz, FC1-4, C1-4, CP1-4. Reference electrode was placed on the right ear lobe and ground electrode was placed on AFz. While in the Rex, subjects follow the visual cue shown on a wearable see-through screen, Google Glass.
2.2
Protocol
During the training stage, 10 trials of each of the following two tasks are performed for 5 s each, one after the other. (a) Move: Imagine moving both hands, (b) Relax: Relax both hands. In both Move and Relax tasks, the subjects are instructed not to
Endogenous Control of Powered Lower-Limb Exoskeleton…
117
move any body part including eyes. Subjects are given a rest period of 10 s between trials to freely move their body. In our protocol, we use a series of cascaded binary classifications to perform a multiclass classification. Our cascaded protocol is defined as following: (1) Cue (1 s), (2) Imagery (1-5 s): classification between walk and turn, (3) Feedback (1 s). If turn intention is detected, the following steps are performed. (4) Imagery (1-5 s): classification between left and right turn, (5) Feedback (1 s). During the imagery period, a bar increases in one of two directions until it reaches either end, where class labels are assigned to both ends. Imagery period ends if the bar reaches a threshold or 5 s has passed, in which case the dominant direction is chosen. We compare our protocol with another commonly used 3-way classification protocol, as used in [13], as a baseline. In this baseline protocol, instead of performing cascaded classification, if the bar is intentionally kept from reaching either side of the ends until timeout occurs, the robot executes the third command.
2.3
Experiment Scenario
In our work, we run both protocols in the setup as shown in Fig. 1. Due to the slow speed of the Rex (4–8 s per action), we design scenarios such that each trial lasts between 4 and 8 min which requires 20–30 motor commands to finish. There are two obstacles placed on each side. Although it may seem trivial, it requires 16 consecutive correct decisions to reach the goal in the optimal path. After running a few practice runs until the user feels comfortable, each subject performs one trial each for both protocols. We test our protocol and the baseline protocol with 5 healthy subjects (1 female/4 males, mean age 29.2 ± 6.2). The same classification method explained in Sect. 2.4 is used in both protocols. All subjects had previous experiences with standard motor imagery (left hand vs right hand movement or hand vs feet movement) but none of them had any experience with the imagery we used in this experiment (both hands vs relax). Fig. 1 (a) A user controlling Rex. (b) Experiment scenario. The user starts in “S” position and tries to reach “G” position. (a)
(b)
118
2.4
K. Lee et al.
Signal Processing and Decoding
The EEG signals are filtered using common average reference (CAR) with 50 Hz notch filter and transformed into power spectral density (PSD) features in the beta band (14–19 Hz) of a 0.5-s sliding window. We use multitaper PSD, which is effective in reducing the estimation bias by averaging over multiple independent estimates from a given sample [12]. The PSD computed for all channels are concatenated to form a feature vector and is used to train a Random Forests classifier (1000 trees, 100 depth limit), which also internally performs feature ranking while training. During the testing phase, the likelihoods are computed in 15 Hz by the classifier and are accumulated over time, represented as the final decision score. The real-time visual feedback is given to the user through a wearable see-through screen. The computed likelihoods Pc ðtÞ (c ∊ {Move, Relax}) at current sampling time t are accumulated over time with a scoring function Sc ðtÞ: Sc ðtÞ ¼ aSc ðt 1Þ þ ð1 aÞPc ðtÞ; Sc ð0Þ ¼ 0:5
ð1Þ
where 0 a\1 is a damping factor to smooth out the noisy output of the classifier, obtained heuristically. We set a ¼ 0:8. Now we define our decision function DðtÞ as: DðtÞ ¼ Dðt 1Þ þ ðSMove ðtÞ SRelax ðtÞÞ; Dð0Þ ¼ 0
ð2Þ
A decision is made and delivered to the robot if jDðtÞj L or t ttimeout , where L is a bar length threshold. The sign of D(t) determines the class label, which is assigned in the beginning of the experiment.
3 Results Table 1 shows the number of errors and total decisions made to fulfill the task. The average completion time was 285 s with the cascaded protocol and 341 s with the baseline protocol, which reflects the lower error rates of the proposed method. It should be noted that we do not know the ground truth label of each decision the user made as each motor command is voluntarily chosen by the user. So instead we estimate a BMI classification error indirectly by counting how many times the chosen action led the robot to move farther from the goal, i.e. assuming the user was always trying to deliver an optimal command. Occasionally, the user gets confused on what motor command to deliver because of the loss of attention or due to the error on path planning. Therefore, estimated errors may be interpreted as an upper bound on the classification error.
Endogenous Control of Powered Lower-Limb Exoskeleton… Table 1 Error estimates. At least 16 correct decisions need to be made to successfully reach the goal position.
ID
119
Proposed
Baseline
Errors
Decisions
Rate
S1
2
19
0.11
Errors 2
Decisions 21
0.10
Rate
S2
0
16
0.00
3
21
0.14
S3
3
20
0.15
4
21
0.19
S4
0
16
0.00
6
21
0.29
S5
1
18
0.06
3
20
0.15
Total
6
89
0.07
18
104
0.17
4 Discussion We have shown a working method for controlling a powered lower-limb exoskeleton using only EEG signals. This work will be extended with more challenging scenarios and larger number of subjects to further validate the feasibility of brain-controlled locomotion.
References 1. Bogue, R.: Robotic exoskeletons: a review of recent progress. Ind. Robot Int. J. 42(1), 5–10 (2015) 2. Millán, J.D.R., et al.: Combining brain–computer interfaces and assistive technologies: state-of-the-art and challenges. Front. Neurosci. 4, 161 (2010) 3. Kwak, N.S., Müller, K.R., Lee, S.W.: A lower limb exoskeleton control system based on steady state visual evoked potentials. J. Neural Eng. 12(5), 056009 (2015) 4. Müller-Putz, G.R., Scherer, R., Pfurtscheller, G., Rupp, R.: EEG-based neuroprosthesis control: a step towards clinical practice. Neurosci. Lett. 382(1), 169–174 (2005) 5. Contreras-Vidal, J.L., Grossman, R.G.: NeuroRex: a clinical neural interface roadmap for EEG-based brain machine interfaces to a lower body robotic exoskeleton. In: EMBC, 2013, pp. 1579–1582 (2013) 6. Pfurtscheller, G., Neuper, C.: Motor imagery and direct brain-computer communication. Proc. IEEE 89(7), 1123–1134 (2001) 7. http://www.rexbionics.com 8. http://www.argomedtec.com 9. Banala, S.K., Kim, S.H., Agrawal, S.K., Scholz, J.P.: Robot assisted gait training with active leg exoskeleton (ALEX). IEEE TNSRE 17(1), 2–8 (2009) 10. Bortole, M., Venkatakrishnan, A., Zhu, F., Moreno, J.C., Francisco, G.E., Pons, J.L., Contreras-Vidal, J.L.: The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study. J. NeuroEng. Rehab. 12(1), 1 (2015) 11. Quintero, H.A., Farris, R.J., Goldfarb, M.: Control and implementation of a powered lower limb orthosis to aid walking in paraplegic individuals. In: ICORR, 2011, pp. 1–6 (2011) 12. Percival, D.B., Walden, A.T.: Spectral Analysis for Physical Applications. Cambridge University Press, Cambridge (1993) 13. Carlson, T., Millan, J.D.R.: Brain-controlled wheelchairs: a robotic architecture. IEEE Robot. Autom. Mag. 20, 65–73 (2013)
Natural User-Controlled Ambulation of Lower Extremity Exoskeletons for Individuals with Spinal Cord Injury Kiran Karunakaran, Ghaith Androwis and Richard Foulds
Abstract Natural-quality, independent ambulation is a prerequisite for community use of lower extremity exoskeletons by individuals with disabilities. In general, current exoskeletons generate pre-programmed gait, where the user cannot exercise volitional control necessary to navigate over uneven surfaces and avoid obstacles. This project introduces an intuitive control strategy that allows the user to determine and sense the exoskeleton movement in real time using trajectories produced by the hands. The concept allows neurally defined ambulation control to be expressed through alternative biological articulators. This novel approach uses admittance control to compute each exoskeleton’s foot position from Cartesian forces exerted by the user’s hand on a trekking pole that is connected to foot through a multi-axis load cell. The algorithm has been evaluated by naïve, non-disabled users who walked a 10 degree of freedom, ½ scale biped robot on a treadmill. The results show that the algorithm produced robot-generated gait kinematics that are similar to human gait kinematics. A human-scale exoskeleton has been developed to allow further exploration of this control method.
Research funded by the Gustavus and Louisa Pfeiffer Research Foundation and the NJIT/Kessler Rehabilitation Engineering Research Center on Wearable Robots for Independent Living (NIDILRR-HSS grant 90RE5021). K. Karunakaran G. Androwis R. Foulds (&) New Jersey Institute of Technology, Newark, USA e-mail:
[email protected] K. Karunakaran e-mail:
[email protected] G. Androwis e-mail:
[email protected] G. Androwis Kessler Research Foundation, West Orange, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_20
121
122
K. Karunakaran et al.
1 Introduction Individuals with paraplegia due to spinal cord injury (SCI) have impaired motor control and sensory feedback which can significantly reduce their ability to walk. The preference of many of these people for upright ambulation [1, 2] over wheelchair use has driven the development and commercialization of advanced rehabilitation exoskeletons. [3] However, for the new technology to effectively augment or even replace wheelchair use, the performance of the user/exoskeleton in community living must approach that of unimpaired ambulation. [4] Thus, the exoskeleton must operate in a way that reflects the user’s needs in the real world. In this paper, we present a novel form of human/robot cooperation that offers the potential to allow users exceptional control over their exoskeleton-assisted walking. We suggest that rather than to detect neurally-coded biological walking intentions via a brain-computer interface(BCI), or endow an exoskeleton with autonomous gait control that can respond to environmental conditions, the user’s natural ambulation can be expressed and sensed through intact articulators and used to control the exoskeleton’s movement patterns. Currently reported user/exoskeleton control strategies have attempted to detect neural signals that represent the user’s gait intentions. This work includes the use of cortically implanted and surface electrodes, as well as hybrid EEG/EMG signal processing. [5–8] To date, the decoding of detailed lower extremity motor control signals has been elusive, with most BCI approaches generating very coarse notions of intention, or on/off triggers of actions that are carried out by the computer controlled exoskeleton. Alternatives to BCI control include the use of sensors to detect ambulation cues from the tilt of the user’s torso, arms and crutches, or the movement of the user’s center of pressure or center of mass. These signals are used to trigger different phases of exoskeleton gait. [9, 10] Neither the BCI nor the sensor-based strategies produce gait that resembles unimpaired human walking, and are not yet adequate for daily use in the real world.
2 Surrogate Articulation of Gait Our work explores the use of an individual’s neurally-planned gait intention, but unlike BCI methods, allows those intentions to be biologically processed to produce natural, walking-like movements of surrogate articulators. We have shown that such movements can be faithfully produced by user’s fingers and hands. [11, 12] It is not known if these controlled movements are linked to ambulation motor areas in the brain, or if there is connectivity between lower and upper extremity cortical control that typically facilitates coordinated movement. Regardless, users have demonstrated fast learning of ‘hand-walking’ that appears quite intuitive.
Natural User-Controlled Ambulation of Lower Extremity Exoskeletons … Fig. 1 ½ scale robot equipped with trekking poles and 3 axis force sensors that is controlled by the walking-like motions of the user’s hands
123
2 DOF hip Trekking pole
1 DOF knee
2 DOF ankle 3 axis force sensor
3 Methods 3.1
Experimental Apparatus
A 1/2 scale biped robot prototype supported by a frame was built to test the hand control mechanism. Each robot leg has 2 links, from hip to knee, and from knee to ankle complying with the 50 % scaled anthropometrics of the human leg. Dynamixel MX series smart servos provide two degrees of freedom (DOF) at the hip, one at the knee and two at the ankle. A foot was designed with provisions to mount force sensors to detect ground contact, and also to attach a 3 DOF force sensor adjacent to the foot of each leg. A trekking pole (carbon fiber rod) is mounted on the 3DOF force sensor and extends upward to the height of the robot’s waist (Fig. 1).
3.2
Admittance Control of Hand-Walking
The user grasps the trekking poles on the ipsilateral side and makes walking-like motions to produce the kinematics similar to the foot. The rigid poles transmit the user’s applied forces in the direction intended to the 3 DOF force sensor on the ipsilateral foot (Fig. 2). An admittance control algorithm converts these forces into Cartesian position of the exoskeleton foot. The Cartesian forces in x and y are sampled at 100 Hz and converted to predicted Cartesian positions and velocities for each 10 ms sample in sagittal plane. This is accomplished by our admittance loop that implements the equation of motion in (1) for all directions using CVode (Ordinary Differential Equation Solver) developed at Eindhoven University [13]. The Cartesian foot position values are converted to the corresponding joint angles and angular velocities of the knee and hip using custom developed inverse kinematics. All angles and velocities as inputs are fed to the corresponding Dynamixel motors to generate the required torque as the output to perform the intended
124
K. Karunakaran et al.
Fig. 2 ½ scale robot ‘walked’ on a treadmill by non-disabled subjects using their hands to produce foot movement of the robot. The biped was supported by a frame. Subjects easily accommodated unexpected treadmill speed changes, and mock obstacles. Performance was not degraded when subjects experienced a cognitive load
movement. The process also allows the user’s hands to feel the movement and ground impact of the exoskeleton’s feet, providing essential feedback. 0
00
X ðt Þ ¼
F ðtÞ B X ðtÞ M M
ð1Þ
Where F (N) is the applied force vector in x and y, M (kg) is the desired mass to be felt by the user (this is the admittance coefficient), B (Ns/m) is the desired 0 00 damping to be felt by the user, and X (m/s) and X (m/s2) are the velocity and acceleration of the robot respectively.
4 Results The results show that the complete gait cycle produced by the control algorithm resembles the normal gait trajectory as shown in Fig. 3. The frame duration of 10 ms that allows an admittance loop at 100 Hz is shown to be sufficient for human operators to feel smooth, nearly passive movements of a robot. Subjects successfully walked the robot on a treadmill with little to no learning period. They produced consistent ankle trajectories in the presence of unexpected changes in x and y Position of Biped Foot
y position (cm)
y position (cm)
6 4 2 0
-20
-18
-16
-14
-12
-10
-8
x position (cm)
-6
-4
-2
0
25
x and y Position of Human Foot
20 15 10 5 -35 -30 -25 -20 -15
-10
-5
0
x position (cm)
5
10
15
Fig. 3 Comparison of the sagittal plane ankle trajectories of the robot foot (left) and human foot (right) when walking on the treadmill. The shapes are quite similar, with a vertical/horizontal excursion ratio of approximately .275 in each case. Direction of walking is to the right
Natural User-Controlled Ambulation of Lower Extremity Exoskeletons …
125
treadmill speed, the presence of obstacles, and while performing an unrelated cognitive task (counting backwards). Our future work involves incorporating balance of the exoskeleton to the current control mechanism.
References 1. Ditunno, P.L., Patrick, M., Stineman, M., Ditunno, J.F.: Who wants to walk? Preferences for recovery after SCI: a longitudinal and cross section study. Spinal Cord 467, 500–506 (2008) 2. Esquenazi, A., et al.: The ReWalk powered exoskeleton to restore ambulatory function to individuals with thoracic-level motor-complete spinal cord injury. Am. J. Phys. Med. Rehabil. (2012) 3. Strickland, E.: Good-bye, wheelchair. IEEE Spectr. 49(1), 30–32 (2012) 4. Kilgore, K.L., et al.: Neuroprosthesis consumers’ forum: consumer priorities for research directions. J. Rehabil. Res. Dev. 38(6), 655–660 (2001) 5. Chéron, G., Duvinage, M., De Saedeleer, C., Castermans, T., Bengoetxea, A., Petieau, M., Sylos Labini, F.: From spinal central pattern generators to cortical network: integrated BCI for walking rehabilitation. Neural Plasticity (2012) 6. Ruiz, A.F., Forner-Cordero, A., Rocon, E., Pons, J.L.: Exoskeletons for rehabilitation and motor control. In: The First IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics, BioRob 2006, pp. 601–606. IEEE, February 2006 7. Frolov, A.A., Biryukova, E.V., Bobrov, P.D., Mokienko, O.A., Platonov, A.K., Pryanichnikov, V.E., Chernikova, L.A.: Principles of neurorehabilitation based on the brain-computer interface and biologically adequate control of the exoskeleton. Hum. Physiol. 39(2), 196–208 (2013) 8. Jose, L., Kilicarslan, A.: Non-invasive brain-machine interfaces to powered exoskeletons for restoration of walking 9. Strausser, K.A., Kazerooni, H.: The development and testing of a human machine interface for a mobile medical exoskeleton. In: 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems (2011) 10. Quintero, H.A., Farris, R.J., Goldfarb, M.: A method for the autonomous control of lower limb exo-skeletons for persons with paraplegia. J. Med. Devices 6(4) (2012) 11. Birmingham, C., Noesner, M., Foulds, R.: Haptic controlled functional electrical stimulation of the lower extremities. In: 2003 IEEE 29th Annual Proceedings of Bioengineering Conference, pp. 287–288. IEEE, March 2003 12. Karunakaran, K., Abbruzzese, K., Xu, H., Ehrenberg, N., Foulds, R.: Haptic proprioception in a virtual locomotor task. In: 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 3594–3597, August 2014 13. Van Riel, N.: Speeding up simulations of ODE models in Matlab using CVode and MEX files. Eindhoven University of Technology (2012)
Real-Time Modeling for Lower Limb Exoskeletons Guillaume Durandau, Massimo Sartori, Magdo Bortole, Juan C. Moreno, José L. Pons and Dario Farina
Abstract Real-time electromyography (EMG) driven musculoskeletal (NMS) modeling estimates internal body biomechanical parameters and motor intentions. This is central for understanding the dynamics of user-exoskeleton interaction and for developing closed-loop user-exoskeleton interfaces that are intuitive and effective in promoting neuroplasticity. This abstract, presents methods and results behind the interfacing between a six degree of freedom lower limb exoskeleton (H2 exoskeleton, Technaid S.L., Spain) and a real-time EMG-driven NMS model of the human lower extremity.
1 Introduction In the past decade the development of exoskeleton technology flourished. Current solutions feature state of the art hardware mechatronics. However, the available human-machine interfaces (HMI) permit little user control. These are mostly based on position controllers using pre-recorded movement data or zero impedance controllers that minimize user-exoskeleton interaction forces. Examples based on position controllers include the Lokomat (Hocoma, Switzerland), ReWalk (ReWalk, Israel), or Ekso (Ekso Bionics, USA). Whereas, the H2 (Technaid S.L., Spain) is an example of zero-impedance controller. Position controllers disregard the user’s motion intention thus preventing active user participation. In this context, the user would passively train with limited engagement and presumably limited induction of neuroplasticity. Zero-impedance controllers provide support only if the user is capable of producing This work was supported by the ERC Advanced Grant DEMOVE [267888]. G. Durandau (B) · M. Sartori · D. Farina Institute of Neurorehabilitation Systems, Universitätmedizin Göttingen, Göttingen, Germany e-mail:
[email protected] M. Bortole · J.C. Moreno · J.L. Pons Cajal Institute, CSIC, Madrid, Spain J.L. Pons Tecnológico de Monterrey, Monterrey, Mexico © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_21
127
128
G. Durandau et al.
detectable forces and movements [1]. In this context, severely impaired individuals are excluded as target patients. Therefore, there is the need for a new class of HMIs that can access internal biomechanical variables in vivo in the intact patient. In this abstract, we show the use of real-time EMG-driven musculoskeletal (NMS) modeling to gain access to a repertoire of internal biomechanical variables including muscle-tendon length, moment arms, force as well as the resulting net joint torque in multiple degrees of freedom (DOFs) in the human lower extremity. We show how this can be achieved using a fully-embedded solution based on low-power, wearable and portable systems, i.e. Raspberry PI 2 (Raspberry Pi Foundation, UK) and six-DOF H2 exoskeleton.
2 Method 2.1 Real-Time EMG-Driven NMS Modeling The real-time modeling framework extends from the offline Calibrated EMGinformed Neuromusculoskeletal Modeling (CEINMS) toolbox [2] and from the Multidimensional Cubic B-Spline (MCBS) software [3]. Our proposed framework combines CEINMS and MCBS in a stand-alone forward dynamics approach (Fig. 1). It determines musculoskeletal forces as a function of measured EMGs and joint angles [4]. Joint angles inform MCBS for the real-time computation of muscle-tendon length (Lmt) and moment arms (Rmt). This uses the OpenSim [5] application programming interface for the offline computation of the LMT nominal values needed for generating B-Spline coefficients. Experimental EMG and predicted Lmt and Rmt inform the CEINMS module for the real-time computation of muscle-tendon forces and joint torques. Model internal parameters that vary non-linearly across individuals anthropometrics are estimated using an offline calibration procedure. This adjusts internal parameters to minimize the difference between predicted and experimental joint torques across a range of calibration movement trials.
2.2 Interface with the H2 Lower-Limb Exoskeleton The H2 is a lower limb exoskeleton with six DOFs including left/right hip flexionextension, knee flexion-extension, and ankle plantar-dorsi flexion. Each joint can be controlled both in position and torque. The H2 sensor information is transmitted on a CAN (Controller Area Network) bus and includes: joint position, motor torque and interaction torque. Each joint is actuated by an electric motor and an harmonic drive that can provide a continuous net torque up to 35 Nm or a peak torque of 180 Nm for short time periods [6]. The H2 was first interfaced with the RP2 via CAN. We implemented a graphical user interface (GUI) running on a separate desktop
Real-Time Modeling for Lower Limb Exoskeletons
129
Fig. 1 Diagram of the interface and interaction between the EMG driven NMS model and the H2 lower-limb exoskeleton
PC connected to the RP2 using an UDP connection (WIFI or LAN) for real-time data monitoring. This GUI plots several real-time variables including: processed EMGs, torques computed by the EMG-driven NMS model and H2 sensor and control information (Fig. 1). This approach (Fig. 1) enables operating the EMG-driven NMS model as part of an open-loop or a closed-loop formulation. The open-loop formulation extract real-time biomechanical variables during natural locomotion or during locomotion wearing the H2. In this application, EMG-driven model based estimates are not used to close the loop with the H2 for control purposes. The GUI displays real-time body parameters and how they are altered as a result of humanmachine interaction. The closed-loop formulation uses model-based computed joint torque to control the exoskeleton (Fig. 1).
2.3 Experimental Protocol and Tests EMGs are recorded (2048 Hz) using a 256-channel EMG amplifier (OTBioelettronica, Italy) from 9 muscle groups: rectus femoris, vastus medialis and lateralis, lateral and medial hamstrings, gastrocnemious medialis and lateralis, soleus and tibialis anterior. Reference joint angles and torques are recorded using the H2 exoskeleton built-in sensors (i.e. sampled at 100 Hz). Furthermore, the developed software is interfaced in real-time with a seven-camera motion capture system (Qualysis, Sweden) at 256 Hz and two in-ground force plates (Bertec, USA) at 2048 Hz. In this study, we used the motion capture system and the force plate to compute joint torques and angles via real-time inverse dynamics and kinematics. We initially used these data
130
G. Durandau et al.
Fig. 2 Real-time torque from inverse dynamic (ID) and from CEINMS of the ankle joint from 4 squat trials
for the EMG-driven model calibration. After calibration, we validate the EMG driven model in real-time on a new set of trials not used for calibration. In Fig. 2, we can see the result of the pilot test for the squat trial. Model-based prediction error was quantified as the root mean squared error (RMS) between experimental and predicted joint moments during a set of validation trials that were not used for calibration. RMS about the ankle and knee DOFs ranged between 13.8 and 12.55 Nm over a maximal amplitude of 79.33 and 92 Nm. This was measured from one subject who performed a squat/calf rise trials. The average Processing time was 3.938e+01 ± 54.15 % ms for the NMS model, 4.136e−01 ± 43.6 % ms for the cubic B-spline computation and 4.830e+01 ± 27.8 % ms for the total delay.
3 Conclusion Future work will focus on the real-time model-based control of the H2 exoskeleton system. This will exclusively use H2 sensory data for model calibration and real-time operation. Different control schemes will be tested including those in the torque and impedance domains. Our proposed model-based control scheme will allow development of efficient and user specific neurorehabilitation schemes. Moreover, it will enable quantifying the progress of the patient during therapy helping the physician to decide to change the rehabilitation strategies and can also help to asses in the long term the effectiveness of the therapy. Finally, the model can easily be improved for computing new body parameters in real-time like joint load, stiffness [7], muscle energy consumption and muscle primitives [8].
References 1. Hidler, J., Nichols, D., Pelliccio, M., Brady, K., Campbell, D.D., Kahn, J.H., Hornby, T.G.: Multicenter randomized clinical trial evaluating the effectiveness of the lokomat in subacute stroke. Neurorehabil. Neural Repair 23(1), 5–13 (2009)
Real-Time Modeling for Lower Limb Exoskeletons
131
2. Pizzolato, C., Lloyd, D.G., Sartori, M., Ceseracciu, E., Besier, T.F., Fregly, B.J., Reggiani, M.: Ceinms: a toolbox to investigate the influence of different neural control solutions on the prediction of muscle excitation and joint moments during dynamic motor tasks. J. Biomech. 48(14), 3929–3936 (2015) 3. Sartori, M., Reggiani, M., van den Bogert, A.J., Lloyd, D.G.: Estimation of musculotendon kinematics in large musculoskeletal models using multidimensional b-splines. J. Biomech. 45(3), 595–601 (2012) 4. Sartori, M., Reggiani, M., Farina, D., Lloyd, D.G.: Emg-driven forward-dynamic estimation of muscle force and joint moment about multiple degrees of freedom in the human lower extremity. PloS One 7, 12 (2011) 5. Delp, S.L., Anderson, F.C., Arnold, A.S., Loan, P., Habib, A., John, C.T., Guendelman, E., Thelen, D.G.: Opensim: open-source software to create and analyze dynamic simulations of movement. IEEE Trans. Biomed. Eng. 54(11), 1940–1950 (2007) 6. Bortole, M., Venkatakrishnan, A., Zhu, F., Moreno, J.C., Francisco, G.E., Pons, J.L., ContrerasVidal, J.L.: The h2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study. J. Neuroeng. Rehabil. 12(1), 54 (2015) 7. Sartori, M., Maculan, M., Pizzolato, C., Reggiani, M., Farina, D.: Modeling and simulating the neuromuscular mechanisms regulating ankle and knee joint stiffness during human locomotion. J. Neurophysiol. 114(4), 2509–2527 (2015) 8. Sartori, M., Gizzi, L., Lloyd, D.G., Farina, D.: A musculoskeletal model of human locomotion driven by a low dimensional set of impulsive excitation primitives. Frontiers Comput. Neurosci. 7 (2013)
Towards Everyday Shared Control of Lower Limb Exoskeletons Tom Carlson
Abstract We take a multi-faceted whole-system approach towards designing and implementing a neural interface for lower limb exoskeletons. This paper highlights some of the initial steps we have taken, from the development of shared control techniques to the characterization of the exoskeleton itself and the identification of brain signals that could be used in such an interface.
1 Introduction Wearable lower limb exoskeletons are designed to help people with paralysis walk, offering them more independence than conventional wheelchairs. Several such exoskeletons are now commercially available, but most require the user to balance themselves with crutches or some other external support. The Rex (robotic exoskeleton by Rex Bionics, NZ) offers full stability support so people who also have upper limb impairments can use it. However, the joystick interface and pre-set motions make it challenging to operate effectively in confined and cluttered environments. Several research groups are striving to achieve a more natural control of such exoskeletons by designing neural interfaces [1]. We believe that despite recent advances, current neural decoding resolutions and accuracies are not yet sufficient for everyday control of these exoskeletons outside laboratory environments. Therefore, the Aspire Create team takes a multi-faceted whole-system approach towards designing and implementing a shared control system that will work in tandem with the neural interface.
T. Carlson (&) Aspire Create, Royal National Orthopaedic Hospital, University College London, London HA7 4LP, UK e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_22
133
134
T. Carlson
Fig. 1 (a) Gait analysis of the Rex using the Motek Medical Grail. (b) Measuring the physical human-exoskeleton interaction forces. (c) Simultaneously measuring interaction forces and muscular activity.
2 Shared Control We are extrapolating our extensive experience in brain-controlled smart wheelchairs, such that the control authority will be shared between the user and the device [2]. The exoskeleton will therefore be able to: perceive its surroundings; use this information to interpret the user intent in this context; and then provide the user with an appropriate level of assistance in completing their desired task.
3 Lower Limb Exoskeletons Our first steps towards implementing such a system, have been to characterize: the exoskeleton motion; the physical human-robot interactions; and the user’s brain activity (Fig. 1). We first performed a gait analysis of the Rex using the Motek Medical Grail system at the Royal National Orthopaedic Hospital, Stanmore, UK. Not only did this allow us to compare the gait with a typical human gait and identify specific differences, such as the dorsi-/plantar- flexion [3], but it also enabled us to build a kinematic model that will be used by the shared controller. Next, we analyzed the profiles of the interaction forces between the lower limbs and the exoskeleton cuffs, which we found to differ significantly for each motion primitive [4]. This real-time information could be exploited both for safety and in the feedback control loop. Our most recent preliminary electroencephalography (EEG) results show significant differences in the mu band power over the motor cortex, when able-bodied participants wanted to change their state (e.g. stand up, take a step, turn in place etc.) [5].
Towards Everyday Shared Control of Lower Limb Exoskeletons…
135
4 Discussion and Future Work Combining the results discussed here, we will implement a shared control system, such that the exoskeleton will be able to interpret the user’s intent (from EEG)— given the context of the situation—and then generate appropriate sequences of movements to achieve the desired task safely. Acknowledgments The author would like to thank G. Barbareschi, C. Holloway, R. Loureiro, JdR. Millán, DZ. Morgado Ramirez, A. Rathore, E. Sanchez, R. Richards, M. Thornton, M. Wilcox and A. Zervudachi for their contributions to this ongoing project and Rex Bionics for providing the exoskeleton for our experiments.
References 1. Kilicarslan, A., Prasad, S., Grossman, R.G., Contreras-Vidal, J.L.: High accuracy decoding of user intentions using EEG to control a lower-body exoskeleton. In: Proceedings of the Conference of IEEE Engineering in Medicine and Biology Society (EMBC), pp. 5606–5609 (2013) 2. Carlson, T., Millán, J.D.R.: Brain-controlled wheelchairs: a robotic architecture. IEEE Rob. Autom. Mag. 20, 65–73 (2013) 3. Barbareschi, G., Richards, R., Thornton, M., Carlson, T., Holloway, C.: Statically vs dynamically balanced gait: analysis of a robotic exoskeleton compared with a human. In: Proceedings of the Conference of IEEE Engineering in Medicine and Biology Society (EMBC), pp. 6728–6731 (2015) 4. Rathore, A., Wilcox, M., Ramirez, D.Z.M, Loureiro, R., Carlson, T.: Quantifying the human-robot interaction forces between a lower limb exoskeleton and healthy users. In: Proceedings of the Conference of IEEE Engineering in Medicine and Biology Society (EMBC) (2016) 5. Zervudachi, A., Sanchez, E., Carlson, T.: Preliminary EEG characterisation of intention to stand and walk for exoskeleton applications. In: Proceedings of the Conference of NeuroRehabilitation (ICNR) (2016)
Part V
Biomechanics and Neurophysiological Studies with WRs
Joint-Level Responses to Counteract Perturbations Scale with Perturbation Magnitude and Direction Mark Vlutters, Edwin H.F. van Asseldonk and Herman van der Kooij
Abstract To realize a lower extremity exoskeleton that can provide balance assistance in a natural way, an understanding of human balance control is a necessity. In this study, we investigated how the angle, torque and power of the ankle, knee and hip joints changed in response to balance perturbations during walking. Nine healthy young adults walked on an instrumented treadmill and received pelvis perturbations of various magnitudes and directions at the instance of toe-off right. An open source musculoskeletal modeling package (OpenSim) was used to perform inverse kinematics and inverse dynamics. Subjects modulated the ankle torque in the (left) stance foot with the magnitude and direction of the perturbation. Also in gait phases following foot placement, subjects addressed ankle torques to mitigate the remaining effects of the perturbation. The results presented here support the use of ankle actuation in lower extremity orthoses for natural and cooperative balance control.
1 Introduction Most current exoskeletons are unable to stay upright without assistance and guidance of its user. Paraplegic users, for example, often require crutches to prevent falling. To have an exoskeleton assist its user in maintaining balance instead,
This work was supported by the BALANCE (Balance Augmentation in Locomotion, through Anticipative, Natural and Cooperative control of Exoskeletons) project, partially funded under grant 601003 of the Seventh Framework Program (FP7) of the European Commission (Information and Communication Technologies, ICT-2011.2.1). M. Vlutters (&) Biomechanical Engineering Group, University of Twente, Enschede, Netherlands e-mail:
[email protected] E.H.F. van Asseldonk H. van der Kooij University of Twente, Enschede, Netherlands e-mail: {e.h.f.vanAsseldonk,h.vanderkooij}@utwente.nl © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_23
139
140
M. Vlutters et al.
preferably in a natural and human-like manner, an understanding of human balance control is of major importance. Investigating balance responses following perturbations can give insight in the human balance controller. When and how are the different lower extremity joints controlled to maintain upright posture? Furthermore, insight into joint angles, torques, and power during perturbation recovery can provide guidelines for exoskeleton hardware specifications. In this study, pelvis perturbations were used to elicit balance recovery responses in walking human subjects. We investigated how joint-level responses alter with perturbation magnitude and direction, and provides insight in the ranges of motion, torque, and power of the ankle, knee, and hip joints during the recovery.
2 Materials and Methods 2.1
Experimental Setup and Protocol
Nine healthy young adults walked on a custom dual-belt instrumented treadmill (MotekForce Link, Culemborg, Netherlands), while keeping arms crossed over the abdomen. Walking speeds were 2.25 km/h and 4.50 km/h, scaled to the subject’s leg length. For each speed, subjects first walked a 2 min unperturbed baseline trial. In subsequent trials, perturbations were applied to the pelvis using one of two motors (Moog, Nieuw-Vennep, Netherlands), connected with a lever arm to a pelvic brace (Distrac, Hoegaarden, Belgium), worn by the subject. The motors were located at the side and the rear of the treadmill. Perturbations were applied at randomly selected instances of toe-off right (TOR), and consisted of 150 ms block pulses of a force magnitude equal to 4, 8, 12 and 16 % of the subject’s body weight. Perturbation directions were forward (positive), backward (negative), outward (positive, away from stance leg) and inward (negative, toward stance leg). Each perturbation type was repeated 8 times, leading to 256 perturbations per subject. Kinematic data of various landmarks on the lower extremities, pelvis, trunk and head [1] were collected at 100 Hz using a motion capture system (Phoenix Technologies Inc, Vancouver, Canada). Ground reaction forces, subject-motor interaction forces, and EMG data were collected at 1000 Hz. More details about the experimental setup can be found in [2].
2.2
Data Processing
Data were processed using Matlab (R2014b, Mathworks, Natick, US) and OpenSim 3.3 [3]. Joint angles and velocities were calculated using inverse kinematics (IK), joint torques using inverse dynamics (ID), and joint power by multiplying the joint velocities and torques. The model used for the IK and ID calculations in OpenSim
Joint-Level Responses to Counteract Perturbations Scale
141
was the default gait2354 model, scaled to subject specific dimensions using the kinematic landmarks from each subject’s static measurement. In the ID, the collected ground reaction forces as well as the subject-motor interaction forces were specified as external force. All data were cut in sequences from (1) perturbation onset (at TOR) to perturbation end, (2) perturbation end to the subsequent heel strike right (HSR), (3) HSR to subsequent toe-off left (TOL), (4) TOL to subsequent heel strike left (HSL), (5) HSL to subsequent TOR. For each subject, all sequences were resampled to 50 samples and averaged over the repetitions. The repetition averages of each subject were used to obtain averages and standard deviations over subjects.
3 Results Here, results are only shown for the plantar- and dorsiflexion ankle torque and ankle power, in response to anteroposterior (AP) perturbations for the 2.25 km/h walking speed (Fig. 1). Subjects scaled their left ankle torque with perturbation magnitude and direction during the left single support phase, directly following the perturbation. During this phase subjects extracted energy following forward perturbations, and injected energy following the larger backward perturbations, as can be observed from the power. Using the right ankle, subjects delivered an increased plantarflexion torque directly following HSR in response to the larger forward perturbations, resulting in increased energy extraction (negative power). In the second double support phase following the perturbation (HSL-TOR), subjects generated more positive power following backward perturbations as compared to forward perturbations. Surprisingly, the perturbation effects are more pronounced in the right ankle during this second double
Fig. 1 Ankle plantar- and dorsiflexion torques (A) and power (B) in response to anteroposterior pelvis perturbations during 2.25 km/h walking. Top row: left ankle, bottom row: right ankle. Colors indicate the different perturbation magnitudes. Shaded gray area indicates the baseline standard deviation. Standard deviations of the perturbation data are not shown to prevent image cluttering. Data were made dimensionless using subject weight (m*g) and height (l).
142
M. Vlutters et al.
support phase, than in the power delivered by the left ankle during the first double support phase (HSR-TOL) after the perturbation.
4 Discussion Joint angles, torques, and power were obtained using IK and ID. The ankle torques and power show that an ankle strategy is actively addressed in the recovery from pelvis perturbations during walking. One might expect a decreased plantarflexion torque during the push-off directly following forward perturbations (Pert.end-HSR), but an increase was observed instead. An explanation could be that subjects attempt to keep their center of mass at approximately the same height to prevent having to strongly redirect the body vertically [4], or prevent forward body rotation as in [5]. Both require leg extension through plantarflexion. The strong decrease in left ankle plantarflexion torque directly following the perturbation (Pert.end-HSR) allows subjects to quickly regain forward velocity and return to the desired gait cycle. Consequently, no strong ankle torque deviations are observed in the subsequent gait phases following backward perturbations. The larger variability between conditions in right peak ankle power (HSR-TOL) compared to left peak ankle power (HSL-TOR) might be related to the subject repositioning on the treadmill, which likely does not occur until the second step (HSL).
5 Conclusions The presented results can give insight in human balance control on a joint level. Future work consists of finding controllers that can generate such joint-level responses.
References 1. Dumas, R., Chèze, L., Verriest, J.P.: Adjustments to McConville et al. and Young et al. body segment inertial parameters. J. Biomech. 40, 543–553 (2007) 2. Vlutters, M., van Asseldonk, E.H.F., van der Kooij, H.: Center of mass velocity based predictions in balance recovery following pelvis perturbations during human walking. J. Exp. Biol. 209 (2016, to be published) 3. Delp, S.L., et al.: OpenSim: open-source software to create and analyze dynamic simulations of movement. IEEE Trans. Biomed. Eng. 54, 1940–1950 (2007) 4. Donelan, J.M., Kram, R., Kuo, A.D.: Mechanical work for step-to-step transitions is a major determinant of the metabolic cost of human walking. J. Exp. Biol. 205, 3717–3727 (2002) 5. Pijnappels, M., Bobbert, M.F., van Dieën, J.H.: How early reactions in the support limb contribute to balance recovery after tripping. J. Biomech. 38, 627–634 (2005)
Metabolic Energy Consumption in a Box-Lifting Task: A Parametric Study on the Assistive Torque Mohammad S. Shourijeh, Moonki Jung and Michael Damsgaard
Abstract This study showcases effect of adding assistive torques to the hip, knee, and ankle joints in the sagittal plane on the total human body metabolic energy expenditure using the AnyBody musculoskeletal modeling system. To this goal, a box-lifting task was targeted and metabolic energy was computed for several cases including when each of the three joints was assisted at a time. Simulation results showed that the hip joint assistance affects the total metabolic energy consumption more than the knee and ankle joints.
1 Introduction Metabolic energy consumption has been shown to be a serious criterion in human movements. Simulation studies have used this measure to provide information on optimality level of human motion [1]. Rate of VO2 consumption, which is deemed to be equivalent to overall rate of metabolic energy consumption, have been simulated in musculoskeletal dynamic simulations by means of simple [2] to complex [3] mathematical models. Additionally, metabolic energy has been used and speculated to be important in efficacy of exoskeletons. How an assistive device contributes to both the local joint-level and total human body has been a design objective for the This work has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 680754 (The MovAiD project, www.movaid. eu). M.S. Shourijeh (&) M. Jung M. Damsgaard AnyBody Technology, Aalborg, Denmark e-mail:
[email protected] M. Jung e-mail:
[email protected] M. Damsgaard e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_24
143
144
M.S. Shourijeh et al.
neuro-rehabilitation engineers. Experimental measurement of VO2 consumption have been used for assessing performance of human body with and without wearing assistive devices. However, VO2 rate only reflects the total metabolic energy rate and is neither informative about effects at the local muscles or joints nor the contribution of the local elements to the total metabolic energy consumption. Therefore in design of assistive devices, if the question is to evaluate efficiency of the device at the joint level or in some applications, essentially which joint must be assisted most, computer simulations have the potential to play an important role. The goals of this study were (1) to exemplify use of musculoskeletal modeling for assistive wearables in a box-lifting task and (2) to determine which joint is the most crucial to be assisted within the desired movement.
2 Methods AnyBody modeling software [4] has been used to simulate the box-lifting task. Metabolic energy rate model was a modified model of [3]. Ground reaction forces were predicted by the reaction force prediction tool in AnyBody to keep the model dynamic equilibrium; for more details, please see [5].
2.1
Musculoskeletal Model (MSM)
The full human body model in the AnyBody Managed Model Repository (AMMR v1.6.3) was used for this study; see Fig. 1. The model includes 63 segments, and 834 muscles. The model was 75.6 kg and 1.80 m. Three-element Hill-type muscle model was chosen for the muscles, which consists of a contractile element (CE), a parallel elastic element (PE), and a serial elastic element (SE) [6].
Fig. 1 The musculoskeletal model in AnyBody applied for the box-lifting task. Ground reaction forces and moments are predicted by AnyBody
Metabolic Energy Consumption in a Box-Lifting Task …
2.2
145
Box-Lifting Movement
The vertical kinematics of the box was prescribed by an assumed 6th order polynomial with zero velocity and acceleration for the initial and final points. Following that, the kinematics of the model was predicted using the center of mass driver in AnyBody while keeping the center of pressure within the model feet region in order to maintain model balance.
2.3
Metabolic Energy
The model for the muscle metabolic power E_ was defined as _ E_ ¼ H_ þ W
ð1Þ
which models the decrease in the muscle internal energy. In Eq. (1), H_ is the heat rate generated by the muscle defined as the sum of basal, activation, maintenance, and shortening/lengthening heat rates. W_ is the external work done by the muscle and defined as W_ ¼ fce ðlce ; vce ; aÞvce
ð2Þ
where lce and vce are length and velocity of the contractile element (CE) in the muscle model, respectively; fce is the CE force function and a is muscle activation. The total body metabolic energy Etot was defined as Etot ¼
N Z X i¼1
tf
E_ i dt
ð3Þ
t0
where t0 and tf are initial and final movement time, which were set to 0 and 3 s, respectively; i refers to muscle index, and N is the total number of muscles in the model.
2.4
Assistive Torque
The assistive torque was modeled as a rotational actuator, torque of which is proportional to the joint angle as
146
M.S. Shourijeh et al.
Tassist ¼ Kh
ð4Þ
where K (Nm/rad) is the pseudo-stiffness of the assistive device that can be negative, zero (no assist), or positive. In a parametric study, K was changed between [−500, 500] Nm/rad for each joint at a time that seemed to be a reasonably large range for this study.
2.5
Box Interaction with the MSM
A cube (0.4 0.4 0.4 m3) with a mass of 5 kg, with Ixx = Iyy = Izz = 0.768 Kgm2 was assumed for the box. The interaction between the hands and the two sides of the box was modeled as two weld joints. Box bottom was also assumed to move from 0.1 to 0.6 m of the floor.
3 Results Figure 2 depicts the total metabolic power for the four cases in this study: unassisted, and optimally-assisted at hip (K = +50), knee (K = +50), and ankle (K = −100) joints. The total metabolic energy for these cases were 30.3, 25.6, 29.2, and 29.7 kJ, respectively.
Total Metabolic Power (kW)
15.0 13.0 11.0
Unassisted Hip +50 Knee +50 Ankle -100
9.0 7.0 5.0 0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 Time (s)
Fig. 2 Simulated total metabolic power of the box-lifting task for unassisted (black), optimally hip-assisted (green), knee-assisted (blue), and ankle-assisted (red) scenarios
Metabolic Energy Consumption in a Box-Lifting Task …
147
4 Discussion All the three optimal assisted cases led to less metabolic energy required to perform the task. Out of the three joints, hip joint was found to be most influential on the design criterion (i.e. over 15.5 % less metabolic energy). It must be noted that the hip flexion is negative in the model; therefore, a positive pseudo-stiffness for the hip joint will lead to a positive hip extensor torque, which is desired for assistance in the lifting phase. The optimal joint to assist can be dependent on the style of lifting (stooped versus squat). It must be pointed out that in the simulations of this study, the model used the hip more than the knee for the lifting task. The absolute angle variations for the hip and the knee joints were 1.58 and 0.524 rad, respectively, which showed that the lifting was more like a stooped style. It is speculated that if a human subject chooses to perform a squat style lifting, knee will be the optimal joint for assistance. The less is the metabolic energy required from the human, the more comfortable will be the human to perform the task; however, the required assisted torque from the external device might be constrained by the motor size, material properties, etc., which might lead to less assistive torque generation capability. Therefore, computer modeling of the assistive devices must be done with consideration of design and fabrication limitations in order to reach a realistic optimal design.
5 Conclusions and Future Work Effect of adding external assistive torques to the hip, knee, and ankle joints was studied using the AnyBody software. Total metabolic energy over the period of the box-lifting motion was computed and utilized as the design variable. As metabolic energy measure has been broadly used in studies for design optimization of the exoskeletons, musculoskeletal modeling environments such as AnyBody that can provide such information in detail at the joint and muscle levels can significantly contribute. Model kinematics was assumed to be the same between the four cases. However, as the lifting styles might be subjectively different, studying how a range of lifting styles from stooped to squat affects the metabolic energy consumption could be a future research study.
148
M.S. Shourijeh et al.
References 1. Shourijeh, M.S., McPhee, J.: Forward dynamic optimization of human gait simulations: a global parameterization approach. J. Comput. Nonlinear Dyn. 9, 031018 (2014) 2. Voigt, M., Bojsen-Møller, F., Simonsen, E.B., Dyhre-Poulsen, P.: The influence of tendon Youngs modulus, dimensions and instantaneous moment arms on the efficiency of human movement. J. Biomech. 28, 281–291 (1995) 3. Umberger, B.R., Gerritsen, K.G., Martin, P.E.: A model of human muscle energy expenditure. Comput. Methods Biomech. Biomed. Eng. 6, 99–111 (2003) 4. Damsgaard, M., Rasmussen, J., Christensen, S.T., Surma, E., De Zee, M.: Analysis of musculoskeletal systems in the AnyBody modeling system. Simul. Model. Pract. Theory 14, 1100–1111 (2006) 5. Fluit, R., Andersen, M.S., Kolk, S., Verdonschot, N., Koopman, H.F.J.M.: Prediction of ground reaction forces and moments during various activities of daily living. J. Biomech. 47, 2321–2329 (2014) 6. Zajac, F.E.: Muscle and tendon: properties, models, scaling, and application to biomechanics and motor control. Crit. Rev. Biomed. Eng. 17, 359–411 (1988)
Analysis of the Movement Variability in Dance Activities Using Wearable Sensors Miguel Xochicale, Chris Baber and Mourad Oussalah
Abstract Variability is an inherent feature of human movement, but little research has been done in order to measure such a characteristic using inertial sensors attached to person’s body (wearable sensors). Therefore the aim of this preliminary study is to investigate the assessment of human movement variability for dance activities. We asked thirteen participants to repeatedly dance two salsa steps (simple and complex) for 20 s. We then used a technique from nonlinear dynamics (time-delay embedding) to obtain the reconstructed state space for visual assessment of the variability of dancers. Such reconstructed state space is graphically linked with their level of skillfulness of the participants.
1 Introduction Variability is an inherent feature that occurs not only within individual but also between individual systems of movement. Newell and Corcos stated that the movement variability (MV) increases or decreases as a function of practice which is linked with the diminution or increment of skill [1]. In sport biomechanics, for instance, Preatoni et al. stated two important facts about the MV: (i) MV should not be treated as a noise that needs to be removed and (ii) conventional approaches can M. Xochicale is supported by the National Council of Science and Technology—CONACyT Mexico. The support is gratefully acknowledged. M. Xochicale (&) C. Baber School of Electronic Electrical and System Engineering, University of Birmingham, Birmingham, UK e-mail:
[email protected] C. Baber e-mail:
[email protected] M. Oussalah Center for Ubiquitous Computing, University of Oulu, Oulu, Finland e-mail:
[email protected].fi © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_25
149
150
M. Xochicale et al.
only quantify the overall variability. Hence, Preatoni et al. examined non-linear methodologies (entropy measures, dynamical systems theory approaches, and principal component analysis) that are able to deal with and measure variability. This research concluded that the choice of analysis to be used for a particular movement is dependent on the movement in question [2]. Despite the previous efforts of researchers in biomechanics and sport science in measuring the MV, little research has been done with wearable sensors to both quantify the MV and link the MV with the skill assessment of users. For instance, Velloso et al. [3] automatically assessed the quality of weight-lifting activity. Further examples of skill assessment using wearable sensors were investigated on music violin players [4] or medical students doing surgical activities [5]. Little work has been done regarding the use of nonlinear tools using wearable sensors. For instance, Liao et al. used the Empirical Mode Decomposition for activity recognition using accelerometer data [6]. The works of Sama et al. [7] and Frank et al. [8] used the time-delay embedding technique for gait recognition using inertial sensors. We therefore believe that the use of nonlinear tools will provide better measurements and expand the understanding of the variability and skill assessment of activities using wearable sensors. For the current work, we are interested in the question of how the time-delay embedding and PCA techniques can provide insights into the variability and dexterity of dancers. To this end, we consider the performance of a set of steps from Salsa dance and visually compare the variability across dancers.
2 Methods 2.1
Time-Delay Embedding
The aim of the time-delay embedding is to reconstruct a D-dimensional manifold M of an unknown dynamical system sðtÞ from a time series xðtÞ. The time-delay reconstruction, time delayed copies of the available time series xðtÞ, is define as: xðtÞ ¼ ðxðtÞ; xðt sÞ; xðt 2sÞ; . . .; xðt ðm 1ÞsÞÞ where m is the embedding dimension and s is the embedding time-delay. We follow a modified version of the False Nearest Neighbors and the mutual information algorithms to respectively determine the embedded values of m and s [9].
2.2
Framework for the Experiment
The raw data is collected from triaxial accelerometer, gyroscope and magnetometer sensors. For instance, the time series, ax , with a length of N samples is used to obtain the time-delay embedded matrix, Efax g, with m rows and N ðm 1Þs
Analysis of the Movement Variability in Dance Activities …
151
columns. Then, the PCA is applied to obtain, via eigenvalues ðk1 ; . . .; km Þ of eigenvectors ðt1 ; . . .; tm Þ, the principal components ðPC1 ; . . .; PCm Þ of the time-delay embedded phase space which also provide the percentage of variance (POV) per component.
2.3
Participants
Thirteen participants with different years of experience in dancing were invited to dance basic salsa steps: eleven novice dancers (4 female, 7 male), one intermediate (male) dancer (4 years of experience) and one (male) expert dancer (14 years of experience). The design of the experiment was approved by the University of Birmingham ethics approval process. All participants provided informed consent prior to participation.
2.4
Experiment Design
Each participant was shown a series of video clips (recorded by the expert dancer) demonstrating basic salsa steps. Each video clip showed one step repeated several times for 20 s. For the analysis in this work, we reported two Salsa step patterns: step 1 which is mambo and step 2 which is side crossover. Participants watched the video clip and were then asked to copy the steps in time to music. The video was played during the data collection (so that participants did not have to rely on their memory of the steps). Data were collected from the IMUs and recorded. The analysis reported will focus on data from particular axis of the magnetometer (mz for step 1 and my for step 2) taken from the sensor mounted on the left ankle.
2.5
Data Collection
Data from triaxial accelerometer, gyroscope and magnetometer sensors were collected at a sampling rate of 50 Hz using a Razor 9DOF IMU with Bluetooth (BlueSMiRF Silver). The IMUs were attached to custom-made bracelets worn by participants.
152
M. Xochicale et al.
3 Results Figure 1 illustrates the 2-D reconstructed state space for the novice, intermediate and expert dancers. For the time-delay embedding algorithm, we used m = 10 and s = 6 [9]. The reconstructed state spaces visually helped us to distinguish different levels of dexterity. It is immediately noticeable that the shape of the state spaces for each level (novice, intermediate, expert) appears visually similar across step 1. As the participants are meant to be performing the same action, this similarity is to be expected. The state spaces also show a tighter and less varied pattern for the expert than for the other dexterity levels. This suggests that the expert is producing more repeatable and more consistent actions than the other dexterity levels. While this is to be expected, the reconstructed state spaces provide interesting illustrations of this phenomenon. For step 2, which is a more complicated sequence of movements, one
Fig. 1 2-D reconstructed state spaces and percentage of variance using bar plots from the PCA for both participants’ experience and steps
Analysis of the Movement Variability in Dance Activities …
153
can see a marked contrast across dexterity levels. Again, the expert is showing a consistent and repeatable action. The intermediate participant is showing a consistent action but this is different to that of the expert, and the novice is showing a pattern which appears disjointed and noisy. Indeed, for the novice dancer, the state space reconstruction of step 2 seems to have more in common with their state space for step 1 than it does with the other dancers performing step 2. On the other hand, for step 2 the percentage of variance values (bar charts) present a decreasing tendency for the components as the expertise level increased. However, for step 1 we can only say that the first two components have the highest values across the remained components and no evident tendency is shown as the expertise of the participants goes from novice to expert.
4 Conclusion and Future Work Although the time-delay embedding is subject to the embedded parameters (m and s), the technique is useful to visually present the differences among levels of skillfulness. We believe that movement variability (MV) is an ongoing trend towards extending the understanding of human movement with potentially promising applications in the field of human-robot interaction. From this, there are three areas that we are going to investigate: (i) collect data from a wider range of individuals (gender and age) and from additional sensors, (ii) undertake a wider review of nonlinear techniques that can be used for the assessment of MV using wearable sensors, and (iii) explore the use of Hidden Markov Models and Deep Neural Networks for automatic recognition of the variability.
References 1. Newell, K.M., Corcos, D.M. (eds.): Variability and Motor Control, 1st edn. Human Kinetics Publishers Inc., United States of America (1993) 2. Preatoni, E., Hamill, J., Harrison, A.J., Hayes, K., Emmerik, R.E.A.V., Wilson, C., Rodado, R.: Movement variability and skills monitoring in sports. Sports Biomech. 12(2), 62–92 (2013) 3. Velloso, E., Bulling, A., Gellersen, H., Ugulino, W., Fuks, H.: Qualitative activity recognition of weight lifting exercises. Proceeding AH ‘13 Proceedings of the 4th Augmented Human International Conference, pp. 116–123 (2013) 4. Van Der Linden, J., Schoonderwaldt, E., Bird, J., Johnson, R.: MusicJacket—combining motion capture and vibrotactile feedback to teach violin bowing. IEEE Trans. Instrum. Meas. 60, 104–113 (2011) 5. Khan, A., Mellor, S., Berlin, E., Thompson, R., McNaney, R., Olivier, P., Plotz, T.: Beyond activity recognition: skill assessment from accelerometer data. UBICOMP (2015) 6. Liao, M., Guo, Y., Qin, Y., Wang, Y.: The application of EMD in activity recognition based on a single triaxial accelerometer. Bio-Med. Mater. Eng. 6 (2015)
154
M. Xochicale et al.
7. Sama, A., Ruiz, F.J., Nuria, A., Perez-Lopez, C., Catala, A., Cabestany, J.: Gait identification by means of box approximation geometry of reconstructed attractors in latent space. Neurocomputing 121, 77–88 (2013) 8. Frank, J., Mannor, S., Precup, D.: Activity and gait recognition with time-delay embeddings. AAAI Conference on Artificial Intelligence, pp. 1581–1586 (2010) 9. Cao, L.: Practical method for determining the minimum embedding dimension of a scalar time series. Physica D 110, 43–50 (1997)
Part VI
New Developments in Wearable Rehabilitation Robotics
Real Time Computation of Centroidal Momentum for the Use as a Stability Index Applicable to Human Walking with Exoskeleton Je Hyung Jung, Lidwine Van Opheusden, Pierre Barralon and Jan F. Veneman Abstract This abstract presents the real time computation of Centroidal Momentum (CM) in human walking while addressing its applicability as a stability index to human walking with exoskeleton. To this end, a real time full body motion capture suit solution was employed and it was tested both in steady over ground walking and in walking with tripping events. Results show that observation of CM is able to detect abnormal state of human gait in real time thanks to its inherent physical meaning.
1 Introduction Over the last decade, a variety of studies on both upper and lower limbs exoskeletons have been presented and shown that the exoskeletons can not only enhance human power in industrial or military applications but also reduce societal burden for the care of ageing population by enabling independent daily living [1]. Specifically, large attention on lower limb exoskeletons has been paid as they show huge potential to augment and recover walking capability for elderly and physically impaired people. This work has been supported by the partially EU funded project BALANCE, FP7-ICT2011.2.1 grant no. 601003. J.H. Jung (&) P. Barralon J.F. Veneman Health Division, TECNALIA, Mikeletegi Pasealekua 1-3, 20009 Donostia-San Sebastian, Spain e-mail:
[email protected] P. Barralon e-mail:
[email protected] J.F. Veneman e-mail:
[email protected] L. Van Opheusden University of Twente, Drienerlolaan 5, 7522 NB Enschede, Netherlands e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_26
157
158
J.H. Jung et al.
For practical and reliable use of lower limb exoskeletons interacting with a human user, however, still several issues remains and have to be resolved. One of them is how to monitor the stability of a human walking wearing an exoskeleton and detect abnormal states of the walking when the user meets unpredicted disturbances such as a slippery terrain or when stumbling on uneven pavement. Recently, to deal with this topic, Centroidal Momentum (CM), referring to linear and angular momenta at Centre of Mass (CoM), has been introduced and analysed [2] in the context of the EU FP7 project BALANCE (http://balance-fp7.eu/). Preliminary analysis results of CM performed in off-line fashion with data sets recorded in human walking, especially treadmill walking with lateral pelvic perturbations, revealed that monitoring CM is able to distinguish perturbed walking from unperturbed one, thus allowing detection of perturbations during walking [3]. While the off-line analysis showed applicability of the CM as a stability index, it is required to examine the behaviour and performance of CM computation under a real time implementation for the final application where CM is used as a trigger to initiate the balance control (e.g. foot placement adaptation) of an exoskeleton. This abstract deals with this subject and as a first step, presents the real time computation of CM during human natural and perturbed over ground walking (without exoskeleton). The trials were carried out with one subject wearing a full-body motion capture suit from Xsens Company. CM is described shortly in Sect. 2 and a platform employed for its real time computation together with pilot test results appear in Sect. 3. Section 4 contains conclusions as well as future tasks.
2 Centroidal Momentum Centroidal Momentum (CM) can be expressed with a vector whose components are calculated by ~¼ H
n X i $i i i ~ ~ rCM mi ~ vCM þ I x rCM ~ vCM ~
ð1Þ
i¼1
~ L ¼ M~ vCM
ð2Þ
~ is the angular momentum at CoM, ~ where H L the linear momentum at CoM, n the i ! i number of body segments, ~ rCM , v CM and mi are, respectively, the position and $i
~i velocity of the CoM of each segment, and the mass of i-th segment and I and x are the i-th segment’s inertia tensor and angular velocity, respectively, and ~ vCM and M are, respectively, the linear CoM velocity and the total weight of the body [4]. In order to minimize data variance between different subjects, the angular and linear ~ by NH ¼ M j~ momenta are normalized by dividing H L by NL ¼ M j~ vCM jD and ~ vCM j, where D is the CoM height of the body.
Real Time Computation of Centroidal Momentum …
159
3 Real Time Computation of CM 3.1
Demonstration Platform
To demonstrate the real time computation of CM, we used the real time motion data acquisition platform from Xsens. Its name is Xsens MVN (Fig. 1) and consists of a full body suit with 17 Inertial Measurement Units (IMUs), two portable WIFI transmitters and motion visualization software (MVN Studio). Using the IMUs signal and anthropometric data, the kinematic data of 23 body segments are measured and transmitted to MVN Studio through WIFI communication. The data from the suit are refreshed at 120 Hz. In MVN Studio, a human-like 3D avatar imitates, in real time, body movement captured by MVN suit. In addition, MVN Studio provides online streaming of captured motion data, position and orientation of 23 body segments, to Matlab application through UDP (User Datagram Protocol). Hence, the code for real time computation of CM was implemented in Matlab and executed. Real time plotting is also realized to visualize behavior of all elements of CM. Since CM requires the linear and angular velocities, they were numerically calculated in the Matlab codes. Under this implementation, computational time of CM was roughly 3.5 ms, which is considered fast enough to use it in real time, while the visualization for the feedback takes 300 ms in average.
3.2
Demonstration During Natural Overground Walking
The implemented system was investigated in natural over ground walking. Computed CM in real time well corresponds to that obtained in off-line analysis, particularly, showing that angular momentum follows a small periodical pattern around zero during straight walking [4].
3.3
Demonstration During Walking with Tripping Events
In the second test, the subject wearing the suit was asked to fake a tripping situation while walking. Computed CM in real time was visualized as depicted in Fig. 2. Fig. 1 Overview of the platform used for the real time computation of CM with Xsens solution
160
J.H. Jung et al.
From the figure, it can be seen that particular elements of CM, which are angular momentum in Y direction and linear momentum in X direction, show simultaneously large peaks at tripping instances, which means large deviation from the pattern produced in natural walking. From the biomechanical point of view, the variations result from fast rotation of the trunk around the medio-lateral axis together with fast translation in the walking direction, which usually happen when tripping. From this observation, it can be anticipated that real time monitoring of certain elements of CM enables to estimate a type of motion currently occurring, thus detect abnormal states during walking. In addition, it is observed that at turning instances, the angular momentum in vertical direction and linear momentum in medio-lateral direction, show large peaks simultaneously, as can be expected from the biomechanical interpretation.
Fig. 2 Computed normalized Centroidal Momentum in real time during walking with tripping events. X, Y, and Z denote anterio-posterio, medio-lateral, and vertical direction respectively. ① indicates tripping event instance while ② indicates turning instance
(a) Calculated normalized angular momentum at CoM
(b) Calculated normalized linear momentum at CoM
Real Time Computation of Centroidal Momentum …
161
4 Conclusion In this paper, we presented the real time computation of Centroidal Momentum and investigated its use as a stability index during human walking. The results show that CM is computed fast enough for real time use in commercially available body motion capture suit while being able to monitor an actual state of balance during walking thanks to its inherent physical meaning. Next step will be the integration of the CM approach into an exoskeleton and investigate adoptability of the CM as the stability index for human walking with the exoskeleton, which might be affected by interaction between human and exoskeleton. IMU, as is well known, is sensitive to electromagnetic (EM) fields so the leg kinematic information of the exoskeleton directly measured with angle sensors will be incorporated together with IMUs signals in order to improve the robustness of the approach. Acknowledgments Authors thank Henk Luinge from Xsens for providing their platform for the study.
References 1. Dollar, A.M., Herr, H.: Lower-extremity exoskeleteon and active orthoses: challenges and state-of-the-art. IEEE Trans. Robot. 24(1), 144–158 (2008) 2. Jung, J.H., Gutiérrez, I., Veneman, J.: The use of centroidal momentum analysis for defining a stability index for walking with an exoskeleton. In: TAR 2015: Technically Assisted Rehabilitation Conference, Berlin, Germany (2015) 3. Jung, J.H., Veneman , J.: Centroidal momentum analysis for defining a stability index for human-exoskeleton interactive walking: perturbation detection in human gait. Assistance and Service Robotics in a Human Environment Workshop in conjunction with 2015 IROS, September 28 2015–Oct. 2, 2015, Hamburg, Germany 4. Herr and, H., Popovic, M.: Angular momentum in human walking. J. Exp. Biol. 211(4), 467– 481 (2008)
A Versatile Neuromuscular Exoskeleton Controller for Gait Assistance: A Preliminary Study on Spinal Cord Injury Patients Amy R. Wu, Florin Dzeladini, Tycho J. H. Brug, Federica Tamburella, Nevio L. Tagliamonte, Edwin van Asseldonk, Herman van der Kooij and Auke J. Ijspeert
Abstract We investigated the capabilities of a reflex-based neuromuscular controller with a knee and hip gait trainer worn by a subject with a complete spinal cord injury. With controller assistance, this subject was able to reach a walking speed of 1.0 m/s. Measured joint torques agreed reasonably well with those of healthy subjects. The controller was also robust, recovering from manual swing foot perturbations. These preliminary results are promising for future implementation of neuromuscular controllers on wearable prototypes for real-world walking conditions.
1 Introduction Robust and reliable controllers of gait assistive devices for patients with neurological disorders must balance healthy-like walking function with promotion of motor recovery. We take a biologically-inspired approach with a controller based on a neuromuscular model developed by Geyer [2]. This neuromuscular controller (NMC) generates walking by activating simulated muscle reflex loops based on gait state, combining muscle-tendon dynamics to produce lower-limb joint torques. While the full controller can reproduce torques at the ankle, knee, and hip, its modular structure permits use for any combination of these joints. No pre-defined movement pattern is needed, and few sensors are required. Ground contact detection is needed to switch between stance and swing reflexes, and joint angles are used to calculate simulated A.R. Wu (B) · F. Dzeladini · A.J. Ijspeert BioRobotics Laboratory of the École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland e-mail:
[email protected] F. Tamburella · N.L. Tagliamonte Laboratory of Robotic Neurorehabilitation in the Neurological and Spinal Cord Injury Rehabilitation Department A of Fondazione Santa Lucia, Rome, Italy T.J.H. Brug · E. van Asseldonk · H. van der Kooij Department of Biomechanical Engineering of the University of Twente, Holland, The Netherlands H. van der Kooij The Netherlands and the Department of Biomechanical Engineering of the Delft University of Technology, Delft, The Netherlands © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_27
163
164
A.R. Wu et al.
muscle state. This flexibility enables active recruitment of the patient’s own neuromuscular system, an important component of motor function recovery [3]. In simulation, this controller is also robust against perturbations [1] and can adapt to non-steady state behaviors [5]. We performed a preliminary evaluation of this controller on a lower-limb exoskeleton worn by a subject with a complete spinal cord injury (SCI). We hypothesize that the controller dynamics will enable the subject to walk at different walking speeds and recover from small perturbations.
2 Materials and Methods We implemented the neuromuscular controller on LOPES, a knee and hip gait trainer from University of Twente [4]. Only the knee and hip control modules were used in the sagittal plane; the device had no ankle actuation. With knee and hip angles and stance detection as inputs, the controller commanded knee and hip torques to the exoskeleton. To illustrate the controller’s effect on gait, we present results from one SCI subject. This subject (male, 24 years of age, mass 63 kg, height 1.8 m) has a complete lesion (AIS A) at T7 (traumatic, 60 months from time of lesion). During the experiment, the subject was body weight supported (38 % of body weight BW) and used the handrails (33 % of BW). We compared the subject’s NMC-assisted gait with that of one healthy subject (“HC” for Healthy with NMC), and two healthy subjects during shod walking, termed “N” for Normal. As a preliminary study on controller robustness, we also applied manual perturbations, push and pull, to the right swing foot. The manual push in the forward direction produced a longer stride, and the manual pull in the posterior direction impeded the swing foot and created a shorter stride. We recorded knee and hip torques provided by the controller at three different speeds (0.7, 0.9, and 1.0 m/s). The mean of these torques was compared with Normal data, the average joint torques calculated from inverse dynamics of two healthy subjects walking at 1.0 m/s. The perturbation trial was performed at 0.6 m/s. All subjects provided written informed consent prior to the study, according to Institutional Review Board procedures.
3 Results The SCI subject in this study was able to walk in LOPES with NMC assistance from 0.7 m/s up to 1.0 m/s. This subject’s gait was mostly symmetric, as left and right step duration and step length were similar. Therefore we only include results from the right leg. The magnitudes of measured joint angles and torques of this subject increased with greater walking speed (Fig. 1), as observed in previous studies [6].
A Versatile Neuromuscular Exoskeleton Controller for Gait Assistance …
KNEE
Angle (deg)
30
HIP Ext
Ext
Flx
Flx
0 -30
0.7 m/s 0.9 m/s 1.0 m/s 1.0 m/s (HC) 1.0 m/s (N)
-60 Toe-off
-90 50
Torque (N-m)
Fig. 1 Knee and hip angles and controller torques at three speeds (solid lines), compared with Healthy with NMC (HC, dotted lines) and Normal (N, dashed lines). Joint angle and joint torque magnitudes increased with speed. Hip and knee torques deviated from Normal during early and late stance. Gait cycle is defined as two consecutive right heel-strikes
165
Ext
Ext 0 Flx
Flx -50
-100
0
50
100 0
Gait Cycle (% of stride)
50
100
Gait Cycle (% of stride)
We found some differences between NMC-provided torques and Normal torques (Fig. 1). For example, near 50 % of gait, the controller provided more hip flexion torque but lacked knee flexion torque. Despite these torque differences, the joint angle trajectories did not seem greatly affected. The controller was also able to maintain gait despite perturbations to the swing leg. The push perturbation, which induced a longer stride (1.21 m vs. 1.07 m preperturbation), created larger hip extension and knee flexion torques during swing (Fig. 2). To compensate, the post-perturbation stride had less knee extension during stance and less knee flexion and more hip flexion during swing, with the overall effect of producing a shorter stride (0.98 m). The pull perturbation had a similar but
Hip Torque (N-m)
50
PUSH PERTURBATION
PULL PERTURBATION
Toe-off Ext
Ext
Flx
Flx
0
-50
Pre-pert. Perturbed Post-pert.
-100 40
Knee Torque (N-m)
Fig. 2 Hip and knee torques during two types of swing leg perturbations. The push perturbation created higher hip extension and knee flexion torques, while the pull produced the opposite. The controller compensated for both in the next stride. Gait cycle is defined as two consecutive right heel-strikes
30 20 10
Ext
Ext
0 Flx
Flx
-10 0
50
Gait Cycle (% of stride)
100 0
50
Gait Cycle (% of stride)
100
166
A.R. Wu et al.
opposite effect. The short stride length from the impeded foot (0.96 m) was followed by a longer stride (1.27 m). The pull induced more hip flexion and less knee flexion torques. The compensatory behavior was a larger knee extension torque during stance and greater hip extension and knee flexion torques near the end of swing.
4 Discussion We demonstrated the capabilities of the neuromuscular controller on a lower-limb exoskeleton with one SCI subject. Our preliminary results showed that the controller enabled the subject to walk at various speeds, including near healthy (1.0 m/s), and recover from swing foot perturbations. This was accomplished solely with joint angles and ground contact as controller inputs. While the SCI subject’s joint angles were similar to Normal, the controller torques deviated near late stance. Because NMC results from the healthy subject (see Fig. 1, dotted line) also exhibited similar behavior, we believe these deviations were due to the lack of ankle actuation and unmodeled virtual muscles at the ankle joint. We could not compare controller torques against the device’s transparent mode (i.e. compensation of passive dynamics) because the subject could not volitionally initiate and sustain gait. We also had tested the controller with six SCI subjects but felt the results were not directly comparable. Each subject had unique neurological symptoms, and controller settings were tuned for their specific walking ability. Several factors could affect controller behavior, such as body weight support, handrail usage, and the absence of ankle actuation. The controller was also not scaled or optimized to the subjects’ anthropometry and walking speed. Nonetheless the controller exemplified enough versatility to generate gait patterns tuned to the subjects’ dynamics. Further work is needed to provide online subject-specific assistance (i.e. tuning individual virtual muscles or particular joints). Such assistance could start with offline tailoring based on subject’s anthropometry and clinical assessments, but then be further tuned by evaluation of subject effort, gait symmetry, and similarity to healthy gait, with the overall goal of promoting positive subject feedback.
5 Conclusion We performed a preliminary evaluation of a biologically-inspired, neuromuscular controller on a lower-limb exoskeleton worn by a subject with paraplegia. This subject was able to walk at multiple speeds and recover from swing foot perturbations. These auspicious results have important implications towards the design of controllers for wearable devices, which must be amenable to various environmental conditions and promote intuitive and unobtrusive human-machine interaction.
A Versatile Neuromuscular Exoskeleton Controller for Gait Assistance …
167
Acknowledgments This work is supported by EU research program FP7-ICT-2013-10 (SYMBITRON, grant #611626, coordinated by University of Twente). The authors thank Gijs van Oort of the University of Twente for help in conducting this study.
References 1. Dzeladini, F., van den Kieboom, J., Ijspeert, A.J.: The contribution of a central pattern generator in a reflex-based neuromuscular model. Front Hum. Neurosci. 8(371), (2014) 2. Geyer, H., Herr, H.: A muscle-reflex model that encodes principles of legged mechanics produces human walking dynamics and muscle activities. IEEE Trans. Neural Syst. Rehabil. Eng. 18(3), 263–273 (2010) 3. Israel, J.F., Campbell, D.D., Kahn, J.H., Hornby, T.G.: Metabolic costs and muscle activity patterns during robotic-and therapist-assisted treadmill walking in individuals with incomplete spinal cord injury. Phys. Ther. 86(11), 1466–1478 (2006) 4. Meuleman, J., van Asseldonk, E., van Oort, G., Rietman, H., van der Kooij, H.: LOPES II-design and evaluation of an admittance controlled gait training robot with shadow-leg approach. IEEE Trans. Neural Syst. Rehabil. Eng. 24(3), 352–363 (2016) 5. Song, S., Geyer, H.: A neural circuitry that emphasizes spinal feedback generates diverse behaviours of human locomotion. J Physiol. 593(16), 3493–3511 (2015) 6. Zelik, K.E., Kuo, A.D.: Human walking isn’t all hard work: evidence of soft tissue contributions to energy dissipation and return. J. Exp. Biol. 213(Pt 24), 4257–4264 (2010)
Introducing a Modular, Personalized Exoskeleton for Ankle and Knee Support of Individuals with a Spinal Cord Injury Cory Meijneke, Shiqian Wang, Victor Sluiter and Herman van der Kooij
Abstract In the Symbitron Project, one of the main objectives is to develop a safe, bio-inspired, and personalized wearable exoskeleton that enables individuals with a spinal cord injury (SCI) to walk without additional assistance, by complementing their remaining motor function. The first target group of five subjects, have enough hip control to keep themselves upright, but need support around the ankle and/or knee joint. This paper gives an overview of the design features of the newly developed exoskeleton and shares some details about the design process.
1 Introduction Many exoskeletons have been developed for Spinal Cord Injured (SCI) subjects [1–4]. These exoskeletons are mostly designed to support for full lower limb paralysis and require the use of crutches. The exoskeletons to be developed in the Symbitron project aim to support SCI subjects with various levels of impairments. To achieve this, two groups of five subjects have been selected: (1) subjects who have impaired motor function at the ankle and/or knee, but sufficient function at the hip, and (2) subjects with fully impaired lower limbs. The first Symbitron Wearable Exoskeleton (WE1) is
The work presented here was supported by the EU within the SYMBITRON project (FP7-ICT-2013-10 contract #611626). C. Meijneke (&) S. Wang H. van der Kooij Department of Biomechanical Engineering, Faculty of Mechanical, Maritime and Materials Engineering, Delft University of Technology, Mekelweg 2, 2628 CD Delft, The Netherlands e-mail:
[email protected] V. Sluiter Biomechanical Engineering Laboratory, Institute for Biomedical Technology and Technical Medicine (MIRA), University of Twente, Drienerlolaan 5, 7500 AE Enschede, The Netherlands © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_28
169
170
C. Meijneke et al.
developed for the first group of subjects (group I), thus provides knee and ankle support. Because of the modular design of the joints, the ankle and knee modules can also be (re)used as a basis for the second exoskeleton (WE2) suited for group II subjects. Furthermore, the remaining motor function of the group I subjects varies as well, hence for some, the ankle module can give sufficient support, while for others, both the ankle and knee are needed. Commercially available exoskeletons are developed for a large population and are fitted on different users using adjustment mechanisms and custom padding. The Symbitron exoskeletons are personalized i.e. the anatomical features of the subject are accommodated for in the design phase. Special attention is given to the ease of doing on and off the device. Based on experiences with the Mindwalker [5] and Achilles [6] exoskeleton, it is key to reducing setup time and improving comfort during experiments. The following sections discuss the mechanical and electrical design features and list some preliminary specifications of the exoskeleton.
Fig. 1 Rendered 3D model of side and front view of WE1 where (1) denotes the personalized tube and strap to the wearer, (2) knee actuation unit, (3) quick connection mechanism between the ankle and knee modules, (4) ankle actuation unit, (5) personalized tube to the wearer with integrated linkage mechanism to the ankle, (6) personal footplates compatible with sensorized insoles [7]
Introducing a Modular, Personalized Exoskeleton for Ankle …
171
2 Mechanical Design An overview of the design features of WE1 is shown in Fig. 1. The design consists of a knee and ankle module, each with one actuated degree of freedom, respectively flexion/extension and plantar/dorsal flexion. Additionally, the ankle inversion/eversion is free, but has option to be locked. The two modules are connected by hooking them in to each other and locked using a toggling latch. The size and shape of the exoskeleton segments and shells are adapted to body size, shape and anatomical axis (e.g. distance between ankle and knee joint center) of the subjects. Instead of introducing adjustment mechanisms, thus adding mass and complexity, some of the exoskeleton components are designed specifically to fit each subject in the test-groups. An overview of the main features of the actuation unit are shown in Fig. 2. Most components inside the actuator are generic for each actuated joint. The included series elastic element and high resolution sensors enable accurate torque control. The gear ratio and mechanical attachments can be adapted to fit each joint. Fig. 2 Rendered 3D model of Symbitron Actuation Unit where (1) power and EtherCAT communication connectors, (2) Powerful control electronics with integrated IMU, (3) flexible connection to base components, (4) flat high performance electric motor, (5) compact harmonic drive, (6) output flange with integrated end-stops
172 Table 1 Preliminary specifications
C. Meijneke et al. Description Peak output torque Peak output speed Average motor Power Actuation unit mass WE1 mass per leg Torque resolution
Value Knee 70 120 750 1.5 5 0.012
Unit Ankle 100 60
Nm rpm W kg kg Nm
3 Electronic Design The modular design of each actuated joint puts special demands on the electronics. The actuation unit has custom-made EtherCAT slave electronics, which combines a motor controller, inertial measurement unit and various interfaces for additional sensors. The slave has three external connections, i.e., power supply and in/out EtherCAT communication ports. In this way, the actuation units form a distributed sensor and actuator network that allows flexible configurations and enables real-time control of the exoskeleton from a main computer at the trunk of the pilot.
4 Specifications At time of writing of this article, manufacturing of the device is not finished and basic functional test have been one with a test joint. The specification in Table 1 are preliminary.
5 Conclusion This paper showed and overview of a newly developed exoskeleton for knee and ankle support of SCI subjects. In an effort to get the best fit between the exoskeleton and subject in a simple and lightweight manner, multiple component sets tailored to fit all subjects in the test-group are made. The modular design of the exoskeleton enables a flexible configuration and easy doing on and off the device. Thanks to the modular actuation unit, future extension of the exoskeleton to the hip and trunk is also made relatively simple. Acknowledgments Great effort has been put into the design and manufacturing of the exoskeleton by Wouter Gregoor and Tom Phillips from the Electrical and Mechanical Support Division at Delft University of Technology. Thanks to effort and patience of the subjects and staff of Fondazione Santa Lucia, we obtained detail data and scans for out design. The exoskeleton footplates where provided by Össur.
Introducing a Modular, Personalized Exoskeleton for Ankle …
173
References 1. 2. 3. 4. 5.
Indego website. http://www.indego.com/indego/en/home (2016). Accessed 15 Apr 2016 Rex Bionics website. http://www.rexbionics.com/ (2016). Accessed 15 Apr 2016 EksoBionics website. http://www.eksobionics.com/ (2016). Accessed 15 Apr 2016 ReWalk Robotics website. http://rewalk.com/ (2016). Accessed 15 Apr 2016 Wang, S., Wang, L., Meijneke, C., van Asseldonk, E., Hoellinger, T., Cheron, G., Ivanenko, Y., La Scaleia, V., Sylos-Labini, F., Molinari, M., Tamburella, F., Pisotta, I., Thorsteinsson, F., Ilzkovitz, M., Gancent, J., Nevatia, Y., Hauffe, R., Zanow, F., van der Kooij, H.: Design and control of the MINDWALKER exoskeleton. IEEE Trans. Neural Syst. Rehabil. Eng. 4320(c) (2014) 6. van Dijk, W., Meijneke, C., van der Kooij, H.: Evaluation of the Achilles Ankle exoskeleton. IEEE Trans. Neural Syst. Rehabil. Eng. 4320(c), 1–1 (2016) 7. Crea, S., Donati, M., De Rossi, S.M.M., Oddo, C.M., Vitiello, N.: A wireless flexible sensorized insole for gait analysis. Sensors (Basel) 14(1), 1073–1093 (2014)
Towards Exoskeletons with Balance Capacities Herman van der Kooij, Edwin H.F. van Asseldonk and Mark Vlutters
Abstract Current exoskeletons replay pre-programmed trajectories at the actuated joints. Towards the employment of exoskeletons with more flexible and adaptive behavior, we investigate human balance control during gait. We study human balance control by applying brief force pulses at the pelvis in different directions, with different amplitude, and applied at different phases of the gait phase. The observed changes were dependent on the phase at which the perturbation was applied and the walking velocity. From the results we concluded that foot placement was the dominant strategy in the frontal plane, center of pressure (CoP) modulation in the double support phase was utilized in the sagittal plane, and the duration of the swing and double support phase changed. Without the ability to control the CoP through an ankle torque, humans also used a foot placement strategy in the sagittal plane. The center of pressure with respect to the center of mass at the end of the double support phase was linearly related to velocity of the center of mass at the end of the preceding swing phase, which is in agreement with extrapolated center of mass or capture point based stepping strategies previously applied in simple models.
This work was supported by the BALANCE (Balance Augmentation in Locomotion, through Anticipative, Natural and Cooperative control of Exoskeletons) project, partially funded under grant 601003 of the Seventh Framework Program (FP7) of the European Commission (Information and Communication Technologies, ICT-2011.2.1) and the SYMBITRON project (FP7-ICT-2013-10 contract #611626). H. van der Kooij (&) Biomechanical Engineering Group, University of Twente, Enschede, Netherlands e-mail:
[email protected] E.H.F. van Asseldonk M. Vlutters University of Twente, Enschede, Netherlands e-mail:
[email protected] M. Vlutters e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_29
175
176
H. van der Kooij et al.
1 Introduction Most current exoskeletons are unable to stay upright without assistance and guidance of its user. Paraplegic users, for example, require crutches to prevent falling. In the EU Balance and Symbitron projects we aim to support balance control by wearable exoskeletons. Since human balance control during walking is still poorly understood we performed a series of experiments to analyze balance recovery during gait in attempt to predict human balance control strategies.
2 Materials and Methods 2.1
Experimental Setup and Protocol
Nine healthy young adults walked on a custom dual-belt instrumented treadmill (MotekForce Link, Culemborg, Netherlands), while keeping arms crossed over the abdomen. Walking speeds were 2.25 and 4.50 km/h, scaled to the subject’s leg length. For each speed, subjects first walked a 2 min unperturbed baseline trial. In subsequent trials, perturbations were applied to the pelvis using one of two motors (Moog, Nieuw-Vennep, Netherlands), connected with a lever arm to a pelvic brace (Distrac, Hoegaarden, Belgium), worn by the subject. The motors were located at the side and the rear of the treadmill. Our experimental setup is shown in Fig. 1. Fig. 1 Experimental setup
Towards Exoskeletons with Balance Capacities
177
Fig. 2 Pin-foot boots worn by subjects in one of the experiments to study the effect of foot placement in the sagittal plane, to prevent modulation of the cop by changes in ankle torque
Perturbations were applied at randomly selected instances just after toe-off right (TOR), and consisted of 150 ms block pulses of a force magnitude equal to 4, 8, 12 and 16 % of the subject’s body weight. Perturbation directions were forward (positive), backward (negative), outward (positive, away from stance leg) and inward (negative, toward stance leg). Each perturbation type was repeated 8 times, leading to 256 perturbations per subject. Kinematic data of various landmarks on the lower extremities, pelvis, trunk and head [1] were collected at 100 Hz using a motion capture system (Phoenix Technologies Inc, Vancouver, Canada). Ground reaction forces, subject-motor interaction forces, and EMG data were collected at 1000 Hz. More details about the experimental setup can be found in [2]. To exclude the ankle strategy (i.e. modulation of the center of pressure (CoP) by applying an ankle moment), we did a variation of this experiment in which subjects were wearing a pin foot (Fig. 2) while only the backward and forward perturbations were applied at the slow walking speed.
3 Results At heel strike (HS) after the perturbation, recovery from mediolateral (ML) perturbations involved ML foot placement adjustments proportional to the ML CoM velocity. In contrast, for anterior posterior perturbations (AP) no significant AP foot placement adjustment occurred at HS (Fig. 3 left). However, in both directions the
178
H. van der Kooij et al.
Fig. 3 Typical example of one subject. The foot placement of the right leg (i.e. the AP distance of the right foot with respect to the CoM) at heel strike right (HSR) plotted against the AP CoM velocity at HSR
CoM velocity at HS related linearly to the CoP location at the subsequent toe-off (not shown). This relation was affected by the walking speed and was, for the slow speed (not shown), in line with a CoM velocity based control strategy previously applied by others in a linear inverted pendulum model [3]. Finally, changes in gait phase durations (not shown) suggest that the timing of actions could play an important role during the perturbation recovery. Results of this experiment not shown here can be found in [2]. When subject were wearing the pin-foot boots, recovery from AP perturbations also involved AP foot placement, which were proportional to the AP CoM velocity at HS (Fig. 3 right).
4 Discussion These experimental results have implications for the control and design of wearable exoskeletons that aim to support human balance control. In AP directions XCoM at HS can be used to predict the desired CoP location at the end of double support phase. For ML directions xCoM can be used to predict the desired foot placement relative to the CoM. For human-like ML stabilization wearable exoskeletons need actuated hip ab/adduction, and for AP stabilization they require torque controlled ankle actuation.
5 Conclusions The CoP is a main variable humans use to maintain balance during gait. In the ML direction CoP is controlled by foot placement and in the AP direction by ankle torque and foot placement. When ankle torque cannot be used to change the CoP in
Towards Exoskeletons with Balance Capacities
179
the AP direction, humans switch to the foot placement adjustment strategy in the sagittal plane. The CoP at the end of the double support phase is linearly related to the CoM velocity at the end of the preceding swing phase. Timing of foot placement is an important variable for balance control, which is more difficult to predict.
References 1. Dumas, R., Chèze, L., Verriest, J.P., et al.: Adjustments to McConville et al. and Young et al. body segment inertial parameters. J. Biomech. 40, 543–553 (2007) 2. Vlutters, M., van Asseldonk, E.H.F., van der Kooij, H.: Center of mass velocity based predictions in balance recovery following pelvis perturbations during human walking. J. Exp. Biol. 20 (2016). doi:10.1242/jeb.129338 3. Hof, L., van Bockel, R.M., Schoppen, T., Postema, K.: Control of lateral balance in walking. Experimental findings in normal subjects and above-knee amputees. Gait Posture 25(2), 250– 258 (2007)
EMG-Based Detection of User’s Intentions for Human-Machine Shared Control of an Assistive Upper-Limb Exoskeleton A. Accogli, L. Grazi, S. Crea, A. Panarese, J. Carpaneto, N. Vitiello and S. Micera
Abstract To assist people with disabilities, exoskeletons must be provided with human-machine interfaces (HMI) capable to identify the user’s intentions and enable cooperative interaction. Electromyographic (EMG) signals could be suitable for this purpose, but their usability and effectiveness for shared control schemes in assistive devices is currently unclear. Here we developed advanced machine learning (ML) algorithms for detecting the user’s motion intention and decoding the intended movement direction, and discuss their applicability to the control of an upper-limb exoskeleton used as an assistive device for people with severe arm disabilities.
1 Introduction Exoskeletons can operate in cooperation with humans and assist them to perform multiple activities of daily living. One of the goals of the control systems for such devices is to enable the robot to identify the user’s intentions and capabilities by means of a suitable HMI. Shared control schemes can thus be implemented, i.e. the robot can take over or leave the user to control specific aspects of the motion, and adjust the level of assistance/cooperation when needed [1]. In the recent years, EMG signals have been increasingly used for the control of robots [2], still their
This work was supported by the AIDE Project GA 645322, funded by the EU H2020 Framework Programme for Research and Innovation, and by the RONDA Project, funded by Regione Toscana PAR FAS 2007-2013. A. Accogli L. Grazi S. Crea A. Panarese (&) J. Carpaneto N. Vitiello S. Micera Biorobotics Institute, Scuola Superiore Sant’Anna, Pisa, Italy e-mail:
[email protected] S. Micera Center for Neuroprosthetics and Institute of Bioengineering, Ecole Polytechnique Federale de Lausanne (EPFL), Lausanne, Switzerland © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_30
181
182
A. Accogli et al.
usability to implement shared control schemes for assistive exoskeletons is currently under study. Here we aimed to develop fast and accurate EMG-based algorithms for detecting user’s motion intention and decoding movement direction, and to envisage shared control schemes for assisting people with severe arm disabilities wearing an upper-limb exoskeleton.
2 Materials and Methods 2.1
Exoskeleton
In this work we used NESM [3], an upper-limb robotic exoskeleton with four active degrees of freedom allowing for shoulder adduction/abduction, flexion/extension, intra/extra rotation, and for elbow flexion/extension. The controller of the robot implements both a joint position and a joint torque control modes. The first guides each motor to follow a reference trajectory in the joint space, whereas the second applies torque to the actuation units to either be transparent or assist the user’s movement.
2.2
Setup of the Study
A graphical user interface (GUI) was implemented in Labview (National Instruments, Austin, TX, US) with three LEDs (W, C, and E) placed on different positions on a computer screen (15 cm apart, at left, center, and right, respectively; Fig. 1). The subject (healthy, male, 35 years old) was instructed to (i) keep a resting position as long as all the LEDs were off, (ii) move the arm towards the LED which switched on, (iii) stop when the target was reached and keep the position, and (iv) move back to the resting position as soon as the LED switched off. Each of n = 9 sessions consisted of 27 reaching trials. The targets were presented in random order. The LED switched on every 10 s, and remained on for the following 5 s. During each trial, the exoskeleton was in transparent mode. Fig. 1 Snapshots of the experimental setup
EMG-Based Detection of User’s Intentions for Human-Machine …
183
EMG signals from seven muscles of the right shoulder (Trapezius, Anterior and Posterior Deltoid), arm (Biceps and Triceps Brachii) and forearm (Flexor and Extensor Carpi Ulnaris) were recorded at 1.5 kHz with a TeleMyo 2400R (Noraxon Inc., AZ, US), and low-pass filtered at 500 Hz.
2.3
Motion Intention Detection
A Gaussian Mixture Model (GMM) method was used to model the rest and movement phases of the muscular activity during the task. The algorithm first estimated the model on an initial (training) time window; then, the parameters were sequentially evaluated for each new time sample, and an optimal threshold between rest and movement was calculated, minimizing the classification error [4]. The detector evaluated the outputs of multiple GMM algorithms working in parallel on the features MMAV1, SSI, VAR e LOG [5]. We compared the performances of three types of detector based on: (i) single EMG signals, (ii) multiple EMG signals (different detectors working in parallel on single EMG signals, with a majority voting procedure for the final decision), (iii) a compound EMG signal, sEMG ðtn Þ ¼
X#EMGs i¼1
jEMGi ðtn Þ EMGi ðtn DtÞj
ð1Þ
which takes into account the instantaneous variations of all the EMG signals at the same time. Three parameters were used to evaluate the performances: sensitivity ¼
TP TP þ FN
ð2Þ
precision ¼
TP TP þ FP
ð3Þ
latency ¼ htd t0 itrials
ð4Þ
where TP, FN and FP are the number of true positive, false negative and false positive respectively; td is the detected onset time and t0 is the reference onset time, estimated as the time instant corresponding to the 10 % of the peak values assumed by kinematic variables when the LED was on.
2.4
Classification of Movement Direction
Classification problems were solved using Support Vector Machines (SVM), written in MATLAB (Mathworks, Natick, MA, US). Muscle activity recorded in a time window of 300 ms, starting from the movement onset, was used to extract
184
A. Accogli et al.
features to be used as inputs to the classifier, whereas the movement directions (W, C, and E) were used as outputs. A second version of the classifier received the position signals of the four active joints (in the same window used for EMGs, [t0 ; t0 þ 300 ms) as additional inputs. 70 % of the available data was used for training, whereas the remaining 30 % for testing. The mean accuracy across 10 realizations of random sub-sampling validation was evaluated.
3 Results The detector based on the compound EMG signal was capable to detect movement onset much earlier than the other two, even *40 ms before the reference onset time, t0 (Fig. 2a). The three detector types showed to be equally sensitive and precise. The median performances across all muscles were considered for detectors using single EMG signals. Correct classification of movement direction was 61 %, on average, when features of the EMG signals alone were used as inputs. Performance increased to 89 % (Fig. 2b) by adding kinematic information. Fig. 2 A Performances of movement onset detectors. B Confusion matrices of classification performance for movement direction with EMG features alone (left) and with additional kinematic inputs (right)
EMG-Based Detection of User’s Intentions for Human-Machine …
185
4 Discussion GMM is a fast and accurate method for detection of EMG onset [4]. Still, its computational requirements must be minimized for real-time applications. By using the compound EMG signal instead of multiple EMGs processed in parallel, we have shown that it is possible to maintain (or even improve) performance while reducing computational load. SVM classification of movement direction has been studied previously in people with arm disabilities but proved to be inaccurate [6]. Here, we demonstrated that decoding performance can be drastically improved by adding kinematic information from the exoskeleton as input to the classifier. This is expected to strengthen performance even in users with arm disabilities. Future work will be devoted to study the real-time application of the presented algorithms in people with severe arm disabilities. The NESM allows for a bump-less switch between control modes and it is possible to envisage a shared control scheme where the user commands movement start and intended direction (with the device in transparent mode) whereas the exoskeleton executes (switching to position control) or assists the movement thereafter, depending on the user’s residual capabilities.
5 Conclusions Fast and accurate detection of motion intention and classification of movement direction are pivotal to implement useful control schemes for upper-limb assistive devices. HMI based on EMG and advanced ML algorithms represent a suitable solution for this purpose.
References 1. Yu, H., et al.: An adaptive shared control system for an intelligent mobility aid for the elderly. Auton. Robots 15, 53–66 (2003) 2. Singh, R.M., Chatterji, S.: Trends and challenges in EMG based control scheme of exoskeleton robots-a review. Int. J. Sci. Eng. Res. 3, 1–9 (2012) 3. Crea, S., et al.: A novel shoulder-elbow exoskeleton with series elastic actuators, unpublished 4. Liu, J., et al.: Robust muscle activity onset detection using an unsupervised electromyogram learning framework. PloS ONE 10(6) (2015) 5. Phinyomark, A., et al.: A novel feature extraction for robust EMG pattern recognition. J. Comput. 1(1), 71–80 (2009) 6. Cesqui, B., et al.: EMG-based pattern recognition approach in post stroke robot-aided rehabilitation: a feasibility study. J. NeuroEng. Rehabil. 10, 75 (2013)
Part VII
Legal Framework, Standardization and Ethical Issues in WRs
Safety Standardization of Wearable Robots—The Need for Testing Methods Jan F. Veneman
Abstract Wearable Robots, i.e. exoskeletons, are currently entering the market. The specific procedure to allow devices on the market, concerning their product safety depends on the application domain they are sold for (medical—industrial— personal care). Safety Standards, such as produced by ISO and IEC are important tools to demonstrate safety for specific devices, but at this moment there are only very few specific standards available and no specific testing methods in place for Wearable Robots. Standardized testing methods that do not require human subject testing can make the road to the market easier and better affordable. Such test methods need to be based on validated experimental data, and may require the development and definition of specific targeted test benches or equipment.
1 Introduction At this moment, Wearable Robots (WRs), such as exoskeletons for rehabilitation and worker support, start finding their way into the market, while the regulatory situation around their use is not very well established. The current market introduction can be seen as ‘experimental’, which may have unclear and unpredictable effects on the future of companies marketing this type of robots. The currently available WR products can in general be considered as a first generation of devices that only provide basic functionality, operate at low power and apply straightforward control. The WR devices that currently emerge in academic and other research environments, the ‘next generation’, goes much further in
This work has been partially supported by the EU funded project BALANCE, FP7-ICT-2011.2.1 grant no. 601003. J.F. Veneman (&) Health Division, TECNALIA Research and Innovation, Mikeletegi Pasealekua, 1, 20009 Donostia-San Sebastian, Spain e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_31
189
190
J.F. Veneman
Fig. 1 The regulatory environment of Wearable Robots is, as any, characterized by regulation that prescribes how to realize and assure acceptable (safe) products while occurrences of damage due to products already in use, where no prior regulation was in place, is set through litigation. This resembles communicating vessels: if no specific regulation is in place, product safety will be defined through litigation around liability issues [1]
versatility, power capacity and functionality, and hence in potential safety threats to the user and the environment. The realization of product safety for a technology like WR is based on one side on regulation, which applies before a device is put on the market, and which allows or forbids specific products; and on the other side on litigation, which takes place after a specific case of damage has occurred and the liability has to be settled (see also Fig. 1).
2 Regulation of Wearable Robots The ‘ex-ante’ regulation is a framework of rules that prescribes certain aspects of the technology, for example product safety. In Europe the general guidelines are set in the Council Directive of 14 June 1993 concerning medical devices (MDD— Medical Device Directive) [2], and Directive 2006/42/EC of the European Parliament and the Council of 17 May 2006 on machinery, and amending Directive 95/16/EC (MD—Machinery Directive) [3]. The regulations directed by the MDD and the MD indicate the requirements that producers need to respect in order to obtain an EC marking allowing the commercialization of their product. These regulations typically demand the manufacturer to demonstrate product safety, and this is typically performed by applying (on itself voluntary) international standards, as they provide secured methods for implementing certain features in technology, for example procedures on how to implement, analyze and demonstrate safety of new devices before they enter the market. For example product safety for medical devices that belong to the class of Medical Electrical Equipment or Medical Electrical Systems is in principle defined in the IEC 60601-1 Edition 3.1 [4]. In fact, ‘the 60601’ is a large family of standards which includes specific standards for certain aspects or certain types of
Safety Standardization of Wearable Robots …
191
medical devices. Wearable Robots that are to be sold would typically choose to comply with this standard in order to demonstrate their safety. This standard is based on a risk assessment and risk management approach, and basically demands from the manufacturer to identify the main risks of use of their specific device. In this context it provides general points of attention on different aspects and different potential hazards (electrical, mechanical) as well as indicates specific mitigation measures. Concerning regular industrial robots the ISO 10218-1 [5] for robots and ISO 10218-2 [6] for robot systems and integration set similar safety standards, complemented by ISO/TS 15066 [7] concerning collaborative robot safety. Although these standards all contain elements that apply also for Wearable Robots, they are not specific and sometimes not adequate for Wearable Robots. In order to address these and other gaps in the standardization, Working Groups under ISO and IEC flag are addressing such new developments. Under the recently established ISO TC299 Robots and Robotic Devices, several working groups are active in standardization topics relevant to Wearable Robots. For Medical Devices these groups collaborate with IEC TC62 (sub A for general and sub D for particular aspects). Between ISO and IEC the ACART (Advisory Committee on Applications of Robot Technology) has been established to coordinate the future approach to robotics on a higher level. Concretely the currently active groups of interest to Wearable Robots are: ISO/TC 299/WG 2 Personal care robot safety ISO/TC 299/WG 3 Industrial safety ISO/TC 299/JWG 5 Joint ISO/TC 299—IEC/SC 62A—IEC/SC 62D: Medical robot safety; With subgroups: JWG 35: Surgical Robot Safety and JWG 36: Rehabilitation Robot Safety ISO/TC 299/WG 6 Modularity for service robots A first concrete recent output of this work relevant to Wearable Robots, is the publication of ISO 13482 [8]: Robots and robotic devices—Safety requirements for personal care robots, that covers wearable robotics like exoskeletons under the term “restraint-type physical assistant robots”. Beside of this worldwide work, currently, as far as known, only Japan is working on their own standards, as the Japanese market is ahead on certain applications. For example JIS B 8446-2 Part 2: Low powered wearable type physical assistant robot is highly relevant to Wearable Robots. An important landmark in the Wearable Robot regulations is the recent publication of the US Food and Drug Administration (FDA, main regulator of medical devices in USA) on exoskeleton classification [9], where the FDA sets out to classify exoskeletons as “class II (special controls)”, describing how to demonstrate safety of such devices and which special controls are in place. This actual clearance procedure for the US marker requires extensive subject testing to demonstrate safety, and the permitted product use is limited. At this moment, to have
192
J.F. Veneman
exoskeleton type devices cleared by FDA or CE marking, usually requires extensive subject trials, as manufacturers cannot demonstrate their product’s safety by complying with specific standards.
3 Need for Testing Procedures Although, for medical devices, patient trials may always be needed for novel devices to demonstrate their effectivity, Wearable Robots that fit prior established categories, or in industrial or personal care domains, could largely benefit from standards that prescribe specific testing procedures that allow benchmarking product safety. Testing methods are for example available for orthoses and prostheses in ISO 22523 [10]: “External limb prostheses and external orthoses - Requirements and test methods”. This standard includes specific test methods and requirements for example for strength of orthoses. However, this standard as is, does not address any specific properties that relate to robotic, or active orthoses and prosthesis, which fall under the definition of Wearable Robots. As an example: for Wearable Robots one may want to determine interactive force and displacement/velocity generated at the contact surfaces between the wearable robot and the human skin to ensure that the physical stress or strain to the user will remain at safe levels during normal operation. Which levels are safe should be based on validated experimental data, gathered from test subjects. Such a hypothetical test might further require the design of a specific dummy that copies all essential characteristics of a human limb, as well as being equipped with sensor to measure for example the pressure distribution. Preferably one would have an active dummy that can generate realistic movement so that the reaction forces to normal human movements can be determined in the dummy.
4 Conclusion This abstract summarizes the current state of regulation regarding product safety as applies for Wearable Robots, as well as the ongoing work in safety standardization in ISO and IEC context. In the last section the interest of establishing specific testing procedures and equipment that would facilitate safety benchmarking of Wearable Robots is indicated, and first illustrative examples are sketched. We believe that definition of such benchmarks and development of specific testing equipment could greatly facilitate further market introduction of safe Wearable Robots.
Safety Standardization of Wearable Robots …
193
References 1. Bertolini, A., Palmerini, E.: Regulating robotics: a challenge for Europe. Upcoming Issues of EU Law, 94–129 (2014) 2. Medical Device Directive; Official Journal of the European Communities, L 169, 12 July 1993 3. Machinery Directive; Official Journal of the European Communities, L 157, 9 June 2006 4. IEC 60601-1:2005; Medical electrical equipment—Part 1: general requirements for basic safety and essential performance 5. ISO 10218-1:2011; Robots and robotic devices—safety requirements for industrial robots – Part 1: Robots 6. ISO 10218-2:2011; Robots and robotic devices—safety requirements for industrial robots— Part 2: robot systems and integration 7. ISO/TS 15066:2016; Robots and robotic devices—collaborative robots 8. ISO 13482:2014; Robots and robotic devices—safety requirements for personal care robots 9. FDA publication on Medical Devices; Physical Medicine Devices; Classification of the powered exoskeleton (2015). https://federalregister.gov/a/2015-03692; 80 FR 9600; pp. 9600–9603 10. ISO 22523:2006; External limb prostheses and external orthoses—requirements and test methods
The Potential and Acceptance of Exoskeletons in Industry Michiel P. de Looze, Frank Krause and Leonard W. O’Sullivan
Abstract Worldwide, a significant interest in wearable robots or exoskeletons does exist, also from an industrial background. This paper provides an overview of assistive exoskeletons that have specifically been developed for industrial purposes. It discusses their potential in increasing performance and flexibility on one hand and in reducing the mechanical loads on workers involved in manual work on the other. From this it is concluded that exoskeletons have the potential to increase performance under specific conditions and to reduce physical loads significantly. However, several technical issues hinder mainstay practical use of exoskeletons in industry until now. One main issue concerns the human-machine interaction which stands in the way of acceptance. This issue and its linkage to ethics and standardization will be discussed during the conference.
1 Introduction Despite the on-going trend in automation in industry, many workers are still exposed to physical workloads due to material handling (over 30 % of the work population in the EU), repetitive movements (63 %), and awkward body postures (46 %) [1]. To reduce the health risk involved, one approach might be the use exoskeletons.
M.P. de Looze (&) F. Krause TNO, Leiden, The Netherlands e-mail:
[email protected] F. Krause e-mail:
[email protected] M.P. de Looze VU University, Amsterdam, The Netherlands L.W. O’Sullivan University of Limerick, Limerick, Ireland e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_32
195
196
M.P. de Looze et al.
An exoskeleton can be defined as a wearable, external mechanical structure that enhances the power of a person. The main application area of exoskeletons has been for medical/rehabilitation purposes where the devices are aimed to support physically weak, injured, or disabled people [2]. In this paper, we determined the potential of exoskeletons to be useful in industry settings from the perspectives of industrial performance and workers health, and address the shortcomings as it comes to acceptance. To that purpose, we performed a stakeholder analysis and a literature review. The issue of acceptance in relation to ethics and standardization will be discussed during the conference on the basis of field tests in the automotive industry with the so-called RoboMate exoskeleton.
2 Methods 2.1
Stakeholder Analysis
Eight experts in the field of manufacturing industry who could elaborate on the business and social needs and the value of human work in production environments were approached and consulted by interview. They were interviewed about the main general trends in manufacturing industry, about the current and future developments in the value of human work within these trends, and about the needs in relation to any type of exoskeleton to be potentially used on the manufacturing shop floors from the business perspective.
2.2
Literature Review
A literature review was performed by an electronic literature search using the Scopus search engine. We focused on exoskeletons developed for use in occupational fields to support shop floor workers perform physically demanding activities. Papers considering other applications, e.g. rehabilitation, medical, tele-operations, military, and virtual reality, were excluded. We included all types of exoskeletons, i.e. passive and active, anthropomorphic or not, and lower body, upper body and full-body exoskeletons. Single-joint exoskeletons covering the hand and wrist only, were excluded.
2.3
Acceptance
The issue of acceptance will be discussed based on the outcome of the field tests with the RoboMate exoskeleton, which has been developed recently and will be tested in the automotive industry during summer time in 2016.
The Potential and Acceptance of Exoskeletons in Industry
197
3 Results 3.1
Stakeholder-Analysis Results
The following trends in manual work, of relevance for the potential use of exoskeletons in manufacturing, were stressed by the stake-holders. • Due to mechanization and automation, manual work on the shop floor decreases, while more people are involved in the pre-production planning, programming, engineering • Due to increased needs for flexibility, more human workers are required on the shop floor, particularly in highly-automated environments • Due to the ageing of the work force, technology-based support is required to keep older workers productive and safe • Due to the shortage on the labour market, one should increase the attractiveness of factory work for new employees. • Due to prevalence of injury (back, shoulder, arms), technology-based support is required to reduce the physical loads on workers. Main business and social needs of relevance for the application of exoskeletons put forward in the interviews: • • • • •
Need Need Need Need Need
to to to to to
increase the flexibility of production increase the flexibility of workers increase the productivity of manual work increase the quality of manual work reduce the physical load and the risks of injury.
The stakeholders indicate that an exoskeleton could be most useful in the following situations: • Flexible production cannot be achieved through automation because of frequent changes of activities, product types and order sizes, • Weights of parts are just too large to be safely handled by a person while task execution is not on a specific location (thus, local cranes are not an option), • Preventions of manual handling injuries is required. Finally, it was stressed that the main benefit of the application of an exoskeleton above any type of robot system (classical robots, full-automation systems or humanoid robots), would be that, specifically in dynamic environments, one will fully profit from the human’s creativity and flexibility, while he is the one in charge, and there is thus no need for robot programming or teaching of robots.
198
3.2
M.P. de Looze et al.
Literature Review Results
The search resulted in 40 papers in which an exoskeleton with an industrial purpose was described. In these papers a total of 26 different industrial exoskeletons were described These were broken down as 20 upper body, 4 full body, and 2 lower body exoskeletons, with 19 being active (among others [3–8]) and 7 passive (among others [9–13]). These exoskeletons were most frequently aimed to give support in stooped working postures, static holding of a load, and dynamic lifting (and lowering) of a weight. Some studies also mentioned carrying as an activity to be supported. Finally, some job specific activities were mentioned, i.e. patient lifting and transfer (for three different exoskeletons), construction work, agricultural and overhead carpentry work. For 13 out of the 26 industrial exoskeletons, some evaluations of the physical load reductions were performed. Generally, positive effects, either tested statistically or not, have been reported for the physiological (EMG) and biomechanical parameters, both for the passive and the active exoskeletons. For the passive exoskeletons, 10 to 40 % reductions in back muscle activity during dynamic lifting and static holding have been reported. For the active exoskeletons, muscle activity reductions up to 80 % have been reported.
4 Discussion and Conclusions Exoskeletons may be specifically useful where workers have to handle weights which are above health threshold limits, where materials handling takes place at various spots (worker needs to be mobile), and where the production tasks are too difficult to fully automate. Exoskeletons have the potential to considerably reduce the underlying factors associated with work-related musculoskeletal injury. At this point however, several issues hinder acceptance in industry. One main issues, the human-machine interaction, will be discussed in detail on the basis of the outcome of fieldlab testing of the RoboMate exoskeleton, as well as its implications for standardization. Acknowledgments This research was supported by the European shared cost project Robo-Mate, funded under the Seventh Framework Program [grant number FP7-2013-NMP-ICT-FOF].
The Potential and Acceptance of Exoskeletons in Industry
199
References 1. Eurofound, 2012. Fifth European Working Conditions Survey, Publications Office of the European Union, 978-92-897-1062-6. Publications Office of the European Union, Luxembourg (2012) 2. Viteckova, S., Kutilek, P., Jirina, M.: Wearable lower limb robotics: a review. Biocybern. Biomed. Eng. 33(2), 96–105 (2013) 3. Kobayashi, H., Aida, T., Hashimoto, T.: Muscle suit development and factory application. Int. J. Autom. Technol. 3(6), 709–715 (2009) 4. Muramatsu, Y., Kobayashi, H., Sato, Y., Jiaou, H., Hashimoto, T., Kobayashi, H.: Quantitative performance analysis of exoskeleton augmenting devices-muscle suit-for manual worker. Int. J. Autom. Technol. 5(4), 559–567 (2011) 5. Kim, W.S., Lee, H.D., Lim, D.H., Han, C.S.: Development of a lower extremity exoskeleton system for walking assistance while load carrying. In: Proceedings of the Sixteenth International Conference on Climbing and Walking Robots, Sydney, Australia, pp. 35–42, 14–17 July 2013 6. Kawabata, T., Satoh, H., Sankai, Y.: Working posture control of robot suit HAL for reducing structural stress. In: 2009 IEEE International Conference on Robotics and Biomimetics (ROBIO), pp. 2013–2018. IEEE (2009) 7. Yu, W., Rosen, J.: A novel linear PID controller for an upper limb exoskeleton. In: 2010 49th IEEE Conference on Decision and Control (CDC), pp. 3548–3553. IEEE (2010) 8. Kadota, K., Akai, M., Kawashima, K., Kagawa, T.: Development of power-assist robot arm using pneumatic rubbermuscles with a balloon sensor. In: The 18th IEEE International Symposium on Robot and Human Interactive Communication, 2009. RO-MAN 2009, pp. 546–551. IEEE (2009) 9. Abdoli-Eramaki, M., Stevenson, J.M.: The effect of on-body lift assistive device on the lumbar 3D dynamic moments and EMG during asymmetric freestyle lifting. Clin. Biomech. 23, 372–380 (2008) 10. Barret, A.L., Fathallah F.A.: Evaluation of four weight transfer devices for reducing loads on the lower back during agricultural stoop labor. Paper number 01–8056 of the ASAE Meeting, Sacramento, USA (2001) 11. Godwin, A.A., Stevenson, J.M., Agnew, M.J., Twiddy, A.L., Abdoli-E, M., Lotz, C.A.: Testing the efficacy of an ergonomic lifting aid at diminishing muscular fatigue in women over a prolonged period of lifting. Int. J. Ind. Ergon. 3, 121–126 (2009) 12. Lotz, C.A., Agnew, M.J., Godwin, A.A., Stevenson, J.M.: The effect of an on-body personal lift assist device (PLAD) on fatigue during a repetitive lifting task. J. Electromyogr. Kinesiol. 19(2), 331–340 (2009) 13. Sadler, E.M., Graham, R.B., Stevenson, J.M.: The personal lift-assist device and lifting technique: a principal component analysis. Ergonomics 54(4), 392–402 (2011)
Wearable Robots: A Legal Analysis Andrea Bertolini
Abstract The paper provides a complete overview of the legal issues involving wearable robots, in particular ranging from their definition and applicable regulation at the European level, civil liability issues of producers and users as well as a brief approach to issues of human enhancement.
1 Introduction Wearable robots are sophisticated devices intended to substantially improve the quality of life of users, by helping them recover lost functions or acquire new ones. However, their novelty also triggers a series of legal concerns that are not easily addressed through existing regulation. The first necessary effort is thus that of describing the legal framework within which design needs to occur, so as to possibly minimize costs—ex ante and ex post costs associated with legal rules—as well as understand the ethical implications of human enhancement and how the legal system might react to that, by enabling or limiting these technologies. The analysis addresses the legal definition of wearable robots, liability rules and insurance law issues, and human enhancement.
Scientific Independence of young Researchers (SIR) project RELIABLE contributed to the financing of this research. A. Bertolini (&) Private Law Scuola Superiore Sant’Anna (SSSA), Pisa, Italy e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_33
201
202
A. Bertolini
2 Legal Definitions Wearable robots are products1 intended to be implanted or worn by the human user to substantially improve the quality of life of people with disabilities and other users intending to recover lost functions or acquire new ones.2 The wide variety of devices which fall under this definition can however be framed in quite different ways in a legal perspective, in particular with respect to European directives regulating safety standards. A wearable robot qualifies as a medical device for the purposes of the Medical Device Directive3 (in particular so long as an exoskeleton is concerned) or as an Active Implantable Medical Device4 (should prostheses requiring some sort of medical procedure to be installed onto the human body be included within the broad definition). All these regulations address safety standards the devices need to conform to in order to receive the EC marking enabling their distribution on the European market. It is however clear that those directives were not specifically tailored for wearable robots and it could be disputed—both from a technological and legal perspective— whether those rules are indeed adequate. For instance, it could be claimed that adequate technological standards for safety purposes should directly address the control mechanism and its degree of invasiveness, providing detailed benchmarks for each. At the same time, it shall be stressed that while the CE marking is strictly necessary to enable the distribution of the device it does not suffice to exclude liability once the product is used and eventually involved in an accident.
3 Liability and Insurance Being a product, a wearable robot—no matter what specific kind—falls under the application of the European Defective Product Directive.5 Pursuant to this regulation, the producer—meaning the legal person manufacturer of a finished product, the producer of any raw material or of a component part and any person who presents himself as producer by trademark or name—is objectively responsible of all damages deriving from the use of his device (art. 2). Therefore, should any accident occur when the user wears the device the producer might be called into compensate damages suffered. Given the wide variety of 1
Bertolini [1]. Bertolini [2]. 3 Council Directive of 14 June 1993 concerning medical devices (henceforth MDD), OJ L 169, 12.07.1993, pp. 1–43. 4 Council Directive of 20 June 1990 on the approximation of the laws of the Member States relating to active implantable medical devices (90/385/EEC) (henceforth AIMDD), OJ L 189, 20.7.1990, p. 17. 5 Council Directive 85/374/EEC of 25 July 1985 on the approximation of the laws, regulations and administrative provisions of the Member States concerning liability for defective products (henceforth DPD). 2
Wearable Robots: A Legal Analysis
203
situations where a wearable robot could be used, it is hard to anticipate the kind of damages the producer might be bound to repay. Indeed, a wearable robot is intended to follow the user in all day activities, whose range is limitless. The same technological failure however, could lead to extremely different consequences, according to the action performed in that moment. Damages could be trivial if the action is simple and does not involve others, but could also amount to something severe if the activity underwent entails the collaboration with other persons or the use of a device— eventually even a vehicle. For instance, a misleading interpretation of the biological signal utilized to control the robot by the human-machine interface, might lead to different accidents, such as the falling of the wearer or a crash. This dramatically increases ex ante uncertainty and causes the insuring of the devices to be problematic. Moreover, insurance companies calculate premiums by taking into account the likelihood of the occurrence of a harmful event and the expected economic loss they would face in that case. Lack of sufficient statistical data about the frequency of malfunctioning, the amount of damages normally caused, and the application of existing norms—given the absence of precedents and pertinent case-law—might impede that operation. This will cause a delay in the offering of adequate insurance products for wearable-robot manufacturers, exposing them to greater legal risks, ultimately disincentivizing the marketing of such devices.
4 Human Enhancement Wearable robots represent a relevant technology for the purpose of improving the overall living conditions of many individuals, in particular and foremost persons with disabilities. Favouring their development on the side of national and European legislators might also be deemed a legal obligation as emerging from art. 4, let. g of the United Nations Convention on the Rights of People with Disabilities (UNCRPD). However, at the same time those technologies could be used for pure enhancement purposes.6 Enhancement7 is not per se illicit, however it needs to be framed under existing constitutional and supranational principles. These include the notion of human dignity (art. 1 EUCFR)8 and of equality (art. 20 EUCFR)9 but also principles of self determination. How these balance out is to be ascertained in the single case at hand, nonetheless the possibility that regulations restrain the possibility to resort to enhancement through wearable robots needs be considered.
6
Lucivero and Vedder [5]. Coenen et al. [3]. 8 Feldman [4]. 9 Swindells [6]. 7
204
A. Bertolini
Most certainly if the robot is not permanently attached to the human body and its use would not entail a permanent corporal alteration, then self determination of the user would possibly prevail. Otherwise relevant constitutional concerns might emerge and subsequently limitations implemented either though ad hoc regulation or case law.
5 Final Considerations Wearable robots are products highly regulated at the European level. Applicable norms refer to their safety requirements, upon which their testing needs to take place and which regulate their certification for purposes of resale on the European market. Such norms however, do not exclude the possibility of the producer to be held liable when damages occur as a consequence of the use of the device. These are regulated by the DPD which poses a relevant burden on the manufacturer and researcher. This circumstance, together with the existing limited statistical data available about possible technical failures of the devices causes the seizing of adequate insurance covering to be problematic, ultimately delaying innovation. This stifling effect of liability rules can be analysed in detail and directly depends upon the intrinsic technological and functional characteristics of wearable robots, intended to follow the wearer in most of his daily activities, thence used in multiple, and ex ante non-restrained, scenarios. Finally, because wearable robots might be used as human enhancement devices, their use might be subject to specific limitations due to relevant constitutional principles, which govern bioethics. In particular, a use for non-merely rehabilitative purposes might be seen as problematic, and require attentive analysis, in particular taking into account whether the use of the device demands a medical procedure with permanent or reversible consequences.
References 1. Bertolini, A.: Robotic Prostheses. Guidelines on Regulating Robotics. E. Palmerini (2014) 2. Bertolini, A.: Robotic prostheses as products enhancing the rights of people with disabilities. reconsidering the structure of liability rules. Law Comput. Technol. 29(2–3), 116–136 (2015) 3. Coenen, C., Schuijff, M., Smits, M., Klaassen, P., Hennen, L., Rader, M., Wolbring, G.: Human Enhancement. Science and Technology Options Assessment on Human Enhancement (2009). https://www.itas.kit.edu/downloads/etag_coua09a.pdf 4. Feldman, D.: Human Dignity as a Legal Value—Part I. Public Law, 682 (1999) 5. Lucivero, F., Vedder, A. (eds.): Beyond Therapy vs. Enhancement? Multidisciplinary Analyses of a Heated Debate. Pisa University Press (2013) 6. Swindells, F.: Economic inequality and human enhancement technology. Hum. Mente J. Polit. Stud. (26), 213 (2014)
A Verification Method for Testing Abrasion in the Use of Restraint Type Personal Care Robots Yoji Yamada, Mao Xuewei, Kengo Yoshida, Yasuhiro Akiyama and Shogo Okamoto
Abstract Physical stress hazards are identified as a part of the safety requirements for personal care robots as described in ISO 13482. We conducted a safety verification study to clarify an inherently safe condition region in the shear stress—time relationship: The higher the shear stress is, the smaller number of times is needed for blister generation. For validating the verified safety data, we also built a testbed where a manipulator is used to optimally control the position and force of a cuff in contact with a piece of porcine skin for the purpose of reproducing the contact conditions as close as possible to those obtained when a human wears a robot.
1 Introduction Based upon strong demands from industry, various safety test methods have been developed in Japan in the framework of a nation-wide R&D project for putting personal care robots (PCRs, hereafter) into practical use. The abstract paper reports a part of the project activities the outcome of which includes a compilation and proposal of a V&V (Verification and Validation) document to be standardized in compliance with the safety standard ISO 13482 for PCRs. We investigated to identify 62 items in ISO 13482 that require new tests for one or more of the three types of PCRs. Based upon the investigation, 34 test methods are developed and described in a standard form of V&V test methods. In the test methods, the author’s research group develops the test of physical hazard characteristics for restraint type physical assistant robots (PARs, hereafter). Physical stress hazards are identified from the statement in 5.9.2 of ISO 13482 stating that a PCR shall be designed to minimize or reduce physical stress or strain Y. Yamada (&) M. Xuewei Y. Akiyama S. Okamoto Safety Intelligence Group, Department of Mechanical Science and Engineering, Nagoya University, Nagoya, Japan e-mail:
[email protected] K. Yoshida Kansai Electric Power Co. Inc., Osaka, Japan © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_34
205
206
Y. Yamada et al.
to its user due to continuous use. We concluded through discussions with a Japanese company manufacturing PARs that the risk of generating skin wounds becomes also not negligible in such a use of restraint type PAR as to wear their cuff parts each of which directly contacts human skin without any clothing in between a cuff surface and the skin of the wearer. Later, it was also reported that friction blisters were observed to be generated considerably easily in the use of the PAR by a wearer suffering liver cirrhosis. The following sections describe the test method in detail.
2 Verification Test Method for Abrasion Risk We develop a blister generation method for establishing a safety verification system focusing on skin friction trauma [1]. Porcine skin as experimental subject was excised from the anterior part of a pig’s shank shortly after its death, and delivered to our lab within sixteen hours under 4 degree centigrade. Before the experiments, each of the skin surface was shaved by a scissor, cleaned by alcohol, and the excessive fat underneath was also cut off. In our study, the representative direction (shear direction) of force component is selected for emulating the skin surface rubbing phenomena. For conducting a horizontal reciprocating rubbing action, a rheometer (DHR: TA Instrument Co.) is used to exert oscillating rotational rubbing on the porcine skin with high calibration accuracy. A piece of the porcine skin was extended and firmly clamped between the plate with a sheet of sandpaper (40#) attached and a fixation frame. After a certain period of rubbing under oscillating shear force, the structure of the porcine skin was examined by serial sectioning at intervals of 10 lm and the micro-section is analyzed by haematoxylin and eosin (H&E), and the stained sections were observed with a microscope sequentially. The sections of both original porcine skin without any rubbing and the microscopic appearance of porcine skin after the first rubbing experiment are shown in Fig. 1a, b. It can be seen from the figures that the main epidermal structure of porcine skin, including the stratum corneum, stratum granulosum, stratum spinosum, and epidermal basal layer, keeps intact before rubbing test. To confirm the availability of the verification data by use of porcine skin, the characteristics of blister generation should match those obtained in the previous blister generation test using human skins as subjects, which was conducted by P. F. D. Naylor in mid 1950s [2]. After some discussion on the similarity of mechanical properties for both the human and the porcine skin, we combined the verification data obtained through the use of both kinds of skins. In our experimental model, we set shear stress as a unit of the interactive force between the porcine skin and the rubbing head. We calculated to draw a regression curve from the inherently safe time points after statistical analysis including outlier one, regarded it as a pattern of inherently safe threshold, and combined it with the human subject’s blister generation conditions as shown in Fig. 2.
A Verification Method for Testing Abrasion …
207
(a). Appearance of the structure of an original porcine skin surface (intact stratum spinosum indicated by arrow)
(b). Intra-epidermal clefts (indicated by arrow) after 1200s of rubbing under shear stress value 3.2×104 Pa
Fig. 1 a Appearance of the structure of an original porcine skin surface (intact stratum spinosum indicated by arrow). b Intra-epidermal clefts (indicated by arrow) after 1200s of rubbing under shear stress value 3.2 104 Pa Fig. 2 Combination of porcine skin’s and human subjects’ blister generation conditions
208
Y. Yamada et al.
3 Validation of the Verified Data We conducted an experiment validating a cuff attachment in the use of a PAR after confirming the feasibility of using porcine skin for a safety verification test. We focused on a human subject’s “stand-up and sit-down” motion which is considered to exhibit the largest level of relative displacement between the cuff and the skin surface of the subject. The interaction force between the robot’s cuff and the user’s thigh skin can be exerted repeatedly by the experimental apparatus consisting primarily of a manipulator and dummy skin built for the validation experiments. Figure 3 shows the overview of the experimental apparatus where we adopted an adaptive controller called IFT (Iterative Feedback Tuning) to reproduce on the dummy skin the cuff motion represented by both relative displacement generated and interactive force exerted respectively at a contact point between human skin and the cuff of a PAR [3]. Figure 4 illustrates the block diagram of the manipulator controller and Fig. 5 depicts the force reproduction in the cuff sliding direction. Three pieces of porcine skin were tested during the series of experiments, where every test lasted for 1800s. The smallest contact area between urethane and porcine skin was about 12 10−4 m2 and the largest shear stress applied on the porcine skin was calculated approximately as 2.3 104 Pa to verify the safe operation condition.
Fig. 3 Overview of the experimental apparatus
Fig. 4 Control system of the manipulator
A Verification Method for Testing Abrasion …
209
Fig. 5 Controlled force exerted at a contact point on the cuff in the sliding direction
4 Conclusion In the paper, the authors mainly deal with his recent research and development of a safety verification test method: We have conducted a novel study to discuss a safe condition of not generating blisters at mechanical contact between human skin and a cuff of a restrain type personal care robot. Such wounds as blisters are estimated to have a high risk because the probability of occurrence is high while the severity is low. An increase in the number of porcine skin samples for applying them to the dummy skin—cuff system is a part of ongoing work for biomedical verification with higher accuracy. Acknowledgments This work was supported by NEDO (New Energy and Industry Organization), Japan.
References 1. Mao, X.W., et al.: Development of a novel test method for skin safety verification of physical assistant robots. In: Proceedings of the IEEE International Conference on Rehabilitation Robotics (ICORR’15), pp. 319–324, Singapore, Aug 2015 2. Naylor, P.F.D.: The skin surface and friction. Br. J. Dermatol. 67(7), 239–8 (1955) 3. Yoshida, K. et al.: Development of a safety validation test equipment for severity estimation of wounds caused by physical assistant robot. In: Proceedings of the RSJ/SICE/JSME 20th Robotics Symposia, Karuizawa, Nagano, Japan, pp. 483–488, March 2015 (in Japanese)
Part VIII
Benchmarking in WRs and Related Communities
Kinematic Comparison of Gait Rehabilitation with Exoskeleton and End-Effector Devices Byung-Woo Ko and Won-Kyung Song
Abstract Recently, various gait rehabilitation robots have been used as therapy in clinical fields for stroke, spinal cord injuries, and several neurological disorders. We investigated the kinematic differences with joint trajectories of two types of gait rehabilitation robots, i.e., exoskeleton and end-effector devices. Furthermore, we compared the end-effector device’s stair climbing and descending motions to actual motions. The exoskeleton device shows larger hip and knee angle than the end-effector device during gait. However, exoskeleton ankle joint was restricted in dorsiflexed position. The end-effector device’s stair climbing motion was similar to actual stair motion, although there was a delayed and lower maximum flexion. Compared with the actual motion, the stair descending motion had a lower maximum flexion angle for both hip and knee joints in the end-effector device. In addition, the end-effector device’s ankle trajectory was aligned with the dorsiflexion angle, while descending to the bottom stair.
1 Introduction Gait rehabilitation is a tough task for patients and therapists. Particularly, therapists need to take physical efforts for patients with severe conditions who have difficulty in walking independently. Nowadays, several gait rehabilitation systems are launched in the market for reducing these physical efforts and enlarging the time and
This work was supported by the R&D Program of MOTIE/KEIT [10045164, Development of an overground gait rehabilitation robot technology with a success rate over 90 % in gait intention detection based on biosignal interface for various gait rehabilitation of stroke patients]. B.-W. Ko W.-K. Song (&) Research Institute, National Rehabilitation Center, Samgaksan-Ro 58, Gangbuk-gu, Seoul 10122, South Korea e-mail:
[email protected] B.-W. Ko e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_35
213
214
B.-W. Ko and W.-K. Song
number of gait rehabilitation interventions. These systems consist of a balance support module including body weight support, a robot module, and other modules such as virtual reality module for improving motivations. Gait rehabilitation robots could be mainly divided into two types: exoskeleton and end-effector robotic devices [1]. The exoskeleton-type device has an external structural mechanism with joints and links in correspondence with a human body. However, the end-effector-based device has footplates that are mounted on the robot [2]. The end-effector-based robot does not need to be accurately aligned with the joints. The exoskeleton-based robot operates via smaller end-effector forces [3]. Robot-assisted gait training provides versatile control, but the expensive robot devices lead to efficiency debates in comparison with other conventional training techniques [2]. Some studies show significantly higher rates of independent walking during end-effector-based training compared with exoskeleton-based training [4]. The end-effector-based robot presents new features including climbing up and down the stairs [5]. The purpose of this study was kinematic comparison of gait trajectory of body joints, i.e., hip, knee, and ankle joints, provided by gait rehabilitation robots: exoskeleton-based and end-effector-based robots. In addition, we analyzed the up and down stair climbing motions for the end-effector-based robot. The analyzed data could be used to optimize intervention as well as for designing overground wearable robotic devices.
2 Materials and Methods 2.1
Robot Systems: Exoskeleton and End-Effector Devices
In this study, we used two types of gait rehabilitation robots as shown in Fig. 1. One is Lokomat robotic gait orthosis (Hocoma AG) system that consists of a treadmill, a dynamic unloading system, and two light-weight robotic actuators that attach to the subject’s legs. The hip and knee joints are actuated by small DC motors and linear ball screw assemblies [6]. These motors do not require a high torque.
Fig. 1 Gait rehabilitation robot systems (left Lokomat, right G-EO system)
Kinematic Comparison of Gait Rehabilitation with Exoskeleton …
215
The other is G-EO System Evolution (RehaTechnology). This device allows securing the subjects with a harness while they stand on the footplates of the machine. The footplates have three degrees of freedom each, allowing control of the length and height of the steps and footplate angles [7].
2.2
Procedure and Instrumentation
One healthy male subject with no known neurological injuries or gait disorders participated in this study. All experimental procedures and risks were fully explained prior to his participation. The subject walked at a comfortable pace (2.2 km/h) using the Lokomat, and a physical therapist who has used the Lokomat for over three years adjusted the step length of the Lokomat until the subject felt comfortable with the gait pattern. As with standard clinical practice, the Lokomat was operated in the position control mode with 100 % guidance force. A foot lifter was used in this study for blocking the foot drop during gait. The G-EO system could adjust the ankle dorsiflexion angle at initial contact and the plantar flexion at toe off. Thus, we adjusted the gait speed, step length, and ankle angle of the G-EO until the subject felt comfortable with the gait pattern. The hip, knee, and ankle range of motion in sagittal plane data during gait, stair climbing, and descending were captured using a flexible goniometer (SG Series; Biometrics Ltd) connected to a wireless transceiver (Delsys Trigno). The stair climbing and descending motions were performed on real stairs that consists of total 12 stairs, which had a step height of 17 cm. We used foot-switches under each heel for detecting the gait cycle and start points of stair climbing and descending motions. The start points of stair climbing and descending motions were set to the timing of heel rise to perform each motion, respectively. To comparison of gait motion with robot devices, we added Winter’s gait data [8].
3 Results and Discussion 3.1
Comparison of Gait Motion Trajectory
Generally, Lokomat shows greater hip joint trajectory than G-EO during gait. In late stance phase, G-EO kept the hip flexed and showed the lower hip flexion in swing phase compared with Lokomat. The knee joint trajectory of G-EO was kept flexed throughout gait. At initial contact, G-EO shows greater knee flexion and smaller swing phase than Lokomat. The Lokomat ankle trajectory was small and limited compared with G-EO because a foot lifter is applied to the ankle of the Lokomat. The foot lifter consists of a loop that is fastened around the ball of the foot to prevent the
216
B.-W. Ko and W.-K. Song
Fig. 2 Hip, knee, and ankle trajectories during gait. The vertical line represents toe-off. Normal data means Winter’s [8] gait data
patient’s foot drop in the treadmill while walking. Therefore, the Lokomat ankle trajectory was limited to perform only the plantar flexion (Fig. 2).
3.2
Comparison of Stair Climbing and Descending Motion
G-EO’s stair climbing motion shows similar trajectories to the real stair climbing motion. However, the G-EO indicated a delayed and lower hip flexion compared with the real stair motion. These trends of delayed and lower maximum flexion are displayed similarly for the knee joint. The ankle trajectory of the G-EO shows larger plantar flexion when the foot is lifted until it is laid on the stair while climbing each stair. The stair descending trajectory of the G-EO presents lower maximum hip flexion compared to the real stair descending motion. These trends were also observed for a knee joint. The ankle joint trajectory showed maximum plantar flexion while descending the bottom stair, but the G-EO showed increased dorsiflexed ankle trajectory during the stair descending motion (Fig. 3). The two types of wearable robotic devices can have similar concepts of general stationary gait rehabilitation robots, such as Lokomat and G-EO. A wearable robotic device has an external mechanism, like ReWalk [9] and Robin-H [10], that corresponds with the human joints. Additionally, a wearable device has foot and ankle posture control with respect to the body trunk. The robot gait training is repeat the training in the context of the best obtained by limiting the degree of freedom confined to the artificial within the hardware and software framework of the patients. For the patients who can endure end-effector driven movements with
Fig. 3 Hip, knee, and ankle trajectories during stair climbing and descending (Normal actual stair climbing and descending motions)
Kinematic Comparison of Gait Rehabilitation with Exoskeleton …
217
trunk control, weight bearing and shifting, the end-effector devices allow dexterous ankle movements for the subjects with some motivation of self-determined hip and knee movements. An exoskeleton device effectively blocks foot drop using a foot lifter in the case of severe patients. In view of gait adaptation, exoskeleton device is much better than with end-effector device. However, an end-effector-based device is more effective to encourage a user’s motivation.
4 Conclusion In this study, we confirm the kinematic difference characteristics of the training motion of two types of gait rehabilitation robotic devices. Even though the analysis results are restricted to two devices, the results represent distinguished points for joint kinematics, such as ankle and other joints. For developing overground wearable robotic devices, the kinematic analysis results could be applicable. Further studies are required to evaluate not only the kinematics but also the kinetics of gait rehabilitation robots.
References 1. Díaz, I., Gil, J.J., Sánchez, E.: Lower-limb robotic rehabilitation: literature review and challenges. J. Robot., 1–11 (2011) 2. Sale, P., Franceschini, M., Waldner, A., Hesse, S.: Use of the robot assisted gait therapy in rehabilitation of patients with stroke and spinal cord injury. Eur. J. Phys. Rehabil. Med., 111– 121 (2012) 3. Veneman, J.F., Kruidhof, R., Hekman, E.E.G.: Design and evaluation of the LOPES exoskeleton robot for interactive gait rehabilitation. IEEE Trans. Neural Syst. Rehabil. Eng., 379–386 (2007) 4. Mehrholz, J., Pohl, M.: Electromechanical-assisted gait training after stroke: a systematic review comparing end-effector and exoskeleton devices. J. Rehabil. Med., 193–199 (2012) 5. Hesse, S., Waldner, A., Tomelleri, C.: Innovative gait robot for the repetitive practice of floor walking and stair climbing up and down in stroke patients. J. Rehabil. Med. (2010) 6. Hidler, J., Wisman, W., Neckel, N.: Kinematic trajectories while walking within the Lokomat robotic gait-orthosis. Clin. Biomech., 1251–1259 (2008) 7. Hesse, S., Waldner, A., Tomelleri, C.: Innovative gait robot for the repetitive practice of floor walking and stair climbing up and down in stroke patients. J. Neuroeng. Rehabil. 7 (30) (2010) 8. Winter, D.A.: The Biomechanics and Motor Control of Human Movement, 3rd edn. Wiley, New York (1990) 9. Esquenazi, A., Talaty, M., Packel, A.: The ReWalk powered exoskeleton to restore ambulatory function to individuals with thoracic-level motor-complete spinal cord injury. Am. J. Phys. Med. Rehabil., 911–921 (2012) 10. Heo, W.H., Kim, E., Park, H., Jung, J.-Y.: A gait phase classifier using a recurrent neural network. J. Inst. Control Robot. Syst., 518–523 (2015)
Evaluating the Gait of Lower Limb Prosthesis Users Stephanie L. Carey, Kyle B. Reed, Amanda Martori, Tyagi Ramakrishnan and Rajiv Dubey
Abstract Outcome assessments are vital in facilitating periodic, episodic and ongoing evaluation of persons with limb loss. There are many outcome measures used to quantify prosthetic fit, alignment, comfort, functionality and usability of lower limb prostheses. However, many measures are subjective, difficult to implement in a clinical setting and lack psychometric evaluation. This study used an immersive Computer Assisted Rehabilitation Environment (CAREN) virtual reality system with an instrumented spilt-belt treadmill and real time motion capture system as a research tool to evaluate and compare the gait of lower limb prosthesis users and non-amputees as a preliminary study to determine the effectiveness and appropriate use of outcome measures. The use of the CAREN system providing more real world scenarios such as ramps, inclines and unexpected inclines helped evaluate the hill assessment index (HAI) and the combined gait asymmetry metric (CGAM).
S.L. Carey (&) K.B. Reed T. Ramakrishnan R. Dubey Department of Mechanical Engineering, University of South Florida, Tampa, USA e-mail:
[email protected] K.B. Reed e-mail:
[email protected] T. Ramakrishnan e-mail:
[email protected] R. Dubey e-mail:
[email protected] A. Martori Department of Chemical and Biomedical Engineering, University of South Florida, Tampa, USA e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_36
219
220
S.L. Carey et al.
1 Introduction Lower limb amputations are prevalent in the U.S., often due to complications from vascular issues caused by diseases such as diabetes as well as trauma from war related injuries or motor vehicle accidents [1]. In the United States, there are more than 2 million people who have lost a limb and that number is expected to double by 2050 [1]. On average the healthcare costs are $500,000 per person over a 5-year period following limb loss, and additional prosthesis costs over the 5-year period can reach $450,000, with additional rehabilitative costs [1]. Prostheses are often rejected or underused due to problems with control, function, training, comfort or fit. The prevalence and expenses involved in lower limb amputations necessitate specific and effective tools and outcome measures for prosthesis prescription, evaluation and rehabilitation. Outcome assessments are vital in facilitating periodic, episodic and ongoing evaluation of persons with limb loss. Third party reimbursors are demanding justification and evidence for payment of services. There is a need for valid and reliable outcome assessments to quantitatively measure prosthetic fit and patient performance. The Hill Assessment Index is a 12 level ordinal scale developed to address different characteristics of hill ascent and descent but is a high subjective measure [2]. Learning how a prosthetic leg functions on inclines and declines may assist with prosthetic design and training. Measuring asymmetry of prosthetic gait is also of great importance to prosthesis designers, users and therapists. Most amputees have many asymmetric biomechanical parameters including spatial, temporal, kinetic and kinematic. This is because of the inherent asymmetric change in force and motion capabilities in their limbs. However, kinematic and dynamic symmetry is not possible in an asymmetric system [3], and symmetry may not even be necessary for a gait to be considered normal and unimpaired [4]. Since individuals with amputations are inherently asymmetric, aiming for a gait pattern close to that of a symmetric person may not the ideal approach. To understand how multiple parameters interact, we use the Combined Gait Asymmetry Metric (CGAM) to represent the level of asymmetry using five gait parameters [5].
2 Methods 2.1
The CAREN System
The CAREN system (Fig. 1) consists of D-Flow control software, a 180º projection screen and projectors, a 6-degree of freedom platform with a built-in instrumented dual-belt treadmill, and 10 Vicon motion capture cameras. Reflective markers were positioned according to the Vicon Plug-in-Gait lower limb model and motion capture data were processed in Vicon Nexus. An YXZ cardan sequence with two
Evaluating the Gait of Lower Limb Prosthesis Users
221
Fig. 1 Bilateral transtibial amputee walking on the CAREN system
proximal segments was used to calculate joint angles. In order to compare hip, knee, and ankle flexion of all participants, gait trials were normalized to percent of the gait cycle
2.2
Data Collection and Analysis
A testing protocol (IRB # Pro00018519) was approved by the University of South Florida’s Institutional Review Board to collect data with the CAREN while non-amputees and amputees walk on a treadmill at a self-selected speed at various elevations and with unexpected gait perturbations. Two preliminary studies using the CAREN system to evaluate and compare the gait of lower limb prosthesis users were completed. Five non-amputees, four females and one male with a mean age 26 yr ± 10 participated and one bilateral transtibial amputee (TTA), a 50 year old diabetic male, 15 (leg 1) and 13 years (leg 2) post-amputation participated in the first study. Gait trials included walking on level ground, 5° incline and decline, and a ±3° cross slope for able-bodied and ±2° cross slope for the bilateral amputee. The methods for non-amputees in [6] were similar for the amputee. A second preliminary study was conducted with one non-amputee participant and one 37 year old female with a right unilateral transfemoral amputation fitted with two different socket designs: the IRC and VAS brimless. In order to understand how multiple parameters interact, the Combined Gait Asymmetry Metric (CGAM) was used to represent the level of asymmetry using five gait parameters. In addition to typical walking, the five randomized scenarios were foot slip, tread deceleration, missing step (4° pitch), height change (±5° roll), and unstable ground. The perturbations utilized the split belt treadmill and were performed at three speeds of 0.5, 0.9, and 1.3 m/s. The perturbations were randomized to prevent anticipation by the subject.
222
S.L. Carey et al.
3 Results For the first study, average hip, knee (Fig. 2), and ankle flexion were calculated illustrating a comparison between elevations, as well as a comparison between non-amputee and amputee gait. The bilateral TTA demonstrated significantly reduced flexion compared to the non-amputees across all joints and phases of the gait cycle. The greatest difference occurred in knee flexion where the non-amputee maximum across all elevations was approximately 60° and the amputee maximum was approximately 40°. Hip extension was also about 10° less across all elevations for the TTA and hip hiking was evident during the swing phase. Lastly, there was a reduction in both ankle dorsi and plantar flexion for the TTA. This was to be expected as the amputee did not use a multi-axial prosthetic foot. For the second study, the gait biomechanics with spatial, temporal, kinematic, and kinetic parameters are used to calculate the CGAM. Both prosthetic gaits showed a higher magnitude compared to the able-body gait, as expected. The CGAM was calculated for all perturbations at each speed and are presented for walking at all three speeds. These speeds are shown for comparison to our earlier experiment on able-bodied subjects at the same speeds. Figure 3 depicts the CGAM magnitudes with the respective sockets and speeds.
Fig. 2 Average knee kinematics at various elevations: level (black), uphill (red), downhill (blue), and cross slope (green) walking in non-amputees (solid) and bilateral transtibial amputee (dashed)
Evaluating the Gait of Lower Limb Prosthesis Users
223
Fig. 3 Combined Gait Asymmetry Metric magnitudes for a Able- Body subjects, b Gait with VAS-Brimless Socket, and c Gait with IRC Socket
4 Discussion Providing objective data on how lower limb prosthetic function at various inclines and side slopes will aid in benchmarking to improve prosthetic design and training. These data showed the differences in hip, knee, and ankle kinematics at various elevations with the CAREN, as well as a comparison between normative and bilateral transtibial amputee gait. This information can improve upon subjective outcome measures such as the HAI. CGAM is a quantifiable single number representing gait quality that could serve as a clinical measure of biomechanical parameters of gait rather than a qualitative and subjective perspective. In this presented comparison between sockets and able-bodied individuals, it was demonstrated that there are differences in asymmetry and reducing the CGAM magnitude will result in an improved gait pattern.
5 Conclusion This work demonstrated that the CAREN can be used to measure the functional status of an amputee. Future studies with more subjects will examine outcome measures that can be used to track ability, test different devices, and demonstrate the patient’s need for specific types of prostheses. The effects of various prosthetic components and rehabilitation interventions can also be evaluated.
References 1. Sheehan, T.P., Gondo, G.C.: Impact of limb loss in the United States. Phys. Med. Rehabil. Clin. N Am. 25(1), 9–28 (2014) 2. Hafner, B.J., et al.: Evaluation of function, performance and preference as transfemoral amputees transition from mechanical to microprocessor control of the prosthetic knee. Arch. Phys. Med. Rehabil. 88, 207–217 (2007)
224
S.L. Carey et al.
3. Muratagic, H., Handzic, I., Reed, K.B.: Passive kinematic synchronization of dissimilar and uncoupled rotating systems. Nonlinear Dyn. Syst. Theory 15(4), 383–399 (2015) 4. Handzic, I., Reed, K.B.: Perception of gait patterns that deviate from normal and symmetric biped locomotion. Front. Psychol. 6 (2015) 5. Ramakrishnan, T., Muratagic, H.,Reed, K.B.: Combined gait asymmetry metric. In: 38th IEEE Engineering in Medicine & Biology Conference (EMBC) (2016) 6. Martori, A., Carey, S.: Proceedings of the Biomedical Engineering Society Annual Meeting (2015)
Some Considerations on Benchmarking of Wearable Robots for Mobility Jan F. Veneman
Abstract Wearable Robots for Mobility (WR-Mob), i.e. exoskeletons, are currently entering the market. This makes the topic of how to define and measure their performance more relevant and urgent. This abstract provides some considerations that could be taken into account when designing quantitative benchmark metrics that aim to quantify the performance of WR-Mob, focusing on measurement of reduction of metabolic cost and of improvement of balance. The considerations on metrics and their normalization are first steps to well-defined benchmark tests. Proper benchmarks contribute to solid comparison among devices that can be performed in different labs, and thus support a faster progress beyond the state of the art.
1 Introduction As Wearable Robots for Mobility (WR-Mob), such as exoskeletons for rehabilitation and worker support, are finding their way into the market, it becomes increasingly important to measure their performance. Not only to be able to clearly describe the performance that can be expected from products, but also to compare results obtained in different research labs, to quantify the state of the art performance or to set targets for innovation, or to track intermediate development stages, for example when developing an application according to the Technology Readiness Levels (TRLs) [1]. Performance benchmarking of WR-Mob can be either meant to quantify how much benefit using a specific WR-Mob yields relative to not using it, or to quantify the relative performance of one WR-Mob compared to another. As potential uses This work has been partially supported by the EU funded project BALANCE, FP7-ICT-2011.2.1 grant no. 601003. J.F. Veneman (&) Health Division, TECNALIA Research and Innovation, Mikeletegi Pasealekua 1, 20009 Donostia-San Sebastian, Spain e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_37
225
226
J.F. Veneman
and sought benefits of WR-Mob are very diverse, the benchmarks need to be diverse and relate to the intended use and goal related to the specific WR-Mob application. Exoskeletons are promoted to reduce load on workers in manufacturing or other industrial tasks, to support mobility of impaired or frail people, or to provide safe rehabilitation training, or tele-operation of robots, or to improve range of operation of load-carrying soldiers. Each of such scenarios might need their own specific performance benchmarks, but some benchmarks may be more general and relevant among different applications. General performance topics could for example be the influence of a WR-Mob on the metabolic cost of walking, or the influence of a WR-Mob on the ability to maintain balance. As an example, this abstract shortly reviews approaches to benchmark metabolic cost of walking and balance performance in WR-Mob.
2 Metabolic Cost of Walking One of the holy grails of WR-Mob design is to reduce the effort of walking, usually interpreted as maximally reducing the metabolic cost of walking. This is often measured through measuring the oxygen consumption and carbon dioxide production through pulmonary gas exchange measurements over a certain distance walked, usually on a treadmill. However, even if well performed, such measurements are not without problems; outcomes depend on walking speed, loads carried, total weight and physical condition of the walking subjects, implying that conditions need to be normalized. Care has to be taken that no switch from aerobic to anaerobic metabolic regime or vice versa occurs during an experiment or among comparison conditions. WR-Mob unavoidably add weight to the walking person; a proper WR-Mob is fully ‘self-contained’, that is, all necessary components are worn on-body e.g. power supply, controller, and actuators, with no tether to external systems. As a WR-Mob may contain stored energy, for example in a battery, the weight depends on how much energy is contained, and hence is related to the radius of action, as well as on the efficiency of the energy storage in the sense of amount of stored energy over storage mass. Also this should be normalized; or radius of action should always concurrently be reported. To reduce the effect of some individual differences, such as level of fitness, spirometry values can be converted to Power (P[Watt]) through experimentally determined relations, such as by Brockway et al. [2], after which the metabolic rate of standing can be subtracted from the gross metabolic rate of walking [3]. Additionally this value can be normalized to body mass. The full objective metric E P here is the dimensionless specific Cost of Transport, defined as: CoT ¼ mgd ¼ mgv , which allows comparison of values of efficiency of transport across different velocities and masses. Basically the CoT of the task (e.g. walking) in the natural situation should be compared with the CoT with the assistive WR-Mob, in the same
Some Considerations on Benchmarking of Wearable Robots …
227
external conditions (e.g. with different transported mass between conditions due to the added WR-Mob, as the formula includes transported mass). Nevertheless, as an in-between condition the situation that includes only the added mass of the device but not its power support (unpowered operation or using a mock-up) could be obtained in order to distinguish increase and effective reduction. It should be taken into account that, as walking exploits passive structural properties, the optimal walking speed depends mechanically on the mass and its distribution over the body. This implies poor comparison if devices largely alter such parameters among conditions. This could be dealt with through measuring at self-selected walking speed, supposing that a human automatically selects an energy efficient mode of walking, although this is difficult to ascertain. Sawicki and Ferris [4] presented the “Exoskeleton Performance Index (EPI)”, DNet Metabolic Powerg þ
þ muscle defined as: EPI ¼ Average exoskeleton positive mech: power, with gmuscle the ‘muscular efficiency’; that for example asymptotically approaches 0.25 for the ankle [5, 6]. This metric reaches a value 1 when the exoskeleton fully replaces the work of the underlying biological muscles normally performing the positive mechanical work for the supported joint(s). Obviously, when using this metric, comparisons should be performed with the same added weight among conditions. Finally, it should be noted that training with a device may improve its effectiveness for the trained subject.
3 Balance Performance With the potential use of exoskeletons to improve the ability to deal with perturbations, such as is the target of the EC FP7 BALANCE (www.balance-fp7.eu) research project, also balance performance needs to be benchmarked. As known from studying the human, several subsystems compose the ability to avoid falls, and this composition is different between standing and walking. Subsystems also contribute differently depending on whether there is opportunity to anticipate the specific threats to balance or not. Therefore, the performance benchmark for balance should be designed for the specific purpose of the device and the comparison. Considering the ability to react to unpredictable perturbations, a benchmark requires means to provide random, well-quantified, repeatable perturbations, for example the “pelvic pusher device” developed within the BALANCE project by the University of Twente, which can provide perturbations designed with a specific force-over-time profile [7]. Reactions to perturbations can probably best be benchmarked by basic metrics such as “maximum perturbation that can be dealt with” (as impulse: force integrated over time) in standing or in walking, comparing the conditions “using the WM-Mob with supportive control” and “human only”. Like in the metabolic cost benchmarks, wearing the WM-Mob without supportive control, but with added weight (device
228
J.F. Veneman
powered off, or operated in transparent mode; – or using a mock up), can be considered as intermediate condition of comparison. To compare “maximum perturbation that can be dealt with” among conditions that are substantially different, requires adequate normalization of the perturbation impulse. Important for normalization are: the size of the Base of Support, the height of Center of Mass (or approximated by leg length or body length). In walking, especially concerning the duration of the perturbation, also the speed of walking, and for timing additionally the moment in the gait cycle should be considered. These parameters should be used to design a perturbation that can be considered equivalent among conditions. Gait relevant values are normalized by: m = body mass (“total biped”, including WR-Mob - hence may change importantly through a WR-Mob), l = lower-limb length (hip-joint to floor) and g = gravitational constant. Proper normalization factors can be found in [8], leading to a normalization of perturbation impulse with: This also implies for the perturbation angular impulse: J ^ Jang ¼ ang 3 1 . Additionally, care has to be taken that the size of the Base of Support, ml2 g 2
defined as distance from centered CoG to the edge of the Base of Support in the direction of the perturbation (dBoS,P) should be similar among conditions or lengths qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 may be more correctly normalized to l2 þ dBoS;P , instead of to l being the lower-limb length.
4 Conclusion This abstract provides some considerations that could be taken into account when designing quantitative benchmark metrics that aim to quantify the performance of Wearable Robots for Mobility, focusing on measurement of reduction of metabolic cost and of improvement of balance. The considerations are first steps to well defined benchmark test through which such performance can be objectively benchmarked and properly compared among devices and performed in different labs.
References 1. Torricelli, D., Gonzalez-Vargas, J., Veneman, J.F., Mombaur, K., Tsagarakis, N., del Ama, A.J., Gil-Agudo, A., Moreno, J.C., Pons, J.L.: Benchmarking bipedal locomotion: a unified scheme for humanoids, wearable robots, and humans. IEEE Robot. Autom. Mag. 22(3), 103– 115 (2015) 2. Brockway, J.M.: Derivation of formulae used to calculate energy expenditure in man. Hum. Nutr.: Clin. Nutr. 41C, 463–471 (1987) 3. Mooney, L.M., Rouse, E.J., Herr, H.M.: Autonomous exoskeleton reduces metabolic cost of human walking. J. Neuroeng. Rehabil. 11(1), 151 (2014). http://dx.doi.org/, doi: 10.1186/17430003-11-151 4. Sawicki, S., Ferris, D.P.: Mechanics and energetics of level walking with powered ankle exoskeletons. J. Exp. Biol. 211(Pt. 9), 1402–1413 (2008)
Some Considerations on Benchmarking of Wearable Robots …
229
5. Margaria, R.: Positive and negative work performances and their efficiencies in human locomotion. Int. Z Angew. Physiol. Einschl. Arbeitsphysiol. 25, 339–351 (1968) 6. Margaria, R.: Biomechanics and Energetics Of Muscular Exercise. Clarendon Press, Oxford (1976) 7. Vlutters, M., Van Asseldonk, E.H.F., Van der Kooij, H.: Center of mass velocity-based predictions in balance recovery following pelvis perturbations during human walking. J. Exp. Biol. 219(10), 1514–1523 (2016) 8. Hof, A.L.: Scaling gait data to body size. Gait Posture 4, 222–223 (1996)
Benchmarking Data for Human Walking in Different Scenarios Katja Mombaur, Debora Clever and Alexander Schubert
Abstract Benchmarking bipedal locomotion is an important topic not only for wearable robotics, but also for human movement analysis and humanoid robotics. In this paper we discuss how data from the KoroiBot walking database can be used to establish benchmarking data for human walking in different situations based on direct extraction and on model-based analysis. The goal is to identify human reference values for important quantities defining walking motions which can then serve for evaluating walking motions in the different areas of applications.
1 Introduction Developing efficient measures for benchmarking bipedal locomotion is important for different areas of research, including humanoid robotics, human motion studies in biomechanics or medical application, exoskeletons etc. Among other things, benchmarking allows to compare different robot platforms, technologies or control algorithms, define walking standards, measure progress of human walking during therapy, define training goals etc. The importance of benchmarking has also been acknowledged by the EU and play an important role in several European projects, such as H2R, KoroiBot, WalkMan, Balance and BioMot. The interest in benchmarking of bipedal locomotion has also lead to the formation of the benchmarking initiative [1]. [2] is an attempt to formulate a unified benchmarking scheme based on the views of the five mentioned European projects as well as the results of an online survey performed via the benchmarking list. In [3] we have described the benchmarking approach in the KoroiBot project which does not have the goal to compare robots or technology but instead focuses on key performance indicators to measure The research leading to these results has received funding from the European Union Seventh Framework Program FP7 under grant agreement n◦ 611909 (KoroiBot). K. Mombaur (B) · D. Clever · A. Schubert Robotics & Biomechanics Research Group, Interdisciplinary Center for Scientific Computing, Heidelberg University, Heidelberg, Germany e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_38
231
232
K. Mombaur et al.
progress of individual robot performance in the project. We distinguish technical indicators, computational indicators and high-level indicators such as human-likeness. Especially the first group which is of general interest and has also been extensively discussed in [2]. Technical indicators include general walking indicators such as walking speed, energy consumption and efficiency, or robustness, but also evaluate the particular walking scenario, such as stair height, slope angle etc. If such technical indicators are to be used for robots or other technical systems and their performance is to be compared to human performance, extensive reference values of human performance will be urgently needed. In this talk, we discuss how some of these values can be extracted from human motion capture databases.
2 The Koroibot Project and the Koroibot Motion Capture Database The research presented here has been performed in the context of the KoroiBot project [4] which addresses the issue that todays humanoids still lack the ability to walk in a human-like fashion. This is however crucial for many of the tasks they are supposed to perform, namely support humans in households or in the working environment, or to replace them in places that are too dangerous or too far away, such as disaster sites or space missions. Human gaits are at the same time efficient, robust and versatile but gaits of humanoids or bipedal robots are at best good in one of these areas. This problem is not only linked to the present hardware, but also to a large extent to the control principles and the software used. The idea of KoroiBot is to address this issue by developing novel techniques for the generation and control of motions and applying them to different robot platforms. The KoroiBot methodology is based on a thorough experimental study of human walking in different situations and the development of mathematical models as well as optimization and learning methods. One general idea is to perform the transfer from human to humanoid motions not by a direct mapping of motions which would not be feasible due to the serious constraints of humanoid robots - but via the models and corresponding optimization principles for motion generation and control.
Fig. 1 The KoroiBot walking parcours with different walking scenarios for which also motion recordings have been performed: level ground, rough and soft terrain, beams, easaws, step stones and stairs up and down
Benchmarking Data for Human Walking in Different Scenarios
233
In this context, extensive experimental studies of human walking motions have been performed. This includes many unconstrained and constrained walking scenarios that are important for the KoroiBot project, such as walking on level ground, slopes, stairs up and down, beams, step stones, sea-saws, soft mattresses etc. (see sketch in Fig. 1). Data sets include motion capture data from optical systems in all cases, as well as additional force or pressure distribution data in the feet for some of the sets. Recordings have been performed by KoroiBot partners University of Tuebingen, Weizmann, KIT and LAAS-CNRS. The data is publicly available in the KoroiBot database which comprises by now more than 5000 data sets and is still growing (see [5, 6] for the technical implementation of the database.
3 Human Walking Reference Data In the talk, we will present evaluations of many data sets of the KoroiBot database. Some characteristics can be directly extracted from marker information or force / pressure measurements in the database, others have to be derived via some simple mathematical models. Among others, we are investigating the different properties: • Walking speed: The average walking speed of human subjects is extracted for different scenarios and as a function of the level of difficulty of the particular scenario (e.g. increased slope angles or stair heights). We look at absolute walking speed and relative speed normalized by the subject’s leg length. • Step width: Here the interest is to evaluate to what extent increased difficulty of the scenario leads to an increased step width. • Walking speed variance in difficult terrain: The goal is to evaluate if in difficult scenarios, such as very rough terrain or step stones, subjects tend to maintain their average walking velocity or if they highly adjust it to the requirements of the specific terrain. • Joint trajectories: Different analysis of the individual joint trajectories of humans in different situations are performed. This also includes a study of intersegmental coordination, i.e. the relationship between absolute thigh, shank and foot angles. • Properties related to simple stability criteria, such as the capture point and the center of pressure. • Developments of different types of energy over the course of the motion. • Closness to periodicity for constrained and unconstrained motions.
4 Conclusion & Outlook Together with numbers extracted from other data sources in a similar way, the numbers presented in this talk can later serve as reference data to benchmark how close the performance of humanoid robots or other technical systems such as exoskeletons
234
K. Mombaur et al.
gets to typical human-like behavior, or how much a patient in therapy still has to train in order to be back at his normal walking capabilities. We also work on analyzing advanced stability properties of recorded human data which however requires sophisticated multi-body system models and optimal control techniques and is subject of another talk submitted to ICNR. For analyzing human-likeness of motions of systems with different kinematic and dynamic properties we have developed an online survey with a similarity rating tool [7] that allows to compare captured motions and different types of modified motions. The outcome of the survey will then be used to further refine the definition of the high level benchmarking criterion “human-likeness”.
References 1. Benchmarking initiative: www.benchmarkinglocomotion.org, 2013–2016 2. Torricelli, D., Gonzales-Vargas, J., Veneman, J., Mombaur, K., Tsagarakis, N., del-Ama, A.J., Gil-Agudo, A., Moreno, J.C., Pons, J.L.: Benchmarking bipedal locomotion in humanoids, wearable robots and humans: a unified scheme. IEEE Robot. Autom. Mag. 103115, (September 2015) 3. Schubert, A., Clever, D., Mombaur, K.: Key performance indicators for humanoid walking motions defined in the KoroiBot project. In: International Workshop on Wearable Robotics (WeRob 2014) (2014) 4. The Koroibot project: www.koroibot.eu, 2013–2016 5. KoroiBot Motion Capture Database: https://koroibot-motion-database.humanoids.kit.edu/, 2014–2016 6. Mandery, C., Terlemez, O., Do, M., Vahrenkamp, N., Asfour, T.: The KIT whole-body human motion database. In: IEEE International Conference on Advanced Robotics (ICAR 2015), 329–336 (2015) 7. Walking similarity, online survey: https://orb.iwr.uni-heidelberg.de/ratingapp/similarity/ (2016)
Clinical Gait Assessment in Relation to Benchmarking Robot Locomotion J.H. Buurke, J.F. Veneman and D. Torricelli
Abstract The objective of this paper is to identify clinical assessments that are potentially useful for benchmarking of bipedal locomotion. Results: Several clinical measures for static conditions and clinical measures for motion are suggested. Conclusions: Potentially useful clinical measures are identified. New metrics coming from new ambulant measurement techniques may be a good alternative for the qualitative clinical measures.
1 Introduction In a recent paper on benchmarking bipedal locomotion, Torricelli et al. state that there is a growing awareness of the importance of benchmarking in the field of robotics [1]. The objective of the paper of Torricelli et al. was to define the basis of a benchmarking scheme for the assessment of bipedal locomotion that could be applied and shared across different research communities. For that purpose, a web-based survey was carried out first. Secondly, a common nomenclature was defined based on the work of Gentile [2] and Fleishman [3]. A common nomenclature is crucial, since the target
J.H. Buurke (&) Rehabilitation Technology Research Group, Roessingh Research and Development, Enschede, The Netherlands e-mail:
[email protected] J.H. Buurke Biomedical Signals and Systems, MIRA-Institute for Biomedical Technology and Technical Medicine, University of Twente, Enschede, The Netherlands J.F. Veneman Health Division, Tecnalia Research & Innovation San Sebastián, Madrid, Spain e-mail:
[email protected] D. Torricelli Neural Rehabilitation Group, Cajal Institute, Spanish National Research Council (CSIC), Madrid, Spain e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_39
235
236
J.H. Buurke et al.
group is multidisciplinary and different terms and definitions are often used by various disciplines working in the field of robotics. Subsequently, motor skills were classified, based on Fleishman, listing a subset of significant motor skills related to lower-limb motion and identifying related benchmarks. However, most of these benchmarks describe the biomechanical property of the ability. This paper aims to identify clinical assessments useful for benchmarking of bipedal locomotion. Therefore, the proposed common nomenclature will first be compared to definitions used in clinical practice. Subsequently, potentially useful clinical assessments for benchmarking will be selected, based on their suitability for application in clinical practice.
1.1
Taxonomies Related to International Classification of Functioning
In the proposed scheme for benchmarking of bipedal locomotion, three terms were selected to define a common nomenclature (Fig. 1), based on the work of Magill [4]. These terms are: (a) Skill, defined as a task or activity with a specific goal (i.e. walking); (b) Ability, defined as the independent functional blocks needed to achieve a skill; (c) Performance, defined as the level of achievement of the goal. In healthcare measures are generally defined according to the International Classification of Functioning (ICF) [5]. The ICF is a multipurpose classification designed to serve various disciplines and different sectors and to establish a common language for describing health and health-related states in order to improve communication between different users, such as healthcare workers, researchers, policymakers and the public, including people with disabilities. In the ICF three components are defined (Fig. 2): (a) Body functions; these are the physiological functions of body systems (including psychological functions), (b) Activity; this is the execution of a task or action by an individual, (c) Participation; this is the involvement in a life situation. Furthermore the actual behavior of an individual is influenced by (a) Environmental factors, which make
Fig. 1 The basic components of our benchmarking taxonomy: motor skills, motor abilities, and motor performance as suggested by Torricelli et al.
Clinical Gait Assessment in Relation …
237
Fig. 2 Interactions and components of ICF
up the physical, social and attitudinal environment in which people live and conduct their lives, and (b) Personal factors, which are the particular background of an individual’s life and living, and comprise features of the individual that are not part of a health condition or health states (i.e., habits, lifestyle). Although the nomenclature and definitions differ between the two taxonomies they also seem to have a lot in common. For facilitating the discussion on benchmarking of bipedal locomotion, the taxonomy as proposed by Torricelli et al. is used in this paper.
1.2
Clinical Assessments for Bipedal Locomotion
In clinical practice, and especially in the field of rehabilitation medicine, clinical scales are used for decision making and/or evaluation of patients. These scales are used to describe the abilities (i.e., muscle force, sensation, etc.) and skills (i.e., walking, climbing stairs, etc.). Scales used in clinical practice should by reliable, reproducible and valid. A potential problem for benchmarking is that a lot of these scales are specific for a particular diagnosis, meaning that they are only validated for a specific patient group. A crucial prerequisite of clinical scales for benchmarking is that they are suitable for different patient groups (and healthy elderly). Another essential factor, which may be even more important, is the minimal important difference (MID). This MID is defined as the smallest change in a treatment outcome that a patient would identify as important. Last but not least, measurements should be validated in different languages. In general, existing clinical measures don’t distinguish between body posture or body transport (as is suggested in the paper of Torricelli). Measures either focus at static conditions (i.e., static balance), with or without (internal) perturbations, or at conditions where the body is in motion (i.e., walking, climbing stairs). Therefore, the current paper identifies clinical scales based on their corroboration with abovementioned factors concerning practical applicability, that can be used for benchmarking of motor skills related to lower-limb motion.
238
J.H. Buurke et al.
2 Method Outcome measures were identified from a random selection of various databases, such as: • http://www.meetinstrumentenzorg.nl/ • http://www.rehabmeasures.org/rehabweb/allmeasures.aspx?PageView=Shared • http://geriatrictoolkit.missouri.edu/
3 Results Suggested measures for static conditions: Berg Balance Scale (BBS), Timed Balance test (TBT), Reach test (RT). Suggested measures for motion: 10 m walk test (10 MWT), 6 min walk test (6 MWT), Functional Ambulation Category (FAC), Rivermead Mobility Index (RMI), Timed up and Go (TUG), L-Test (LT), Eight walk test (8 WT), Dynamic gait index (DGI), Functional Gait assessment (FGA), Stairs Test (ST), Falls efficacy Scale (FES).
4 Discussion This paper describes the selection of potentially useful clinical measures for benchmarking of bipedal locomotion. A possible disadvantage of these scales is that they are based on observation and that outcomes are often defined on a general level. New quantitative and reliable metrics might offer a solution for this problem. In the INTERACTION (training and monitoring of daily-life physical INTERACTION with the environment after stroke) project (EC FP7 Strep FP7-ICT-2011-7-287351), an unobtrusive and modular system for monitoring of daily life activities of stroke subjects and physical interactions of upper and lower extremity motor function with their environment was developed and validated. First results showed that the data measured by this system, which consists of inertial motion units (IMU’s) and force sensors, correlate very nicely with clinical scales like the BBS, TUG and 10 MWT. Moreover, it provides additional quantitative information that is essential for discrimination between normal and abnormal (compensatory) movements while performing these tests [6]. Furthermore, measurements using this system can be conducted both in a controlled (movement lab) and uncontrolled (at home) environment [7].
Clinical Gait Assessment in Relation …
239
5 Conclusion Potentially useful clinical measures to benchmark bipedal locomotion were identified using the framework described by Torricelli et al. Important factors guiding the selection of clinical measures are generalizability over different patient groups, validity and reproducibility of the results, in combination with the minimal important difference. New metrics coming from new developments in the area of ambulant measurement techniques may be a good alternative for the existing clinical measures.
References 1. Torricelli, D., González-Vargas, J., Veneman, J.F., Mombaur, K., Tsagarakis, N., del-Ama, A. J., Gil-Agudo, A., Moreno, J.C., Pons, J.L.: Benchmarking bipedal locomotion. IEEE Robot. Autom. Mag. Sept. 2015. DOI:10.1109/MRA.2015.2448278 2. Gentile, M.: Skill acquisition: action, movement, and neuromotor processes. In: Carr, J.H., Shepherd, R.B., Gordon, J. (eds.) Movement science: foundations for physical therapy in rehabilitation, pp. 93–154. Aspen Publishers Inc., Rockville, MD (1987) 3. Fleishman, E.A., Quaintance, M.K.: Taxonomies of human performance. Academic Press, Orlando, FL (1984) 4. Magill, R.A.: Motor learning and control: concepts and applications. McGraw-Hill, New York (2007) 5. International Classification of Functioning, Disability And Health: ICF. WHO Library Cataloguing-in-Publication Data. © World Health Organization 2001. ISBN:92 4 154542 9. http://psychiatr.ru/download/1313?view=name=CF_18.pdf 6. van Meulen, F.B., Weenk, D., Buurke, J.H., van Beijnum, B.F., Veltink, P.H.: Ambulatory assessment of walking balance after stroke using instrumented shoes. J. NeuroEng. Rehabil. 13, 48 (2016). DOI:10.1186/s12984-016-0146-5 7. van Meulen, F.B., Klaassen, B., Held, J., Reenalda, J., Buurke, J.H., van Beijnum, B.-J.F., Luft, A., Veltink, P.H.: Objective evaluation of the quality of movement in daily life after stroke. Front. Bioeng. Biotechnol. 3, 210 (2016). doi:10.3389/fbioe.2015.00210
Part IX
Symbiotic Control of WRs
Attention Level Measurement During Exoskeleton Rehabilitation Through a BMI System Álvaro Costa, Guillermo Asín-Prieto, José González-Vargas, Eduardo Iáñez, Juan C. Moreno, Antonio J. Del-Ama, Ángel Gil-Agudo and José M. Azorín
Abstract Brain-Machine Interfaces based on wearable robots’ control have been proposed in the research field for rehabilitation purposes. The combination of both systems allow the performance of more natural movements and a higher level of involvement of patients on their therapy. Studies focused on this topic should face several issues related to the integration of these systems. The current work is meant to test the accuracy of a real time Brain-Machine Interface based on the detection of gait attention during lower limb exoskeletal rehabilitation. Four users performed the experiment wearing an ankle exoskeleton. The system provides a coefficient between 0 and 1 depending on the level of attention experienced by the subject. These results show good similitude between real and decoded attention level.
1 Introduction Several studies suggest that rehabilitation results could be improved by taking advantage of the human ability to generate physical changes in the brain structure [1]. This brain plasticity could be enhanced during rehabilitation by increasing the involvement of patients on their therapy. BMIs have been proposed in the literature to introduce This research has been funded by the Commission of the European Union under the BioMot project - Smart Wearable Robots with Bioinspired Sensory-Motor Skills (Grant Agreement number IFP7-ICT- 2013-10-611695). Á. Costa (B) · E. Iáñez · J.M. Azorín Brain-Machine Interface Systems Lab, Miguel Hernández University, Av. de la Universidad S/N, 03202 Elche, Spain e-mail:
[email protected] G. Asín-Prieto · J. González-Vargas · J.C. Moreno Human Locomotion Laboratory, Neural Rehabilitation Group, Cajal Institute, Spanish National Research Council, Avda. Doctor Arce 37, 28002 Madrid, Spain A.J. Del-Ama · Á. Gil-Agudo Biomechanics and Technical Aids Units, Physical Medicine and Rehabilitation Department, National Hospital for Spinal Cord Injury, SESCAM, Finca de la Peraleda S/N, 45071 Toledo, Spain © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_40
243
244
Á. Costa et al.
this involvement in the rehabilitation process [2]. These systems are focussed on the decoding of the brain response of people in different situations. This information could be used to fit the rehabilitation to the mental state of the user. Current study works on the integration of a BMI to detect the attention level during gait while wearing an ankle exoskeleton. The online BMI presented is based in offline studies previously performed [3]. During systems integration involving the use of BMIs based on electroencephalograpic measurements, the artifact robustness of the system should be evaluated. On the current work, 3 attention levels are going to be measured to provide an attention coefficient. The main goal of this work is to test how accurate is the proposed BMI system during exoskeleton walking. In this case, the BMI system does not provide information to the exoskeleton although this will be the next step of this research. The BMI system is based on the detection of a high gamma band desyncronization produced during a dual task paradigm related to the attention paid to the gait. This selective attention mechanism has been widely applied in other works based on dual tasks paradigms [4, 5].
2 Materials and Methods 2.1 Ankle Exoskeleton An ankle exoskeleton is attached to the dominant foot of the participants during this experiment. This exoskeleton is being developed in the framework of the BioMot Project (Smart Wearable Robots with Bioinspired Sensory-Motor Skills). This exoskeleton is designed to be a biomimetic mechanical structure with compliant characteristics that allows a natural interface between the patient and the robot.
2.2 EEG Acquisition The cortical signals are recorded using 21 active channels located on the scalp through an elastic cap (actiCap, Brain Products, GmbH, Germany). Electrodes are placed on the motor and premotor cortex with the following distribution according to the 10/10 International System: FZ, FC3, FC1, FCZ, FC2, FC4, C3, C1, CZ, C2, C4, CP3, CP1, CPZ, CP2, CP4, P3, P1, PZ, P2 and P4 with a reference in the right earlobe and AF4 electrode as ground. Electrical signals are amplified by a wireless commercial amplifier (actiCHamp, Brain Products, GmbH, Germany) and digitalized at 500 Hz. A hardware bandpass filter between 0.5 and 100 Hz is applied to remove unnecessary signal bandwidth and also a 50 Hz Notch filter is used to remove power line interference.
Attention Level Measurement During Exoskeleton …
245
2.3 EEG Real Time Processing and Feature Extraction The real time acquisition loop records 0.5 s of EEG data on each iteration. Signal is divided in 1 s epochs with a 0.5 s of overlap between them. Each epoch goes to a processing stage. With a previous 30 s offline register, the Segment Maximum Average (SAM) is computed for each channel and used to standardize the data [6]. The power spectral density of the standardized segments is computed using the Maximum Entropy Method (MEM) [7] with a spectral resolution of 1 Hz. An average value of gamma band (from 30 to 45 Hz and from 55 to 90 Hz to avoid the frequencies affected by the 50 Hz Notch filter) is extracted from each channel. Finally these 21 features are concatenated in a single vector and sent to a classification stage. This processing is repeated for each incoming epoch.
2.4 EEG Classification Features vectors are classified using a Linear Discriminant Analysis (LDA) classifier that fits a multivariate normal density to each group, with a pooled estimate of covariance. The classifier take a decisions every 0.5 s. Each decoded attention level is associated to a coefficient value: 0 for low attention level, 0.5 for standard attention level and 1 for high attention level. After each classification, the value is added to the end of a 10 position buffer (5 s). The final classification is obtained by averaging the values from this buffer. Although the use of this averaged value introduce a small delay in the classification (less than 5 s), it helps reducing the influence of misclassified segments and it is not critical as the attention during rehabilitation do not experience huge variation in a short period of time.
2.5 Experimental Protocol The users were instrumented with the EEG equipment and the ankle exoskeleton was adjusted to the specific size of each participant. On each run of the experiment the user was asked to perform three tasks during straight walking. During the first task the participant was asked to perform some mathematical operations shown on a tablet carried by an experimenter (low gait attention level). On the second task, the participant was asked to walk without any distractions (standard gait attention level). On the third task the participant was asked to follow some marks placed on the floor with an unsteady gait step (high gait attention level). Each task lasts for 30 s. Each experimental session is composed of eight runs. Four healthy participants performed the experiment, all of them were right handed men with ages between 25 and 37 (31.2 ± 5.3). Runs 1, 2, 3 and 4 of each participants are used to create a classification model and classification results are computed for runs 5, 6, 7 and 8.
246
Á. Costa et al.
Fig. 1 Averaged value of the attention coefficient provided by the classifier (in blue) during runs 5, 6, 7 and 8 and real attentional task performed by each participant (in red)
3 Results and Discussion On Fig. 1 the averaged coefficient obtained for each subject during runs 5, 6, 7 and 8 are shown. The attention coefficient provided by the classifier (in red) is compared to the real attention level (in blue) to appreciate how accurate is the real time system during exoskeleton rehabilitation. The attention coefficient provided by the system presents appreciable changes related to the attention level experienced by the participant. On this regard, high attention level is easily distinguished from low and standard attention level. However, between low and standard levels there are lower differences but, the use of the average of the 5 s buffer allows to fix a personalized threshold in the attention coefficient depending on the user to approximate the coefficient provided to the real attention level.
4 Conclusions A real time BMI to decode the attention level during gait has been designed. The system has been validated with four healthy users during the simulation of a lower limb rehabilitation strategy assisted by and ankle exoskeleton that supports dominant leg motion during walking. The exoskeleton provides the assistance to perform a natural gait step. The real time BMI system provides the attention level that the subject experiences during rehabilitation. In future works, this system will be used to modified the exoskeleton assistance. Acknowledgments The authors would like to thank Marta Moltedo, Tomislav Bacek and Dirk Lefeber for their support in the work on the MACCEPA actuator used in these experiments. We also thank Maria del Carmen Sánchez , Guillermo Asín-Prieto, José González and Juan Camilo Moreno for the mechanical design and fitting of the ankle exoskeleton.
Attention Level Measurement During Exoskeleton …
247
References 1. Johnston, M.V.: Plasticity in the developing brain: implications for rehabilitation. Dev. Disabil. Res. Rev. 15, 94–101 (2009) 2. Ramos-Murguialday, A., Broetz, D., Rea, M., Läer, L., Yilmaz, Ö., Brasil, F.L., Liberati, G., Curado, M.R., Garcia-Cossio, E., Vyziotis, A., et al.: Brain-machine interface in chronic stroke rehabilitation: a controlled study. Ann. Neurol. 74, 100–108 (2013) 3. Costa, A., Iáñez, E., Úbeda, A., Planelles, D., Hortal, E., Azorín, J.M.: Frequency and number of neighbors study for attention level classification using eeg signals. In: International Workshop on Wearable Robotics (WeRob 2014) (2014) 4. Zhang, D., Maye, A., Gao, X., Hong, B., Engel, A.K., Gao, S.: An independent brain-computer interface using covert non-spatial visual selective attention. J. Neural Eng. 7, 016010 (2010) 5. Kim, D.W., Hwang, H.J., Lim, J.H., Lee, Y.H., Jung, K.Y., Im, C.H.: Classification of selective attention to auditory stimuli: toward vision-free brain-computer interfacing. J. Neurosci. Methods 197, 180–185 (2011) 6. Costa, A., Salazar-Varas, R., Úbeda, A., Azorín, J.M.: Characterization of artifacts produced by gel displacement on non-invasive brain-machine interfaces during ambulation. Front. Neurosci. 10, 60 (2016) 7. Marple Jr, S.L.: Digital Spectral Analysis with Applications, vol. 1. Prentice-Hall, Inc., Englewood Cliffs, NJ, 512 p. (1987)
Detection of Subject’s Intention to Trigger Transitions Between Sit, Stand and Walk with a Lower Limb Exoskeleton Fernando Trincado-Alonso, Antonio J. del Ama-Espinosa, Guillermo Asín-Prieto, Elisa Piñuela-Martín, Soraya Pérez-Nombela, Ángel Gil-Agudo, José Luis Pons and Juan C. Moreno Abstract In this study we explore a way of controlling a lower limb exoskeleton based on the detection of the user intention by recording and classifying information from force sensors placed on both knees and hips. The classifier is based on Linear Discriminant Analysis and has been tested offline in 5 healthy subjects, obtaining an average accuracy of 91.11 % for the sit-to-stand transition, 72.5 % for the stand-to-walk transition and 70 % for the stand-to-sit transition.
This work is supported by The Biomot Project, funded by the Commission of the European Union under Grant Agreement number IFP7-ICT-2013-10-611695. F. Trincado-Alonso (&) A.J. del Ama-Espinosa E. Piñuela-Martín S. Pérez-Nombela Á. Gil-Agudo Biomechanics and Technical Aids Department, National Hospital for Spinal Cord Injury, Toledo, Spain e-mail:
[email protected] A.J. del Ama-Espinosa e-mail:
[email protected] E. Piñuela-Martín e-mail:
[email protected] S. Pérez-Nombela e-mail:
[email protected] Á. Gil-Agudo e-mail:
[email protected] G. Asín-Prieto J.L. Pons J.C. Moreno Neural Rehabilitation Group, Spanish National Research Council (CSIC), Madrid, Spain e-mail:
[email protected] J.L. Pons e-mail:
[email protected] J.C. Moreno e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_41
249
250
F. Trincado-Alonso et al.
1 Introduction Lower limb exoskeletons are robotic devices whose aim is to assist people with gait deficits during walking. During the last years, several studies have shown the feasibility of these systems in patients with spinal cord injury [1]. However, the effectiveness of such devices in terms of functional outcome remains unclear [2]. One of the challenges of the current systems is to improve the interaction between the robotic device and the human who is wearing it. Some of the state-of-the-art exoskeletons are controlled by a switch button, which is an easy way to control any system, but may not be useful when considering two aspects. Firstly, many of the potential users of lower limb exoskeletons require additional technical aids to walk, such as crutches or walkers. This hinders the use of a button at the same time that the user is walking. The second reason lies on the rehabilitative aim of the exoskeletons. Since the main objective of these systems is to provide a tool that facilitates training of gait, in order to maximize the benefits of such training, it should be as similar as possible to the normal gait. Ideally, the operation of the exoskeleton should be transparent to the user. With this objective, in this study we explore a way of controlling a lower limb exoskeleton based on the detection of the user’s intention by recording and classifying information from force sensors placed on both knees and hips. To this end, a classifier based on Linear Discriminant Analysis (LDA) was developed and tested offline in recording from 5 healthy subjects.
2 Materials and Methods 2.1
Material
The exoskeleton used to assist gait was the H2, a 6 degrees-of-freedom wearable lower-limb orthosis with anthropomorphic configuration [3]. It was equipped with three active joints for each leg: hip, knee and ankle, each of which powered by a DC motor coupled with a harmonic drive gear. Interaction torques are computed on each joint from the information that comes from the strain gauges placed on the exoskeleton. Additionally, 6 potentiometers (one on each joint) are used to calculate the angular position of the joints. In order to detect the intention of movement, a button was located on one of the crutches, in such a way that does not affect the position of the subject when performing the tasks. The experimenter triggered the exoskeleton’s transitions by using a mobile phone application that communicates via Bluetooth with the controller [3]. Data from the sensors are sampled at a 100 Hz rate, and synchronized with the signals of the 2 trigger signals: the one from the button and the other one from the mobile application, controlled by the subject and by the experimenter, respectively. Predefined trajectories, obtained from healthy subjects, were used as the desired gait pattern for the sit-to-stand, walking and stand-to-sit movements.
Detection of Subject’s Intention to Trigger Transitions …
2.2
251
Protocol
After wearing the exoskeleton, five healthy male subjects (average age 30 ± 6.4 years) were instructed to perform the following sequence of transitions with the aid of two crutches, beginning seated on a chair: stand up, start to walk, stop and sit down again. Subjects were asked to push the button right at the moment that they attempt to perform such a transition; namely, try to stand up when they are seated, try to start to walk when they are standing and try to sit down when they have finished to walk. 1 s after the subjects pushed the button to show their intention to move, the experimenter triggered the desired transition of the exoskeleton by pressing the corresponding button of the mobile application. Each subject repeated the aforementioned sequence of transitions 4 times.
2.3
Classifier
From the experiment described in the previous section, we obtained, for each of the 3 transitions (sit-to-stand, stand-to-walk and stand-to-sit), 4 windows of one-second of length with sensors data of the subject’s intention to trigger such transitions. Note that walk-to-stand transition is out of the scope of this study because the exoskeleton was configured to perform a predefined number of steps and then to stop, therefore the stopping is not controlled by the subject. A classifier was developed with the objective of distinguishing between rest period (when the subject stays in the same position) and transition period (when the subject wants to perform a transition). Three features were selected to this end: right knee torque, left knee torque and right hip torque. Rest period was considered between −2 and −1 s (where 0 is considered the beginning of the movement attempt). Transition period was considered between 0 and 1 s. Features for both periods were calculated as the difference of the absolute values of the average of the torque in the corresponding 1 s window and in the immediately prior window, according to the following formulas: Ftransition ¼
Frest ¼
1s 0s 1 X 1 X jTðtÞj jTðtÞj 100 t¼0s 100 t¼1s
1s 2s 1 X 1 X jTðtÞj jTðtÞj 100 t¼2s 100 t¼3s
ð1Þ
ð2Þ
Note that T is the torque and 1 s represents 100 samples. A different LDA classifier was trained and tested for each transition. The classifier for each subject (S1, for example) was always trained with data of the other 4
252 Table 1 Accuracy of classification for 5 healthy subjects
F. Trincado-Alonso et al. Transition
Accuracy (%)
Sit to stand Stand to walk Stand to sit
91.11 % 72.5 % 70 %
subjects (S2, S3, S4 and S5) and tested with his own data (S1). Then, using a 4th order cross-validation method, the average accuracy was computed for each transition.
3 Results The classifier has been tested offline in 5 healthy subjects, obtaining an average accuracy of 91.11 % for the sit-to-stand transition, 72.5 % for the stand-to-walk transition and 70 % for the stand-to-sit transition (Table 1). Interestingly, the classifier for each subject has been trained only with data of the other 4 subjects.
4 Discussion Improving the interaction between the subject and the exoskeleton is essential for the success of these devices in rehabilitation. Users should control these systems in a natural way, as similar as possible to the normal gait. A correct detection and interpretation of the user’s intention of movement is crucial to this end. A previous work correctly detected and classified (90 % of accuracy) different phases of gait and stand with an exoskeleton, using information from inertial sensors [4]. Another study used joint position, velocity and ground reaction forces to the same end, discriminating between 5 states: sit, stand, walking, sit-to-stand and stand-to-sit, obtaining an accuracy over 90 % in all cases [5]. The main contribution of our work with respect to the previous ones is that we developed a classifier for each subject using information from other subjects, whereas in the aforementioned studies a classifier was tailored to a single subject. The use of a common classifier for every subject allows minimizing the preparation time, which is a crucial matter in a rehabilitation procedure. Moreover, the correct detection of the subject’s intention enables a more natural way of controlling lower limb exoskeletons, which could facilitates the effectiveness for rehabilitative purposes. Besides, our work shows that 3 features could be sufficient to obtain a classifier with an acceptable accuracy and, at the same time, with a low computational load, which is important to get fast decision times. In the same direction, each classifier was trained to discriminate between rest and movement intention for each transition, but not between the 3 different movement intentions (sit to stand, walk and stand to sit). In the future online validation, the classifier will be complemented with a state machine, to apply
Detection of Subject’s Intention to Trigger Transitions …
253
for each transition the corresponding classifier, thus minimizing the computational load. Further analyses will be needed to validate the application to patients, which is the final goal of this development, since the interaction torques could be different than the ones of the healthy subjects.
5 Conclusions This work shows that it is possible to determine a common pattern of subject’s intention of moving an exoskeleton, although further analysis will be required to validate online these preliminary results.
References 1. Del-Ama, A.J., Gil-Agudo, A., Pons, J.L., Moreno, J.C.: Hybrid gait training with an overground robot for people with incomplete spinal cord injury: a pilot study. Front. Hum. Neurosci. 8, 298 (2014) 2. Federici, S., Meloni, F., Bracalenti, M., De Filippis, M.L.: The effectiveness of powered, active lower limb exoskeletons in neurorehabilitation: a systematic review. NeuroRehabilitation 37, 321–340 (2015) 3. Bortole, M., Venkatakrishnan, A., Zhu, F., Moreno, J.C., Francisco, G.E., Pons, J.L., Contreras-Vidal, J.L.: The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study. J. Neuroeng. Rehabil. 12(1), 54 (2015) 4. Wang, M., Wu, X., Liu, D., Wang, C., Zhang, T., Wang, P.: A human motion prediction algorithm for non-binding lower extremity exoskeleton. In: IEEE International Conference on Information and Automation, pp. 369–374 (2015) 5. Shen, B., Li, J., Bai, F., Chew, C.: Motion intent recognition for control of a lower extremity assistive device ( LEAD), pp. 926–931 (2013)
The New Generation of Compliant Actuators for Use in Controllable Bio-Inspired Wearable Robots T. Bacek, M. Moltedo, J. Gonzalez-Vargas, G. Asin Prieto, M.C. Sanchez-Villamañan, J.C. Moreno and D. Lefeber
Abstract In this paper, a conceptual design of the two iterations of compliant actuators used within BioMot project, as well as the control strategy used to operate these actuators, is presented. The result of the presented approach are 2 exoskeleton gait prototypes that will be used for incomplete spinal cord injury (iSCI) patients’ gait rehabilitation.
1 Introduction Human locomotion is a product of a complex interplay between the neural and the musculoskeletal systems. One of the outcomes of this interplay is the flexibility of the real-time human adaptability when confronted with both task and environmental constraints. However, despite their indisputable benefits, wearable robots (WRs) are T. Bacek (B) · M. Moltedo · D. Lefeber Department of Mechanical Engineering, R&MM Research Group, Vrije Universiteit Brussel (VUB), 1050 Brussel, Belgium e-mail:
[email protected] M. Moltedo e-mail:
[email protected] D. Lefeber e-mail:
[email protected] J. Gonzalez-Vargas · G.A. Prieto · M.C. Sanchez-Villamañan · J.C. Moreno Bioengineering Group, Consejo Superior de Investigaciones Cientficas (CSIC), 28500 Madrid, Spain e-mail:
[email protected] G.A. Prieto e-mail:
[email protected] M.C. Sanchez-Villamañan e-mail:
[email protected] J.C. Moreno e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_42
255
256
T. Bacek et al.
presenting constraints for human wearers due to the lack of a proper human-robot (HR) interface and robot’s versatility. The main goal of the BioMot project [1] is to explore and to exploit dynamic sensory-motor interplays to achieve seamless interaction and safe locomotion adjusted to the user’s intentions and capabilities. A milestone in the process of moving WRs into an unconstrained daily life environment is to develop and incorporate an adjustable compliance that adapts the human body to the desired dynamic motions. Compliance, important in human locomotion [2], has several advantages over stiff actuation. These include energy efficiency, shock tolerance, ability to explore natural dynamics and safety. Despite the benefits of the pneumatic muscles and Series Elastic Actuators, it was by introducing the Variable Impedance Actuators (VIAs) that the embodiment of the characteristics found in biological systems became possible in the new generation mechatronic systems. Within the BioMot project, MACCEPA [3] is used due to its simplicity, compactness and favourable output characteristic. Apart from this series elasticity principle, parallel elasticity principle is also exploited in the BioMot actuator design, leading to a further energy savings while not compromising versatility, flexibility and adaptability of the exoskeleton device. However, passive compliance itself is not sufficient to ensure safe actuation and it should be complemented by the appropriate low and high level control strategies. Hence, this paper presents conceptual design of the two iterations of VIAs used within BioMot project and the actuators’ accompanying low-level control strategies.
2 Compliant Actuation in WRs for Gait A. Series elasticity MACCEPA is a torque-controlled, variable stiffness actuator that works as a torsion spring that allows independent control of equilibrium position and joint stiffness. It has been recently used to build a modular compliant actuator successfully employed for powering all the joints of a sit-to-stance exoskeleton [4]. In order to improve the performance of these actuators, which suffered several drawbacks, first BioMot exoskeleton prototype introduced a highly efficient spindle-driven MACCEPA, leading to a better distribution of the actuator’s inertia and a more compact design. The conceptual design of the actuators in all the joints of the first BioMot exoskeleton is the same (see Fig. 1), but due to different kinematic/kinetic requirements, their realization is different. The knee actuator exploits knee gait kinematics/kinetics by employing a spring working in parallel to the joint (see next section). The shape of the actuators is guided by an anthropomorphic design approach, leading to a revolute joint-actuators providing a torque in the sagittal plane. All the actuators can provide the same peak torque equal to 25Nm. This value comes from a capability gap analysis conducted within the BioMot project (the capability gap is the difference between the torque provided by a healthy subject and the one provided by an impaired one).
The New Generation of Compliant Actuators for Use in Controllable Bio-Inspired …
257
Although MACCEPA can change spring pre-compression online by using a second motor, only the equilibrium position motor is implemented in the BioMot actuators due to a wearability requirement. The pre-compression can be changed manually by means of a screw and a clamped Kevlar® strap. The second BioMot exoskeleton is a multidimensional upgrade of the first version. Being built for assisting gait, the first exoskeleton’s actuators have a limited ROM for e.g. sit-to-stand, which is thus increased in the second prototype. By putting the spring inside the lever arm, an important novelty, the actuators are made smaller and more compact, leading to an improved wearability and human-robot synergy [5]. Further decrease in weight is obtained by performing a Finite Element Method analysis to remove the unnecessary material while preserving actuator’s structural integrity. Being more compact, the new actuators are flipped at the ankle and the knee joint (e.g. Link1, Fig. 1, is now aligned with the thigh and Link2 with the shank; the opposite is true in the first actuator iteration), leading to a bigger variability in the exoskeleton size. The second iteration of the actuators is also able to provide 50Nm peak at all the joints, which is achieved by a stronger actuator structure, higher gear ratio and a Dyneema® rope. Spring pre-compression is now easily changeable by means of a worm gear and a simple key. B. Parallel elasticity During gait, kinematics and kinetics of the knee joint is such that two clearly different gait phases can be observed: stiff weight-acceptance phase and compliant swing phase [6]. In order to mimic this behaviour more accurately, and to exploit passive energy storing and releasing of the former phase, the spring is added in parallel to the knee joint (Fig. 1) and its existing series elasticity actuation (MACCEPA). This spring, which is a part of an energy efficient quasi-passive on/off mechanism, will be controlled using a feedback position control based on a current gait phase.
Fig. 1 Schematic design of the spindle-driven MACCEPA of the first (red spring only) and the second (blue spring only) BioMot exoskeleton. The main difference is in the placement of the MACCEPA spring. The spring in parallel (in green) is implemented only in the knee actuator. Torque of the actuator depends on its geometry, spring pre-compression and torque angle α. Detailed description of the MACCEPA parameters can be found in [6]
258
T. Bacek et al.
3 Control Strategy The control strategy proposed for the BioMot exoskeleton aims at exploiting the energy storage and compliance properties of the MACCEPA. For example, at the ankle joint the state machine (SM) depicted in Fig. 2 is being used, which aims at increasing the power during the push-off phase of the gait, while keeping to a minimum the HR interaction during the rest of the gait. Hence, the way the motor is actuated changes between three states separated by the specific events of the gait. The first state (swing phase) maintains the interaction torque between the patient and the actuator as close as possible to zero using a zero-impedance control strategy. The second state (dorsiflexion) is triggered by the heel-strike event. In this state, the energy is stored into the spring by a natural forward shank motion. If the patient’s forward movement is not sufficient to generate enough of force to compress the spring (i.e. store the energy), a zero-impedance control as described earlier is used. The third state (push-off) is triggered by an event that marks the start of the heel rise. The vertical ground reaction force profile during walking exhibits two noticeable peaks. The local minimum between these two peaks corresponds to the heel rise event [7]. At this point, the control changes to a feedforward strategy that introduces extra energy to the push-off movement. The main advantage of using this strategy with the MACCEPA is the simultaneous contribution of the energy stored in the dorsiflexion and motor activation, thus improving the efficiency of the energy transfer. Once the toe-off is detected (i.e. the swing phase starts), the state machine goes back to the zero-impedance control.
Fig. 2 SM and control strategies proposed for the BioMot ankle actuator. S1:The minimized interaction torque is approximated by measuring the torque angle α, which is used as a feedback in a PID-based controller; S2:A PID-based position control keeps the lever arm position fixed, thus storing the energy into the spring. Fixed position depends on the patient’s capabilities; S3:The energy from the motor is a result of a fixed step of a current to the motor. Motor current also depends on the patient’s needs
The New Generation of Compliant Actuators for Use in Controllable Bio-Inspired …
259
4 Conclusion Two design approaches of the BioMot exoskeleton that takes advantages of the MACCEPA concept have been shown. Furthermore, the control used at the ankle joint is presented as an example of the control strategies that are being used for this exoskeleton. Tests with able subjects, as well as iSCI patients are currently being carried out to test the mechanical design and the control strategies. Acknowledgments This work is supported by the EC’s 7th Framework Program as part of the project BioMot (Grant Agreement number IFP7-ICT-2013-10-611695).
References 1. BioMot Project: Smart Wearable Robots with Bioinspired Sensory-Motor Skills. http://www. biomotproject.eu 2. Alexander, R.: Three uses of springs in legged locomotion. Int. J. Robot. Res. (Special Issue on Legged Locomotion) 9(2), 53–61 (1990) 3. Van Ham, R., et al.: MACCEPA, the mechanically adjustable compliance and controllable equilibrium position actuator: design and implementation in a biped robot. Robot. Auton. Syst. 55(10), 761–768 (2007) 4. Junius, K., et al.: Mechatronic design of a sit-to-stance exoskeleton. In: 5th IEEE RAS & EMBS International Conference on Biomedical Robotics and Biomechatronics, Sao Paulo, Brazil, pp. 945–950 (2014) 5. Moltedo, M., et al.: Mechanical design of a lightweight compliant and adaptable active ankle foot orthosis. In: 6th IEEE International Conference on Biomedical Robotics and Biomechatronics 2016, submitted for publication 6. Bacek, T., et.al.: Conceptual design of a novel variable stiffness actuator for use in lower limb exoskeletons. In: IEEE International Conference on Rehabilitation Robotics, Singapore, pp. 583–588 (2015) 7. Perry, J., Burnfield, J.M.: Gait analysis: normal and pathological function. J. Sports Sci. Med. 9(2), 353 (2010)
An EMG-informed Model to Evaluate Assistance of the Biomot Compliant Ankle Actuator Elena Ceseracciu, Luca Tagliapietra, Juan C. Moreno, Guillermo Asin, Antonio J. del-Ama, Soraya Pérez, Elisa Piñuela, Ángel Gil and Monica Reggiani
Abstract A main concern that rises when developing active orthoses is how to actively engage the users and monitor how they are affected by the devices. Through EMG-informed neuromusculoskeletal modeling, it is possible to estimate users’ muscle contributions to joint moments generation. We present preliminary results about the application of such models to a subject wearing the BioMot ankle actuator.
1 Introduction The successful development of novel wearable robotic devices for physical assistance of human locomotion requires that they be treated not as independent devices, but as actors in an ecosystem [1], where the environment and especially the user play crucial roles. One of the solutions adopted in the BioMot project is to sense users’ intention and estimate their contribution to motion generation through the use of electromyography(EMG)-informed models. In the present work, the applicability of these models and related technology to the BioMot Ankle Actuator device has been investigated on a healthy subject during treadmill walking.
E. Ceseracciu · L. Tagliapietra · M. Reggiani (B) Department of Management and Engineering, University of Padua, Vicenza, Italy e-mail:
[email protected] E. Ceseracciu e-mail:
[email protected] J.C. Moreno · G. Asin Neural Rehabilitation Group, Cajal Institute, Spanish National Research Council, Madrid, Spain A.J. del-Ama · S. Pérez · E. Piñuela · Á. Gil Biomechanics and Technical Aids Department, National Hospital for Paraplegics, Toledo, Spain © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_43
261
262
E. Ceseracciu et al.
2 Materials and Methods A. Biomot Ankle Actuator The BioMot Ankle Actuator (BAA) is an orthotic ankle device comprising a Mechanically Adjustable Compliance and Controllable Equilibrium Position Actuator (MACCEPA) [2]. To reduce cumbersomeness, the compliance of this actuator can only be changed manually, thus avoiding the need for a second motor on the unit. A flat MACCEPA design is used to better distribute the actuator’s inertia. Parameters and components were determined from simulation studies, targeting a user muscle weakness of 30 %; peak torque that the actuator can provide is 25 Nm. This actuator is compliant, modular, thus able to be used as a part of an exoskeleton or just as active ankle-foot orthosis, and bidirectional, therefore able to assist the patients during every phase of the gait cycle. For the present study, the actuator was controlled in transparent mode: the controller aims at compensating the presence of the device, without interfering with the user motion nor providing additional assistance to the user. Only the left ankle actuator was actually controlled, but the subject also wore the left knee actuator, to provide a measurement for the knee angle that is needed to determine the state of biarticular muscles, and the right side actuators, to maintain a symmetric gait (Fig. 1). B. CEINMS Neuromusculoskeletal (NMS) models reproduce the process with which muscles transform neural commands into movement, and can be used to estimate the forces generated inside the human body. Through the use of surface electromyography, we can indeed estimate the neural commands that are used as inputs for the NMS models: the linear envelope is computed from EMG data, then normalized to maximum voluntary contraction (MVC) value for each muscle. A non linear filter is then employed to model the muscle twitch response, obtaining activation signals that are input to Hill-type muscle models. These consist of an active force generating component (the muscle fibers) in series with a passive one (the tendon). Parameters for
Fig. 1 User wearing the BioMot actuators and Cometa EMG sensors
An EMG-informed Model to Evaluate Assistance …
263
the non linear filter and for the Hill-type models (most relevant are optimal fiber length, tendon slack length, maximum isometric force) are initially estimated from literature or measured from medical images, and then calibrated to ensure that resulting joint moments are consistent with the ones experimentally determined by means of inverse dynamics analysis of reference motions. The NMS models we employ are implemented through the Calibrated EMG-Informed NMS Modelling Toolbox (CEINMS) [3], available at https://simtk.org/home/ceinms. C. Experimental session One male healthy subject (age 30, height 174 cm, weight 65 kg) was recruited for this study, upon giving his informed consent. Active Codamotion (Charnwood Dynamics Ltd., UK) markers were placed on the subject by a physiotherapist (PT), according to Codamotion’s Bilateral Segmental Gait Analysis protocol. The PT also instrumented the subject’s leg with Cometa Wave Wireless EMG (2000 Hz; Cometa srl, Italy) electrodes, on the following muscles: tibialis anterior, medial and lateral gastrocnemius, soleus, peroneus longus. Two force platforms (Kistler AG, Switzerland), placed in the gait laboratory walkway, were used to collect ground reaction force data. A static acquisition of the subject in a standing posture was first recorded. The subject then performed at least 5 walking trials at self selected speed, hitting a force platform with the instrumented leg. MVC trials were performed, to determine normalization values for all subsequent EMG data processing. A generic musculoskeletal model was scaled to match the subject anthropometry measured with the static acquisition. The scaled model was used to run OpenSim analyses (inverse kinematics, inverse dynamics, muscle analysis) on the walking trials. Joint moments estimated through inverse dynamics were used as target for CEINMS model calibration. Muscle kinematics information (muscle-tendon lengths and moment arms) instead constituted the inputs for CEINMS, together with processed EMG data. A calibrated NMS model of the subject was thus obtained, which was used to estimate muscle forces and joint moments generated by the user while walking with the BAA. For the latter condition, the user performed continuous walking on a treadmill, at slow speed for safety reasons. As for CEINMS inputs, EMG data were acquired with the Cometa system and processed online, while joint angles measured by the BAA encoders were assumed to be consistent with the user joint angles. All systems and software were synchronized and connected within a ROS framework as in [4].
3 Results Subject’s ankle moment estimated by CEINMS during the walking trial with the BAA is shown in Fig. 2, together with ankle torque data recorded by the BAA.
264
E. Ceseracciu et al.
Fig. 2 Ankle moments exerted by the BAA (blue line) and by the user, as estimated by CEINMS (red line) during a walking session on a treadmill. Dorsiflexor moment is positive, plantaflexor is negative. Vertical lines show gait events for some of the steps (heel strike: solid lines, toe off: dashed lines), delimiting stance (ST) and swing (SW) phases
4 Discussion A quantitative assessment of BAA influence on user ankle moments was not possible in the current study because of the difference in walking speed, which affects the walking pattern of the user. For example, lower plantarflexor and dorsiflexor peaks when the subject wears the BAA might be attributed to the reduced speed. Nonetheless, the effect of the actuator controller can be clearly seen during the swing phase, where the dorsiflexor moment counteracts gravity, while user plantarflexor moment occurs during stance phase. While not conclusive, these data will be used as yardstick to evaluate the performance of the BAA when providing different assistance levels.
5 Conclusions An EMG-informed NMS model was successfully created and employed to monitor the state and the involvement of a healthy user wearing the BioMot Ankle Actuator. This model will be employed to assess the influence of different control parameters on the performance of the BAA, and possibly to integrate information on the state of the user within a novel control strategy. Acknowledgments This research has been supported by EU-F7 grant BioMot (project no. 611695).
An EMG-informed Model to Evaluate Assistance …
265
References 1. Tucker, M.R., et al.: Control strategies for active lower extremity prosthetics and orthotics: a review. J. NeuroEngineering Rehabil. 12(1) (2015) 2. Van Ham, R., Vanderborght, B., Van Damme, M., Verrelst, B., Lefeber, D.: MACCEPA, the mechanically adjustable compliance and controllable equilibrium position actuator: design and implementation in a biped robot. Robot. Auton. Syst. 55(10), 8–761 (2007) 3. Pizzolato, C., et al.: CEINMS: a toolbox to investigate the influence of different neural control solutions on the prediction of muscle excitation and joint moments during dynamic motor tasks. J. Biomech. 48(14), 39293936 (2015) 4. Ceseracciu, E. et al.: A flexible architecture to enhance wearable robots: integration of EMGinformed models. In: International Conference on Intelligent Robots and Systems (IROS 2015)
Tacit Adaptability of a Mechanically Adjustable Compliance and Controllable Equilibrium Position Actuator, a Preliminary Study Guillermo Asín-Prieto, Shingo Shimoda, José González, M. Carmen Sánchez-Villamañán, José L. Pons and Juan C. Moreno Abstract Current powered exoskeleton (exo) control algorithms for locomotion assistance and rehabilitation are based on assistive, resistive and error augmentation paradigms. Within the assistive controller’s family, assist-as-needed consists in applying a corrective force proportional to the error (actual limb position versus reference pattern). Our final goal is to implement a fully adaptable control mechanism to allow a full lower limb exo to dynamically adapt the gait pattern to each patient. We propose to use a modified version of tacit learning algorithm in combination with a variable stiffness actuator to explore the improvement of the adaptability in comparison to stiff actuators. The preliminary results show that using this concept on a compliant actuator it is possible to modulate a fixed trajectory to adapt to the position limits that are induced by user’s movement capabilities.
1 Introduction Current powered exoskeleton (exo) control algorithms for locomotion assistance and rehabilitation are based on assistive, resistive and error augmentation paradigms. Within the assistive controller’s family, assist-as-needed (AAN) consists in applying a corrective force proportional to the error from the actual position of the limb with respect to a reference pattern (eg. the physiological gait) [1]. This corrective force can be modulated to stimulate the active participation of the user to generate movement. In the literature, AAN has been applied to the control of stiff actuation exos, which gives certain ability to adapt to the patient’s movement capacity [2, 3]. Recently, AAN has been proposed as a control strategy for exos with compliant actuators, such as in the LOPES exo [4]. However, the adaptability of AAN depends on a manually fixed value that specifies the magnitude of the corrective force. Therefore, it is difficult G. Asín-Prieto (B) · J. González · M. Sánchez-Villamañán · C. Pons · J. Moreno Neural Rehabilitation Group, Spanish National Research Council, Madrid, Spain e-mail:
[email protected] S. Shimoda Intelligent Behaviour Control Unit, RIKEN Brain Science Institute, Wako, Japan © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_44
267
268
G. Asín-Prieto et al.
for the exo controller to automatically adapt to the changing characteristics of each patient’s gait capability. In the framework of the European Project BioMot, our goal is to obtain a behavioural flexibility that enhances the patient’s skill acquisition process. In previous work we have successfully used the tacit learning (TL) algorithm to improve the adaptability of a stiff actuated exo [5]. However, due to the rigidity of the exo, this adaptation was limited. Therefore, in this study, we modified the TL algorithm to use it with a variable stiffness actuator (VSA). We called this modification of the algorithm tacit adaptability (TA) since it allows an automatic adaptation of the exo control in real time. We implemented and tested the control strategy with the knee joint of an exo, using a position controller, with the aim of demonstrating that the approach could be moved to a real rehabilitative scenario. Thus, the exo would exert a fixed gait pattern that could be automatically adapted.
2 Materials and Methods A. Variable Stiffness Actuator The actuator we used is based on the intrinsically compliant MACCEPA (mechanically adjustable compliance and controllable equilibrium position actuator) concept [8]. It is based on a compliant torque-controlled rotational actuator with adjustable compliance. The motor is connected rigidly to the lever arm (LA), which is connected to the fixed link (FL) with a spring, as shown in Fig. 1. LA and FL move with respect to the output link (OL). We have used the actuator of the knee, so in a real scenario OL should be attached to the upper leg, while FL should be attached to the shank. In this way, if the motor is set to a specific position, the user is still able to rotate the knee by compressing the spring. This allows to approximate the interaction force between the user and the actuator by measuring the resulting length of the spring (α angle).
Fig. 1 Model and real MACCEPA actuator. Lever arm is rigidly moved by the motor, while fixed link should be attached to the limb of the user. A spring (K) attaches LA to FL. The spring compression is proportional to the interaction force
Tacit Adaptability of a Mechanically Adjustable Compliance and Controllable …
269
Fig. 2 PID control scheme integrating a TA and a reference-tracking PID controller. α represents the deflection between LA and FL
B. Tacit adaptability control TA algorithm is shown in (Fig. 2). This control strategy automatically adapts a reference trajectory to the user’s movement capabilities. In this way, the output of a position PID controller is modulated by a integral gain we call TA constant (TAc). Therefore, if the user is not able of realizing the trajectory, the output of the control adapts proportionally to the interaction torque between the exo and the user. This adaptation is important, since it will automatically reduce the force applied to the user. The response time of the adaptation can be modulated by changing TAc. C. Experimental setup The system was powered with a 24 volts/300 watts power supply. A magnetic encoder measures α, what allows us to infer the interaction between the user and actuator through the spring compression. An absolute optical encoder measures the angle on the user’s joint. The data is acquired in a PC with a PCI CAN card, running Simulink. We recorded 60 seconds per condition sampled at 200 Hz. We fixed the links of the actuator to a rigid structure, and allow the motor to move LA by compressing the spring. The goal with this setup was to explore the adaptability of the controller to predefined reference trajectories. We propose a condition to simulate the situation in which a user is not able to generate any limb movement. In our setup, the actuator was limited to the horizontal plane to avoid the effects of gravity. In this scenario, the mechanical compliance, given by the spring constant (K), absorbs some of the force that the motor applies to the user’s limb. Since the limb cannot move as it is fixed in one position, TA adjusts the reference trajectory, to further reduce the applied force. TA controller was set to follow a 60 degrees sinusoidal trajectory (range of motion of the knee actuator) at a frequency of 0.2 Hz. The range for α is determined by the spring stiffness. The actuator was fixed at three different joint angles (15, 30 and 45 degrees). We collected the LA and the controller output for 5 different TAc (0, 5, 10, 15 and 20).
3 Results Figure 3 shows the evolution of α with different values of TAc. When TAc = 0 the behaviour is a pure PID control, thus no adaptation on LA position trend can be seen. When the TAc takes values greater than zero, TA acts, and the sinusoidal signal is modulated in a way that the neutral position of LA is aligned with the position of FL.
270
G. Asín-Prieto et al.
Fig. 3 Comparison of α angle amplitude for TAc values of 0, 10 and 20 during the experiment. α angle amplitude is represented with a slim line; the thick line shows the trend, calculated as the centered moving average (3000 points window over the 12000 points signal)
4 Conclusion The preliminary results show that using the TA concept on a compliant actuator it is possible to modulate a fixed trajectory reference to adapt to the position limits that are induced by user’s movement capabilities. We tested the control strategy in a setting that simulates the extreme case where the user is not capable of realizing any movement. When no TA is used, we observed that the control output results in a stiffer behaviour of the actuator. On the other hand, a high TAc results in a more compliant behaviour of the actuator. Thus, the adaptation is given by the compliance of the exo as well as by TA. These results point out at the possibility to automatically adapt a fixed reference trajectory to the movement capabilities of the user by modulating the TAc.
5 Futute Work We plan to extend this study to consider other input variables to the TA block, and emulating non-extreme conditions, e.g. a study in dynamic conditions with partial torque contribution. A study involving healthy subjects wearing the full exo and applying this TA approach is to be designed, also exploring the approach for gait cadence adaptability. Compare these results with other AAN-based controllers that use trajectory references, e.g. virtual tunnel with the H2 [6] or hybrid (FES and exo) actuation with Kinesis [7]. Acknowledgments This study has been funded by grant from the European Commission, within the Seventh Framework Programme (IFP7-ICT-2013-10-611695: BioMot - Smart Wearable Robots with Bioinspired Sensory-Motor Skills).
Tacit Adaptability of a Mechanically Adjustable Compliance and Controllable …
271
References 1. Riener, R., Nef, T., Colombo, G.: Robot-aided neurorehabilitation of the upper extremities. Medical and Biological Engineering and Computing, vol. 43(1), pp. 2–10. Springer (2015) 2. del Ama, A.J., Moreno, J.C., Gil-Agudo, Á., Pons, J.L.: Hybrid FES-robot cooperative control of ambulatory gait rehabilitation exoskeleton for spinal cord injured users. In: Converging Clinical and Engineering Research on Neurorehabilitation, pp. 155–159. Springer (2013) 3. Moreno, J.C., Collantes, I., Asín-Prieto, G., Pons, J.L.: Design of better robotic tools adapted to stroke rehabilitation practice. World Cong. Med. Phys. Biomed. Eng. (2012) 4. Lagoda, C., Schou, A.C., Stienen, A.H.A., Hekman, E.E.G., and van der Kooij, H.: Herman: Design of an electric series elastic actuated joint for robotic gait rehabilitation training. In: 2010 3rd IEEE RAS and EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob), pp. 21–26 (2010) 5. Shimoda, S., Costa, Á., Asín-Prieto, G., Okajima, S., Iáñez, E., Hasegawa, Y., Azorín, J.M., Pons, J.L., Moreno, J.C.: Joint stiffness tuning of exoskeleton robot H2 by tacit learning. In: Symbiotic Interaction, pp. 138–144. Springer 6. Bortole, M.: Design and control of a robotic exoskeleton form gait rehabilitation. Ph.D. thesis, Carlos III University, Madrid, Spain (2013) 7. Del-Ama, A.J., Gil-Agudo, Á., Pons, J.L., Moreno, J.C.: Hybrid gait training with an overground robot for people with incomplete spinal cord injury: a pilot study. Front. Hum. Neurosci. 8, 298–298 (2013) 8. Bacek, T., Unal, R., Moltedo, M., Junius, K., Cuypers, H., Vanderborght, B., Lefeber, D.: Conceptual design of a novel variable stiffness actuator for use in lower limb exoskeletons. In: IEEE International Conference on Rehabilitation Robotics (ICORR), vol. 2015, pp. 583–588 (2015)
Part X
Emerging Applications Domains of WRs, Emerging Technologies in WRs
Design and Kinematic Analysis of the Hanyang Exoskeleton Assistive Robot (HEXAR) for Human Synchronized Motion Wansoo Kim, Hojun Kim, Donghwan Lim, Hyungi Moon and Changsoo Han
Abstract The Hanyang Exoskeleton Assistive Robot (HEXAR)-CR50, which performs human synchronized gait motions to augment human power during load carrying, was designed and analyzed in this study. The HEXAR-CR50 was developed for industrial and military purposes to carry a payload of 20–30 kg while walking on level ground and climbing stairs. For the design of the exoskeleton robot with considered the joint functions and motion, we conducted a gait analysis that was based on biomechanics. The parameters for the design were based on the results of the gait analysis. The designed exoskeleton consisted of one leg with seven degrees of freedom (DOF). The simulations were conducted to verify the kinematic synchronized motion using LifeMODTM .
1 Introduction Researchers have studied the lower extremity exoskeleton systems with the purpose of augmenting the physical capability of wearers to walk long distances or to carry heavy objects easily. Depending on the design structure of the alignment to the rotation axis of the robot joint and the rotation axis of the human joint, an exoskeleton may be classified as anthropomorphic, quasi-anthropomorphic, or non-anthropomorphic [1]. For example, exoskeleton systems were developed for power augmentation has been designed quasi-anthropomorphically to carry heavy loads [2, 3]. The Massachusetts Institute of Technology (MIT) designed an W. Kim Human-Robot Interfaces and Physical Interaction (HRI2), Istituto Italiano di Tecnologia (IIT), Genoa, Italy e-mail:
[email protected] H. Kim · D. Lim · H. Moon Mechanical Engineering, Hanyang University, Seoul, South Korea C. Han (B) Department of Robot Engineering, Hanyang University ERICA, Ansan, South Korea e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_45
275
276
W. Kim et al.
anthropomorphic exoskeleton system with spring and variable dampers [4]. However, walking assist device (WAD) has been designed non-anthropomorphically [5]. The purpose of this study was to design the Hanyang Exoskeleton Assistive Robot (HEXAR)-CR50, a lower extremity exoskeleton that is used for power augmentation, and to provide HEXAR-CR50 with human-synchronized gait motion. Therefore, the degree of freedom (DOF) and joint types that allow for human motions and joint functions had to be considered. Simulations were conducted to verify the effectiveness of the design for synchronized motion with a human using LifeMODTM .
2 System Requirements This study focused on the load carrier, which performs the anti-weight function of the load while walking on level ground. The functions of exoskeleton structure in the stance phase is only required to support the load bearing along the gravitational axis, which is perpendicular to the ground. For the purpose of the exoskeleton design, the peak values were analyzed for the angle, torque, and power during human motion, which represented the hip, knee, and ankle joints in the sagittal plane, because the vast majority of the gait is done in the sagittal plane [6]. The peak values can be used to select the following three elements in the design phase: (1) the range of motion (ROM) were considered through the peak value of the joint angle; (2) the joint peak moment were considered in selecting the actuator specifications; and (3) the joint actuating methods were selected through the analysis of the peak power of the joint. The exoskeletons power of the joints must be generated using the active joints. However, it is difficult to implement all the DOF in anthropomorphic exoskeleton robots due to the space limitations caused by structural characteristics. Thus, a quasi-anthropomorphic architecture was selected for efficiency and for synchronized motion to achieve the load carrying goal. Our previous study provided more details about the design concept of the HEXAR-CR50 [7].
3 Mechanical Design of HEXAR-CR50 The HEXAR-CR50 has seven DOF per leg: three for the hip motions, one for the knee motions, and three for the ankle motions (Fig. 1). The active mechanism is applied to the hip and knee joints for extension/flexion movements; the ankle and other side-joint modules use a quasi-passive mechanism. The active mechanism uses a flat brush-less direct current motor (BLDC) motor, which can produce extension/flexion motions in the sagittal plane, with a Harmonic Drive. The specifications of the electrical BLDC motors and the harmonic gear that were selected using biomechanical data. We used a 207 W BLDC motor (Kollmorgen® ) with a 100:1 gear ratio (Harmonic Drive LLC). The quasi-passive mechanism that can generates a force by applying spring elements; it is applied to the
Design and Kinematic Analysis of the Hanyang Exoskeleton Assistive Robot …
277
Fig. 1 The active, quasi-passive, and passive joint arrangement of the HEXAR-CR50 showing the ROM.
dorsi-plantarflexion motion of the ankle joint to compensate for the torque generated in the stance phase and to generate a propulsion force in the toe-off phase. Consequently, the spring constant was determined to be 69.6 N/mm, which resulted from the conversion of the results of the gait analysis to the weight of the exoskeleton [7].
4 Kinematic Simulation with Walking Motions Simulations were conducted to verify the efficiency of the HEXAR-CR50 design in terms of joint arrangement, ROM, and shape. Two assumptions were considered in order to analyze the coordinated motions between the human and the exoskeleton in the simulations. The first assumption was that the contact locations between the human and exoskeleton were the torso, thigh, and foot. The second assumption was that the exoskeleton motions were synchronized completely with the human motions through the wearer. Utilizing LifeMOD software, we also constructed simulation environments to check the human and exoskeleton motions. To verify the ROM and the arrangement of exoskeleton joints attached to the human, the normal gait cycle by the human and the exoskeleton were compared in a 3D spaces. Because, normal gait cycles are represented well in our target motion. Based on the torso link, the human and exoskeleton trajectories of each joints were represented in the frontal, transversal, and sagittal planes to compare their motion ranges. Similar trajectories indicated that the designed exoskeleton could be imitated the human motions well. In the sagittal plane, since the hip and knee joints of the exoskeleton had the same axis as the human joints, no significant difference was shown during the operation. However, in the case of the ankle joint, the dorsi/plantarflexion joint was designed to be close to the ground to represent the ROM, as shown in Fig. 2a. Figure 2b shows the motions in the frontal plane. As the figure shows, the joint trajectories of the human and the exoskeleton were expressed as offsets because the exoskeleton was attached
278
W. Kim et al.
Fig. 2 The trajectories were generated by the all the joints, which as human and exoskeleton, that moved according to the training through LifeMOD
(a)
(b)
(c)
externally, although it was verified that the motion patterns were similar. Figure 2c shows in the transversal plane, in which the motions in the trajectories were expressed with offsets. In particular, the results shown in this Fig. 2c verified that the motions in the internal/external rotation mechanism of the hip joint represented human motions very well. According to the simulation results, also implemented human motion well.
5 Conclusion Through this study, a lower-extremity exoskeleton called HEXAR-CR50 was designed for power augmentation during load carrying. The design specifications of the exoskeleton were based on the results from biomechanical considerations and gait analysis. Based on the these specifications, we have selected the ROM, joint types, and arrangement of the exoskeleton for design procedure. In the design, each leg had seven-DOF and was configured with a combination of active, passive and quasi-passive joints. The designed exoskeleton was verified the design efficiency with respect to the ROM, DOF and structure layout, through the LifeMOD simulations.
Design and Kinematic Analysis of the Hanyang Exoskeleton Assistive Robot …
279
Acknowledgments This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government(MSIP) (No.NRF-2015R1A2A2A01002887)
References 1. Lee, H., Kim, W., Han, J., Han, C.: The technical trend of the exoskeleton robot system for human power assistance. Int. J. Precis. Eng. Manuf. 13(8), 1491–1497 (2012) 2. Kazerooni, H., Steger, R.: The Berkeley lower extremity exoskeleton. J. Dyn. Syst. Meas. Control 128(1), 14 (2006) 3. Bogue, R.: Exoskeletons and robotic prosthetics: a review of recent developments. Ind. Robot. Int. J. 36(5), 421–427 (2009) 4. Walsh, C.J., Endo, K., Herr, H.: A quasi-passive leg exoskeleton for load-carrying augmentation. Int. J. Humanoid Robot. 4(3), 487–506 (2007) 5. Ikeuchi, Y., Ashihara, J., Hiki, Y., Kudoh, H., Noda, T.: Walking assist device with bodyweight support system. In: IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2009, pp. 4073–4079. IEEE (2009) 6. Palmer, M.L.: Sagittal plane characterization of normal human ankle function across a range of walking gait speeds. Master’s thesis, Massachusetts Institute of Technology (2002) 7. Kim, W., Lee, H., Kim, D., Han, J., Han, C.L.: Mechanical design of the Hanyang Exoskeleton Assistive Robot (HEXAR). In: International Conference on Control, Automation, and Systems (ICCAS 2014), pp. 479–484 (2014)
Design and Experimental Evaluation of a Low-Cost Robotic Orthosis for Gait Assistance in Subjects with Spinal Cord Injury Josep M. Font-Llagunes, Daniel Clos, Urbano Lugrís, F. Javier Alonso and Javier Cuadrado Abstract Robotic gait training after spinal cord injury (SCI) is of high priority to maximize independence and improve the living conditions of these patients. Current rehabilitation robots are expensive and heavy, and are generally found only in the clinic. To overcome these issues, we present the design of a low-cost, low-weight robotic orthosis for subjects with SCI. The paper also presents a preliminary experimental evaluation of the assistive device on a subject with SCI. Results show that gait velocity, stride length and cadence of walking increased (24.11, 7.41 and 15.56 %, respectively) when wearing active orthoses compared to the case with standard passive orthoses.
1 Introduction SPINAL cord injury (SCI) is prevalent in society. Worldwide each year more than 250.000 individuals suffer SCI [1]. Walking impairment after injury leads to a decreased quality of life and other serious health conditions, and carries substantial health care costs. Locomotor rehabilitation is reported as a high priority issue for subjects with SCI independent of severity, time after injury, and age [2].
This work has been supported by the Spanish Ministry of Economy and Competitiveness under the project DPI2015-65959-C3-2-R, co-funded by the European Union through ERDF funds. J.M. Font-Llagunes (&) D. Clos Department of Mechanical Engineering and Biomedical Engineering Research Centre of the Technical University of Catalonia, Barcelona, Spain e-mail:
[email protected] U. Lugrís J. Cuadrado Laboratory of Mechanical Engineering of the University of La Coruña, Barcelona, Spain F.J. Alonso Department of Mechanical Energetics and Materials Engineering of the University of Extremadura, Barcelona, Spain © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_46
281
282
J.M. Font-Llagunes et al.
Current gait rehabilitation robots are machines that support the patient’s weight and train the walking motion over a treadmill or feet supports; or lower limb exoskeletons that assist over-ground walking. These robots are generally heavy and expensive, and are only found in the clinic because skilled personnel have to manually fit the robot to the patient and operate it. Moreover, they are adapted to the patient before the treatment, increasing time and health care costs; and, in general, they impose a motion pattern rather than complementing patient’s capabilities. With the aim of overcoming these limitations, this work presents the design, control and experimental evaluation of a low-cost, low-weight and simple robotic orthosis for gait assistance in subjects with SCI. The prototype is intended for patients that can control hip flexion/extension to a certain extent, but lack control of knee and ankle muscles. The design is based on the current passive knee-ankle-foot orthoses that these patients use after rehabilitation. The latter include a knee locking system, which is essential to bear the patient’s weight during stance due to the lack of quadriceps force; and a compliant system that applies a dorsiflexion torque at the ankle to avoid drop-foot gait (klenzak joint). The robotic orthosis presented here improves the passive devices by adding a motor at the knee, that can move or lock the joint, and an inertial measurement unit (IMU) at the shank to detect gait events. The aim of this work is two-fold: first, we present the design of the robotic orthosis; and second, we perform a preliminary experimental evaluation on a subject with SCI. In this case study, the kinematics of walking with passive orthoses is compared with that obtained with the robotic device.
2 Robotic Orthosis Design The proposed lower limb orthosis has two degrees of freedom. The knee joint is powered by an electrical motor in series with a Harmonic Drive gearbox. The ankle is passively actuated by a mechanism that applies the mentioned dorsiflexion torque (klenzak joint). A preliminary design of the orthosis was reported in [3]. The current device weights 2.7 kg per leg, along with a 1.7 kg backpack containing a BeagleBone Black board, the motor drivers and the battery. The bilateral thigh and shank uprights are articulated at the knee, using a standard hinge joint at the medial side and the motor-gearbox set at the lateral side. A footplate with a shoe is hinged to the shank uprights by the compliant klenzak joint, which allows an ankle range of motion between 0º and 20º (dorsiflexion). The orthosis structure is specifically tailored to the patient to avoid adapting the same design to the wide range of morphologies found among subjects with SCI. Figure 1a shows the right robotic orthosis with the elements described later.
Design and Experimental Evaluation of a Low-Cost Robotic …
283
Fig. 1 Robotic orthosis design: a general view showing the knee actuation system and the IMU; b CAD design of the actuation system
2.1
Knee Actuation System
The design and selection of the orthosis actuation system were based on kinematic and kinetic data of the knee joint during normal gait at a normal speed [4]. The most significant criteria for the actuation system selection were specific power (power to weight ratio), system dimensions, and portability of the power supply system. Based on these considerations, a 70 W brushless DC motor (Maxon Motor, Sachseln, Switzerland) was selected, which has a nominal voltage of 24 V and a nominal torque of 128 mNm. A Harmonic Drive gearbox (Harmonic Drive, LimburgLahn, Germany) is coupled to the motor to increase torque and reduce velocity, which offers a large gear ratio with a reduced space (Fig. 1b). The selected gear ratio of 160:1 allows a continuous net torque at the knee of 20.5 Nm and peak torques of 60 Nm according to the driver current limit.
2.2
Sensors and Control
All the sensors are placed on the orthosis mechanical structure in order to avoid issues related to safety, comfort, reliability and donning/doffing process. The sensors used are one IMU and one angular encoder per orthosis. The low-cost 9-DOF IMU (SparkFun Electronics, Niwot, USA) is attached to the shank upright; and incorporates a triple-axis gyro, a triple-axis accelerometer and a triple-axis magnetometer. The orientation and acceleration measurements are sent to the BeagleBone board through a serial interface. The angular encoder is coupled to the knee motor.
284
J.M. Font-Llagunes et al.
The control algorithm uses both IMUs measurements to detect the stance-to-swing transition within the gait cycle. During stance, the knee is fully extended and the motor acts as a brake. When the stance-to-swing transition is detected, based on vertical acceleration and pitch angle of both shanks, the knee motor launches a fixed flexion-extension cycle using a PID position controller with feedforward. This cycle is personalized to the subject in terms of duration, shape and maximum flexion angle.
3 Experimental Evaluation The subject was an adult female 41 years old, mass 65 kg and height 1.52 m; with SCI at T11. In the first experiment, she walked with her usual pair of passive knee-ankle-foot orthoses with the help of two parallel bars. Then, the subject carried out 6 one-hour training sessions wearing the active orthoses and did some specific exercises at home to facilitate adaptation. After this period, a second experiment walking with the active orthoses, also with the help of parallel bars, was performed (Fig. 2a). In order to compare the walking kinematics in the two experiments, 4 consecutive gait cycles were captured each time by 6 optical infrared cameras (Natural Point, Corvallis, USA) that measured the position of 37 optical markers. Then, a computational 3D skeletal model with 18 anatomical segments and 57 degrees of freedom was used to determine the kinematic characteristics of the subject’s gait (Fig. 2b). Table 1 shows kinematic descriptors for one gait cycle during the first experiment and another gait cycle during the second experiment. Gait velocity, stride length and cadence of walking increased (24.11, 7.41 and 15.56 %, respectively)
Fig. 2 Gait of spinal cord injured subject assisted by active orthoses and parallel bars: a acquired motion; b computational model
Design and Experimental Evaluation of a Low-Cost Robotic … Table 1 Kinematic data with passive and active orthoses Gait velocity (m/s) Stride length (m) Cadence (step/min) COM lat. displ. (cm)
285
Ex.1: passive
Ex.2: active
% change
0.17 0.53 38.46
0.21 0.57 44.44
+24.11 +7.41 +15.56
7.89
6.37
–19.31
when wearing active orthoses compared to the case with passive orthoses. Furthermore, the lateral displacement of the subject’s centre of mass (COM) decreased in 19.31 % when the subject walked with active orthoses.
4 Conclusion This paper presents the design and control of a patient-tailored low-cost knee-ankle-foot robotic orthosis for subjects with SCI. This orthosis is equipped with a compact knee actuation system and an IMU at the shank to detect gait events. Preliminary experimental evaluation of this assistive device on a subject with SCI shows that the subject walked faster, and in a more natural and stable way when wearing the designed active orthoses. While the experiments provided promising results, more tests with a larger sample of subjects are needed in order to confirm the improvements when walking with the designed orthosis.
References 1. Bickenbach, J., Bodine, C., Brown, D., Burns, A., Campbell, R., Cardenas, D., et al.: International Perspectives on Spinal Cord Injury. World Health Organization (WHO), Geneva (2013) 2. Ditunno, P.L., Patrick, M., Stineman, M., Ditunno, J.F.: Who wants to walk? Preferences for recovery after SCI: a longitudinal and cross-sectional study. Spinal Cord 46(7), 500–506 (2008) 3. Font-Llagunes, J.M., Lugrís, U., Romero, F., Clos, D., Alonso, F.J., Cuadrado, J.: Design of a patient-tailored active knee-ankle-foot orthosis to assist the gait of spinal cord injured subjects. In: Proceedings of the International Workshop on Wearable Robotics, Baiona, Spain, 2014, paper 54 4. Bovi, G., Rabuffetti, M., Mazzoleni, P., Ferrarin, M.: A multiple-task gait analysis approach: kinematic, kinetic and EMG reference data for healthy young and adult subjects. Gait Posture 33(1), 6–13 (2011)
A Powered Low-Back Exoskeleton for Industrial Handling: Considerations on Controls Stefano Toxiri, Jesús Ortiz, Jawad Masood, Jorge Fernández, Luis A. Mateos and Darwin G. Caldwell
Abstract A powered low-back exoskeleton is being developed to support manual material handling in industry. Controlling this device poses several challenges. At the low-level, the actuation units need to be capable of large torque outputs as well as transparent interaction. At the high-level, the exoskeleton needs to modulate its assistance based on information acquired from the environment and the wearer, so as to maximise its beneficial effect. These challenges have great relevance to industrial applications, where complexity, cost and invasiveness are key to successful deployment. In describing the current progress in the development of the exoskeleton, an attempt is made to highlight and discuss these challenges and possible technical solutions.
1 Introduction In industrial manufacturing processes, manual material handling is one of the most frequent operations. There has been increasing interest in researching the use of wearable assistive devices to reduce the risk of resulting musculoskeletal disorders. S. Toxiri (B) · J. Ortiz · J. Masood · J. Fernández · L.A. Mateos · D.G. Caldwell Advanced Robotics Department, Istituto Italiano di Tecnologia, Via Morego, 30, 16163 Genoa, Italy e-mail:
[email protected] J. Ortiz e-mail:
[email protected] J. Masood e-mail:
[email protected] J. Fernández e-mail:
[email protected] L.A. Mateos e-mail:
[email protected] D.G. Caldwell e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_47
287
288
S. Toxiri et al.
The review presented in [1] summarises existing industry-oriented devices and points at evidence of reduced resulting physical loads on wearers. In [2], we further supported the proposition that the spinal loads are reduced. Building upon previous experience, we report here the progress towards the development of a wearable powered low-back exoskeleton aimed at reducing spinal loads during manual material handling (as part of the Robo-Mate consortium [3]). The present manuscript offers a brief description of the current prototype. The focus is on the controls of wearable robots, and the description is complemented with general considerations relevant to the issues we encountered while developing our prototype. In the following sections, we discuss two aspects separately: the low-level control of the actuators and the high-level assistive strategy.
2 Low-Back Exoskeleton As anticipated, the goal is to reduce the spinal loads during manual material handling. Biomechanically, the compression on the lumbar spine is considered as a key risk factor for musculoskeletal injuries. In this respect, we envisioned a wearable device that would reduce the lumbar compression by reducing the need for muscular activity in the spine [2]. Requirements were extracted using a simple biomechanical model combined with human motion data. In order to provide substantial assistance, we therefore aim at applying torques on the upper body of around 100 Nm on the sagittal plane. As shown in Fig. 1, our prototype exoskeleton spans the lower back and upper legs and its weight rests mainly on the waist. The actuator torques are transferred to the wearer as assistive forces applied onto thighs and torso. The two actuation units
Fig. 1 User wearing the prototype. Different postures of a typical handling task are shown
A Powered Low-Back Exoskeleton for Industrial Handling …
289
are approximately aligned with the hip on both sides. A set of passive joints is aimed at allowing unhindered movements outside the sagittal plane.
3 Low-Level: Actuator Control A highly desirable feature of powered joints in wearable robots is transparent physical interaction. It can be explained as the device not hindering the wearer’s movement unless required, and finds important applications in motor rehabilitation as well as power augmentation. Some important considerations concerning the implementation of said transparent interaction are necessary. Reduction gears are often employed when relatively high torques are required, allowing the use of small and lightweight motors. However, this comes at a cost: achievable speeds are reduced, and output friction and inertia are greatly amplified. The result is a sacrificed control bandwidth, which may significantly limit performances when dealing with dynamic tasks. Force/torque sensors only partially mitigate the above restrictions, as they increase weight, cost and complexity [4]. Hence the motivation to research sensorless solutions, such as those proposed in [5] and [6]. In this respect, we are currently working towards a formal comparison of sensor-based versus sensorless methods to control our prototype actuation units. With reference to the restrictions introduced above, our prototype actuation units present limiting dynamic properties. Each unit includes a flat brushless DC motor (Maxon EC90) and a 160:1 HarmonicDrive. The motors are current-controlled via a commercial control board (Maxon ESCON). Unsurprisingly, the units exhibit very large output friction and inertia, which made for poor performances in transparent interaction in current-control mode. This led to the inclusion of a joint torque sensor, that allows far more accurate control of the torque output. As a result, the exoskeleton is capable of providing over 120 Nm total continuous torque with maximum speeds limited to around 2 rad/s, which still barely keeps up with the human movements during handling tasks. On the one hand, we are considering alternative motor/reduction combinations to allow for higher speeds and therefore more dynamic tasks. On the other hand, we are looking into leaving out the joint torque-sensor while retaining comparable torque control performance.
4 High-Level: Assistive Strategy At the high level, a strategy is needed for the modulation of the assistance with the aim of maximising the beneficial effect on the wearer. Extensive reviews on assistive strategies can be found in [7–9]. In general, one can think of indirect and direct control as the two cases in which the assistance is modulated using information from either the environment (e.g. ground reaction force) or the wearer (e.g. surface
290
S. Toxiri et al.
electromyography, sEMG), respectively. As considered in [7], acquiring direct volitional information from the wearer is especially important when noncyclic tasks are to be supported, as opposed to cyclic ones such as walking. The study in [10] presents a hybrid scheme that uses sEMG to assist the knee during sit-to-stand. However, direct volitional information requires relatively invasive acquisition techniques that may not be compatible with industrial settings, which probably explains the poor diffusion of EMG-based methods outside research laboratories. Our prototype embeds an inertial unit to measure the orientation of the upper torso. This indirect information is used to calculate a torque reference such that the effect of gravity acting on the wearer’s upper body is counteracted by the exoskeleton. Yet, no information on the weight of the external load being handled is available. We are therefore exploring the use of instrumented foot insoles to estimate the change in ground reaction forces as the user picks up and releases objects of different weights [11]. The weight of the external object will also be estimated via instrumented gloves, currently being developed by ZHAW, a Robo-Mate consortium partner [3].
5 Conclusions This contribution highlights and discusses important technical challenges related with controlling a wearable powered low-back exoskeleton. These challenges have great impact to translating robotics research into industrial applications, where complexity, cost and invasiveness are key to successful deployment. Acknowledgments The authors are grateful to the whole Robo-Mate project consortium, particularly to the team in ZHAW and their contribution to the controls. The research leading to these results has received funding from the European Unions Seventh Framework Programme for research, technological development and demonstration under grant agreement n 608979, and from the People Programme (Marie Curie Actions) of the European Unions Seventh Framework Programme FP7/2007-2013/ for research, technological development and demonstration under REA grant agreement n 608022.
References 1. de Looze, M.P., Bosch, T., Krause, F., Stadler, K.S., O’Sullivan, L.W.: Exoskeletons for industrial application and their potential effects on physical work load. Ergonomics 1–11 2. Toxiri, S., Ortiz, J., Masood, J., Fernández, J., Mateos, L.A., Caldwell, D.G.: A wearable device for reducing spinal loads during lifting tasks: biomechanics and design concepts. In: 2015 IEEE International Conference on Robotics and Biomimetics (ROBIO), pp. 2295–2300. IEEE (2015) 3. http://www.robo-mate.eu 4. Zanotto, D., Lenzi, T., Stegall, P., Agrawal, S.K.: Improving transparency of powered exoskeletons using force/torque sensors on the supporting cuffs. In: 2013 IEEE International Conference on Rehabilitation Robotics (ICORR), pp. 1–6. IEEE (2013)
A Powered Low-Back Exoskeleton for Industrial Handling …
291
5. Nef, T., Lum, P.: Improving backdrivability in geared rehabilitation robots. Med. Biol. Eng. Comput. 47(4), 441–447 (2009) 6. Ugurlu, B., Nishimura, M., Hyodo, K., Kawanishi, M., Narikiyo, T.: A framework for sensorless torque estimation and control in wearable exoskeletons. In: 2012 12th IEEE International Workshop on Advanced Motion Control (AMC), pp. 1–7. IEEE (2012) 7. Tucker, M.R., Olivier, J., Pagel, A., Bleuler, H., Bouri, M., Lambercy, O., del R Milln, J., Riener, R., Vallery, H., Gassert, R.: Control strategies for active lower extremity prosthetics and orthotics: a review. J. Neuroeng. Rehabil. 12(1), 1 (2015) 8. Yan, T., Cempini, M., Oddo, C.M., Vitiello, N.: Review of assistive strategies in powered lower-limb orthoses and exoskeletons. Robot. Auton. Syst., 64, 120–136 (2015) 9. Lobo-Prat, J., Kooren, P.N., Stienen, A.H., Herder, J.L., Koopman, B.F., Veltink, P.H.: Noninvasive control interfaces for intention detection in active movement-assistive devices. J. Neuroeng. Rehabil., 11(1), 1 (2015) 10. Karavas, N., Ajoudani, A., Tsagarakis, N., Saglia, J., Bicchi, A., Caldwell, D.: Tele-impedance based assistive control for a compliant knee exoskeleton. Robot. Auton. Syst. 73, 78–90 (2015) 11. Mateos, L.A., Ortiz, J., Toxiri, S., Fernández, J., Masood, J., Caldwell, D.G.: Exoshoe: a sensory system to measure foot pressure in industrial exoskeleton. In: 2016 IEEE International Conference on Biomedical Robotics and Mechatronics (BioRob) (accepted for)
Efficient Lower Limb Exoskeleton for Human Motion Assistance Nazim Mir-Nasiri
Abstract The main advantage of the fully autonomous system is its ability to decouple the weight/mass carrying function of the system from its forward motion function. It makes exoskeleton more efficient by reducing its power consumption, weight and size of the propulsion motors as well as by extending the run time of the batteries. A human machine interface has been achieved by means of flexible sensors to monitor subject shank and ankle movements and subject’s foot pressure. When subject stands on one leg and swings the other one the body weight is fully supported by standing leg exoskeleton structure where the knee joint motion is fully blocked. The join motors are small in size and consume less electrical energy from batteries because they do not support subject’s weight during the walk.
1 Introduction During the last decade, researchers have focused on the development of lower limb exoskeletons that are now applied to several fields, including power augmentation for the military [1], medical assistance [2], and rehabilitation [3, 4]. In such devices human provides control signals while the exoskeleton actuators provide required power for performing the task. On lower extremity exoskeletons, most previous researchers paid their attention in developing walking aid systems for gait disorder persons or aged people [5]. One of those systems is HAL (Hybrid Assistive Leg) developed by Yoshiyuki Sankai of University of Tsukuba was aimed at assisting human leg muscles during walking [6]. The system was based on electromyography (EMG) sensing of human muscles as the primary drive signals. The development resulted in several versions of HAL with the latest HAL-5 in 2009 [7]. The exoskeleton was motor powered on the hip and knee joints, leaving other joints free. The Berkeley Lower Extremity Exoskeleton (BLEEX) [8] was aimed at enhancing human strength and endurance for payload transport. The exoskeleton N. Mir-Nasiri (&) Nazarbayev University, Astana, Kazakhstan e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_48
293
294
N. Mir-Nasiri
incorporates hydraulic actuation on all three sagittal joints and two coronal joints on the hip with all others joints free. The ECUST Leg Exoskeleton Robot (ELEBOT) designed at East China University of Science and Technology (ECUST) shares the similar design goal as BLEEX but with a simplified system [9]. While the above exoskeleton designs require substantial power for operation on low efficiency, an exoskeleton design at Massachusetts Institute of Technology (MIT) [10] attempted to lower the power requirement for load carrying. The exoskeleton has only series elastic actuation at hip sagittal joints, variable damper at the knee joints and spring at ankle sagittal joints. Various problems remain to be solved, the most important remains being the creation of a power and cost efficient system that will allow an exoskeleton to operate for extended periods. Such lighter and cheaper devices are currently important engineering research area in medicine and military [11].
2 Mechanical Design and Components Figure 1 shows the conceptual sketch of the proposed exoskeleton structure in Solid Works. In the figures seat 1 is there to rest subject’s body and support its weight. Each exoskeleton leg has four degrees of freedom: two at the hip 2, one at the knee 3 and one at the ankle 4 to allow legs forward and lateral motions. Cushioned seat 1 in between subject crotch is connected to two parallel rigid pipes 5 at the back the object. A back panel 6 mounted onto the rigid pipes serves as a platform for control and power supply mounting. At hip level, the two parallel rigid pipes extend out to the two hip coronal joints. The link then continues to both sides of the hip 2 where sagittal and transverse joints are located, subsequently to the knee joints 3 and through the ankle joints 4 to the ground. Both exoskeleton legs are attached to subject legs using flexible cuffs 7. Single degree four-bar linkage mechanism 8 with rotary joints at the hip level provides hip-centered lateral rotation of the exoskeleton leg around vertical axis. The remaining three single-degree
Fig. 1 Concept of the exoskeleton structure
Efficient Lower Limb Exoskeleton for Human Motion Assistance
295
Fig. 2 Schematic diagrams of the exoskeleton
parallel axes rotary joints at the hip 2, knee 3 and ankle 4 provide freedom of flexion at the joints. Pneumatic cylinders 9 are used to block the motion at the knee joints 3 when necessary to support the weight. Figure 2 shows schematic diagrams of the exoskeleton. In these Fig. 1 are adjustable telescopic members of the exoskeleton; 2 are dummy pneumatic cylinders that are able to inhibit the motion at the knee joints; 3 and 4 are the sensors to detect motion of subject thigh and shank; 5 are springs to support feet 6 of the exoskeleton; 7 and 8 are flexible belts to fasten exoskeleton to the subject thigh and shank. In the figures M1 and M2 are motors driving the hip and knee joints of each leg; C1 and C2 are solenoid valves of the pneumatic cylinders 2 that are able to inhibit motions at the knee joints; S1 and S2 are flexible strips 3 and 4 with bonded strain gages that are able to sense the tiny motions of subject limbs; F1 and F2 are foot pressure sensors to sense the amount of pressure applied by the ground on the exoskeleton sole 6 during the walk. The pressure at the exoskeleton sole is generated due to the transmission of the weight forces via mechanical structure to the ground while the subject is resting on the seat.
3 Exoskeleton Operation The operation of all actuators, i.e. motors M1, M2 and cylinders solenoid valves C1 and C2 depends on the amount of pressure felt by the feet pressure sensors F1 and F2 (Fig. 2). The pressure on the feet depends on actual posture of the subject and can be classified in three distinct cases. In case 1 the subject is stands on one leg and reading from the corresponding leg foot sensors will be at its maximum value Pb whereas for other leg that will be zero. In case 2 the subject stands still on both legs then the total subject body weight is almost equally shared by both legs and the pressure reading will be Pb/2. In case 3 the subject is in the stage of transiting its weight from one leg to another. As a result pressure reading will go to zero for one leg and increase to maximum for another one. The operation of the cylinders that
296
N. Mir-Nasiri
can inhibit the knee joint motion and the motors based on the following rules of operation that depends on the actual knee angle Ɵ and foot pressure reading Pb: • Motors M1 and M2 of each leg are actuated only and only if Ɵ Ɵmin [AND] the pressure reading from sensors at the foot P Pmin in order to pick up the leg from the ground and take a step • Cylinders solenoid valves C1 and C2 are actuated only and only if Ɵ < Ɵmin. It means the cylinders are actuated when the knee has reached the minimum allowed value to bend. The value of the minimum bending angle Ɵmin depends on usual walk style of the subject The control strategy for the motors M1 and M2 is aimed to make sure that the exoskeleton structure will follow the subject’s leg physical motion without hindering it. The set of sensors S1 and S2 (flexible strips with bonded strain gages) are attached to links of the exoskeleton (Fig. 2). When the subject limbs commence the motion the limbs will touch and bend the strips. The sensors will detect in real time the intention of the subject to move limbs and send the signals to PID controller. The close-loop controller will react immediately by activating hip and knee motors M1 and M2 in order to move the links of the exoskeleton away from the object limbs and thus to restore the original shape of the strips.
4 Conclusion The paper describes the methodology of mechanical design and effective control of a new exoskeleton system to enhance walk capabilities of people. It also can be used for rehabilitation of people with light leg injuries. The core idea is to use exoskeleton to decouple weight carrying capabilities of the legs from its body advancing capabilities. This has been done by special logic and management of exoskeleton actuators. This exoskeleton is power efficient because electrical motors are smaller in size and did not participate in supporting the weight like in all other existing exoskeleton designs.
References 1. Zoss, A.B., Kazerooni, H., Chu, A.: Biomechanical design of the Berkeley lower extremity exoskeleton (BLEEX). IEEE/ASME Trans. Mechatron. 11(2), 128–138 (2006) 2. Suzuki, K., Mito, G., Kawamoto, H., Hasegawa, Y., Sankai, Y.: Intention-based walking support for paraplegia patients with robot suit HAL. Adv. Robot. 21, 1441–1469 (2007) 3. Jamwal, P.K., Sheng, Q.X., Shahid, H., John, G.P.: An adaptive wearable parallel robot for the treatment of ankle injuries. IEEE/ASME Trans. Mechatron. 19(1), 64–75 (2014) 4. Iqbal, J., Baizid, K.: Stroke rehabilitation using exoskeleton-based robotic exercisers Mini Review. Biomed. Res. 26(1), 197–201 (2015)
Efficient Lower Limb Exoskeleton for Human Motion Assistance
297
5. Banala, S., Agrawal, S., Scholz, J.: Robot assisted gait training with active leg exoskeleton (ALEX). IEEE Trans. Neural Syst. Rehabil. Eng. 17, 2–8 (2009) 6. Kawamoto, H., Sankai, Y.: Power assists method based on phase sequence and muscle force condition for HAL. J. Adv. Robot. 19(7), 717–734 (2005) 7. Kawabata, T., Satoh, H., Sankai, Y.: Working posture control of robot suit HAL for reducing structural stress. In: Proceedings of IEEE International Conference on Robotics and Biomimetics (ROBIO), pp. 2013–2018 (2009) 8. Ghan, J., Kazerooni, H.: System identification for the Berkeley lower extremity exoskeleton (BLEEX). In: Proceedings 2006 IEEE International Conference on Robotics and Automation (ICRA), pp. 3477–3484 (2006) 9. Heng, C., Jun, Z., Chunming, X., Hong, Z., Xiao, C., Yu, W.: Design and control of a hydraulic-actuated leg exoskeleton for load-carrying augmentation. In: Proceedings 2006 IEEE International Conference on Robotics and Automation (ICRA), Part I, LNAI 6424, pp. 590–599 (2010) 10. Walsh, C.J., Paluska, D.: Development of a lightweight, underactuated exoskeleton for load-carrying augmentation. In: Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pp. 3485–3491 (2006) 11. Leslie, M.: The next generation of exoskeletons. Mag. IEEE Eng. Med. Biol. Soc. 3(4), 56–61 (2012)
Active Safety Functions for Industrial Lower Body Exoskeletons: Concept and Assessment Jawad Masood, Luis A. Mateos, Jesús Ortiz, Stefano Toxiri, Leonard O’Sullivan and Darwin Caldwell
Abstract Exoskeleton technology can assist human effort performing manual handling tasks in industrial environments. Safety is vital both for the commercial and legal acceptance of industrial exoskeletons. We consider such systems as safety critical as they directly involve humans. Active safety functions can enhance the safety of exoskeletons and prevent accidents and injuries. We present the conceptual and evaluation criteria for Fall Detection, Active Balancing, Active Detachment, Collision Detection and Automatic Motor Braking. This evaluation highlights the safety critical scenarios which can be helpful in future hazard and risk assessment of exoskeletons, and also for safety certification evaluation.
1 Introduction Exoskeleton technology is a vibrant area of research due to its potential in force augmentation both for medical and industrial applications. For industrial handling applications, current EU research includes the Robo-Mate and Spexor projects. But to date, there has been limited commercial success of this type of technology in the J. Masood (B) · L.A. Mateos · J. Ortiz · S. Toxiri · D. Caldwell Istituto Italiano di Tecnologia, Via Morego, 30, 16163 Genoa, Italy e-mail:
[email protected] L.A. Mateos e-mail:
[email protected] J. Ortiz e-mail:
[email protected] S. Toxiri e-mail:
[email protected] D. Caldwell e-mail:
[email protected] L. O’Sullivan University of Limerick, Limerick, Ireland e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_49
299
300
J. Masood et al.
industry. Most industrial exoskeletons on the market to date have focused on passive devices e.g. Fortis. The current Robo-Mate exoskeleton forms the first step towards a real world, active, industrial handling exoskeleton. Robo-Mate is a wearable exoskeleton which is designed to work closely with the user. The system dynamics synchronize with the user natural biomechanics through information flow from sensors. The intended system objective is to provide assistance to the user during lowering, lifting and carrying tasks. This close interaction makes Robo-Mate a safety critical system which can lead the user to Musculo Skeletal Disorders (MSDs) [1]. The aim of Robo-Mate is not only to provide intended functionality of assistance but to enhance the safety of the worker. This ambitious objective has focused our interests on devising intelligent safety functions. We conceptualize four active safety function based on available hardware, software, and safety risk scenarios. The Risk Scenarios are evaluated through the Hazard Analysis and Risk Assessment (HARA) of the Robo-Mate system. These functions are Active Balancing, Fall Detection, Collision Detection and Automatic Motor Braking. But are these safety functions safe enough or do they introduce additional hazards and risks to the user and system? These interesting research questions are yet to be answered. HARA is a technique used in established standards to measure the performance of safety-critical systems. For example, in the automotive industry ISO-26262 provides the guidelines for HARA. Relevant standards for Industrial Exoskeleton Technologies are evolving quickly. The primary standard is ISO 13482-2014 on personal care robot safety. That standard mentions the use of “dummies” which infers instrumented manikins as used in automotive crash research and safety certification. Yet there is a paucity of instrumented manikins for low-velocity contact testing. There is no mentioning of software models in this standard. This paper is an important opportunity to provide safety information to support exoskeleton design and also certification used in an industrial context. In addition, these concepts can be beneficial for the development of other standards on the safety of industrial robots such as ISO 10218/1-2011 and ISO 10218/2-2011. The evaluation process and safety functions presented in this paper can contribute towards developing simulation methods and procedures that can help in exoskeleton safety verification at early design stages.
2 Active Safety Functional Concepts A. Active Balancing (1) Scenario: The safety critical scenario can be defined as the situation when the user is handling a load with varying center of gravity (e.g. fluid in a basin) or when due to some unforeseen condition the system is under the uneven distribution of load.
Active Safety Functions for Industrial Lower Body Exoskeletons … Table 1 Available posture states for Active Safety Functions State Weight Inclination (force sensor) (IMU)
301
Edge (Force sensor + IMU)
90◦
–
Lift up
Initial body weight (base weight) Increase
135◦ –90◦
Lift down
Decrease
90◦ –135◦
Walk Walk while carrying
Initial body weight Initial body weight + Lifted weight
90◦ –95◦ 90◦ –95◦
Force sensor Rising, IMU Falling Force sensor Falling, IMU Rising – –
Stand
(2) Risk: The scenario can risk the health and safety of the user. The uneven distribution of load can lead to dangerous posture and instability of the full system. (3) Definition: The exoskeleton detects the non-homogeneous distribution of external load and adapts the control algorithm accordingly. The information received from foot force sensors, Inertial Measurement Unit (IMU) and hand force sensors can be fused to detect the system unbalancing Table 1. The foot force sensors can detect the weight distribution on each foot before and after user handles the load. The hand force sensors can validate the presence of a load in the hand. The information is fused in the centrally embedded computer. During assistance mode, this information can help to update the torque control algorithm. In this way, the system can generate different torque profiles for left and the right actuator such that the center of the system can remain in sagittal plane. This information can be validated and controlled by IMU. B. Fall Detection (1) Scenario: The safety critical scenario can be defined as the situation when the user falls while wearing Robo-Mate. (2) Risk: The scenario can risk the health and safety of the user. The uneven distribution of load can lead to dangerous posture and instability of the full system. (3) Definition: The exoskeleton detects the user falling. The information received from the IMU in the form of vertical acceleration can be used to detect the fall. The foot force sensors can detect the weight distribution on each foot before and after detection of the fall. The hand force sensors can validate the presence of a load in the hand. The information can be fused in the centrally embedded computer. This function will bring the Robo-Mate into fall mode. This safety function can be enabled during all operational modes. C. Collision Detection (1) Scenario: The safety critical scenario can be defined as the situation when the user handling the load collides with an obstacle in its workspace.
302
J. Masood et al.
(2) Risk: The scenario can risk the health and safety of the user. The collision can lead to dangerous posture and instability of the full system. (3) Definition: The exoskeleton detects the collision by using sensor-less obstacle detection using hybrid torque-motion control strategy. The collision detection information is computed in the central embedded computer. A collision detection warning is generated in the Human Machine Interface (HMI). The user is asked to return to a safe configuration. If the collision warning is repeated three times in critical time (t), then the system enters into halt mode. D. Automatic Motor Braking (1) Scenario: The safety critical scenario can be defined as the situation when the user tries to move outside the system safe range of motion. (2) Risk: The scenario can risk the health and safety of the user. Moving outside the safe range of motion can lead to dangerous posture and instability of the full system. (3) Definition: The exoskeleton function can monitor the movement of the user. A warning is generated as the user moves close to the limits of safe range of motion. If the user moves beyond this range, the actuators provide an opposite torque (i.e. Automatic Motor Braking—AMB) to constraint the users motion beyond the safe range of motion. As the system returns back to the safe range, the actuator returns to the previous mode.
3 Hazard Analysis and Risk Assessment HARA of active safety functions is an important task during design and development of safety critical systems such as exoskeletons. The International Standard Organization (ISO) describes the systematic procedure for this task. For example, the ISO-26262 is the automotive functional safety standard which divides this task into three main sub-tasks. These sub-tasks are Severity of Failure, Controllability during Failure and Probability of Failure. The assessment of these sub-tasks is performed by dividing them into several levels and computing the mutual score. In practice, these assessments are based on intuitive knowledge of the production manager and engineers. This approach is error prone and raises questions on all safety assessment process. There are several tools used in practice [2]. They can calculate the Safety Integrity Levels (SIL) by considering severity, controllability, and probability of failure without explicitly supporting design decisions and analysis. We suggest to use the rigorous simulations to assess these risk [2–4] because it can help in quickly verifying the safety concepts at earlier design stages of exoskeleton.
Active Safety Functions for Industrial Lower Body Exoskeletons …
303
4 Conclusions We present the concepts and assessment for four active safety functions for industrial exoskeletons. The safety scenarios are defined and risks are identified. The evaluation and assessment of these safety functions can be performed by using rigorous simulations because of their capability to express explicit design decision and analysis. We propose to use the SIL to evaluate the functional performance similar to ISO 26262, which is based on severity, controllability, exposure and probability to failure. The presented work can be beneficial in developing the HARA for upcoming standards for industrial exoskeletons. Acknowledgments The authors acknowledge the contributions of the Robo-Mate consortium. The research funding is supported by the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no. 608979.
References 1. Van der Vorm, J., OSullivan, L., Nugent, R., de Looze, M.: Considerations for developing safety standards for industrial exoskeletons (2015) 2. Duracz, A., Bartha, F.A., Aljarbouh, A., Masood, J., Philippsen, R., Eriksson, H., Duracz, J., Xu, F., Zeng, Y., Taha, W., Grante, C.: Using rigorous simulation to support hazard analysis and risk assessment (HARA) in the ISO 26262 functional safety standard. In: ACM Transactions on Embedded Computing Systems (TECS) (2016) 3. Duracz, A., Eriksson, H., Bartha, F.A., Xu, F., Zeng, Y., Taha, W.: Using rigorous simulation to support ISO 26262 hazard analysis and risk assessment. In: IEEE 12th International Conference on Embedded Software and Systems (ICESS) (2015) 4. Masood, J., Philippsen, R., Duracz, J., Taha, W., Eriksson, H., Grante, C.: Domain analysis for standardised functional safety: a case study on design-time verification of automatic emergency breaking. In: The International Federation of Automotive Engineering Societie Congress, FISITA (2014)
SOLEUS: Ankle Foot Orthosis for Space Countermeasure with Immersive Virtual Reality Pierre Letier, Guillaume Fau, Uwe Mittag, Jochen Zange, Joern Rittweger, Moonki Jung, Jo McIntyre and Arnaud Runge
Abstract This paper presents the SOLEUS project, which aims to design innovative countermeasure for astronauts in space, based on the development of an active foot orthosis and immersive virtual reality technologies. The paper introduces the problematics of space countermeasure and describes the expected benefits of the proposed technology. It provides details on the system architecture, components and the simulation tool that has been used to support the design process. Finally, the scientific evaluation strategy for the validation of the system is introduced.
1 Introduction During spaceflight, the human body is subject to physiological adaptation to the microgravity environment. Although some effects, like vestibular disorders, lead to temporary discomfort for some days, other reactions such as bone mineral loss or muscle atrophy affect the physical condition of astronauts in proportion to the time spent in space. These are major concerns for long-duration missions such as those P. Letier (&) G. Fau Space Applications Services, Zaventem, Belgium e-mail:
[email protected] U. Mittag J. Zange J. Rittweger Institute of Aerospace Medicine, German Aerospace Center DLR, Cologne, Germany J. Rittweger Department of Paediatrics and Adolescent Medicine, University of Cologne, Cologne, Germany M. Jung Anybody Technology A/S, Aalborg, Denmark J. McIntyre Tecnalia, Derio, Spain A. Runge TEC-MMG, European Space Agency, Noordwijk, The Netherlands © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_50
305
306
P. Letier et al.
on-board the ISS (several months) and for future planetary exploration missions to the Moon and Mars (several years). Increasing time exposure of astronauts to microgravity for new missions requires a more thorough understanding of issues met by astronauts in microgravity. Current countermeasures are not effective enough, as they only partially mitigate deconditioning effects. The development of new or significantly enhanced countermeasures is of paramount importance. The on-going ESA project SOLEUS aims to develop a new approach of integrated countermeasure device in the shape of a lower-leg boot-exoskeleton associated with 3D head mounted display. It focuses primarily on the neuromotor and mechanical stimulation of the lower leg body segments that are the most heavily affected body parts while astronauts are exposed to microgravity. This approach is supported by immersive Virtual Reality (VR) technologies aiming at providing additional stimulation and information to the user, in order to increase the countermeasure efficiency. This paper presents the status of the project, describing the expected benefits of the device and the system design, supported by the use of musculo-skeletal simulation tools.
2 SOLEUS Project Expected Benefits The SOLEUS project is mainly characterized by the association of active ankle orthosis for each leg and an immersive virtual reality display. Compared to existing approaches for space countermeasure currently used on-board the ISS, the following benefits are expected: • Activation of the muscle-tendon unit (MTU) of the lower legs as a whole, with a focus on the spring-damper behavior and stiffness characteristics of the MTU [1], that is not well addressed by the current use of the treadmill or body press [2]. Optimization of energy storage in the MTU has a direct impact on the longitudinal bone force loading during specific motion (jumping, running). • The use of the VR to create scenarios addressing not only the pure muscle atrophy but also the full functional task including the neurologic transmissions [3]. The system can be used to stimulate different neurological pathways (e.g. balance, locomotor, reflex, patterns generators) to increase the number of motor units activated (MU, association of a neuron and muscle fiber). That can also include tricks and cheats to increase the effect or compensate for the absence of other solicitations (e.g. feel of the gravity vector). • Increased motivation by using the VR for displaying challenging and recreative adapted scenarios. Experiments have suggested that outcomes are better when practice is task oriented, repetitive and adapted to the user [4].
SOLEUS: Ankle Foot Orthosis for Space Countermeasure …
307
Fig. 1 SOLEUS system architecture overview
3 SOLEUS System Architecture Figure 1 represents the SOLEUS system architecture overview that is mainly composed of the exoskeletons and VR sub-systems.
3.1
Exoskeletons Subsystem
The Exoskeleton subsystem is responsible to measure the motion and interaction forces of the user and implements active feedback based on the scenarios exercises. It is composed of two portable lower-leg exoskeletons enabling active control of the ankle joint in the flexion/extension and pronation/supination motions. Each exoskeleton features a mechanical structure, comfortable mechanical fixations to the user, two brushless DC actuators, absolute position and force sensors, and Synapticon boards for low-level control, power amplification, sensors conditioning and communication purposes. The boards are interconnected by a power bus and EtherCAT data bus to the Exoskeleton Controller that is running Orocos for high-level real-time mechatronics control [5]. Each orthosis can provide up to 130 Nm of equivalent torque around the ankle thanks to the use of an offset slider-crack mechanism.
3.2
Virtual Reality Subsystem
The Virtual Reality subsystem is responsible to simulate physics of the exercises scenarios and to create the 3D world environment in which the user will be
308
P. Letier et al.
Fig. 2 Muscular-skeletal simulation of a human ankle wearing the SOLEUS orthosis under load scenario, output of the muscular response under activation of the SOLEUS orthosis
immersed. It is composed of a computing unit running Unity3D [6] and the eVRS framework that allows standard physicians to create and manage easily 3D medical scenarios, and to interface many different hardware components to be used during the experiments. The system allows connecting a 3D head mounted display (Oculus Rift) to render the 3D world computed by Unity and to track user’s head motion. Both subsystems are connected through ROS to exchange measured and control data. The system can also be interfaced with medical instrumentation (e.g. EMG sensors) to measure muscle activity during operations (for medical evaluation). The SOLEUS system is currently under integration for further tests to be carried out during a Scientific Evaluation campaign (Sect. 5).
4 Musculo-Skeletal Simulations The design process has been supported by the AnyBody Technology inverse dynamics software [7]. Based on a realistic model of the lower-leg human muscular-skeletal system and a model of the mechatronics design, we demonstrated the ability of the SOLEUS orthosis to generate the proper muscular response from typical user’s scenarios (Fig. 2). The simulation has also been used to derive the mechanical loads for finite element analyses and mechanical design optimizations.
5 Scientific Evaluation Once integrated, SOLEUS will be tested during a short scientific evaluation with the purpose, to validate the main functions of the system, focusing on the good synchronization between the mechatronics and virtual reality. We will in addition perform preliminary countermeasure tests through different scenarios to assess the expected benefits while analyzing risks of using such technology (e.g. risks of
SOLEUS: Ankle Foot Orthosis for Space Countermeasure …
309
fracture). Beside standard activities like isometric, isotonic and isokinetic stimulations, more advanced scenarios with full 3D immersion of the subject will be tested such as “body jumping” and “balancing”. The goal is to obtain evidences of the activation of the spring-damper behavior of the MTU, as well as increased EMG amplitude signal by the use of the VR rendering.
6 Conclusion This paper presented the SOLEUS system as a wearable technology for the emerging applications of countermeasure for astronauts in space. This technology has also potential for other applications such as rehabilitation and healthcare. Acknowledgments This study if funded by ESA in the framework of a Technology Research program (contract No. 4000112181/14/NL/RA) entitled “Integrated Countermeasures with Biofeedback and Actuators”.
References 1. Farris, D., Robertson, B., Sawicki, G.: Elastic ankle exoskeletons reduce soleus muscle force but not work in human hopping. J. Appl. Physiol. 115, 579–585 (2013) 2. Alkner, B.A., Tesch, P.A.: Efficacy of a gravity-independent resistance exercise device as a countermeasure to muscle atrophy during 29-day bed rest. Acta Physiol. Scand. 181, 345–357 (2004) 3. Chopard, A., Hillock, S., Jasmin, B.J.: Molecular events and signalling pathways involved in skeletal muscle disuse-induced atrophy and the impact of countermeasures. J. Cell Mol. Med. 13, 3032–3050 (2009) 4. Levin, M., et al.: Virtual reality environments to enhance upper limb functional recovery in patients with hemiparesis. In: Advanced Technologies in Rehabilitation (2009) 5. Orocos: http://www.orocos.org/ 6. Unity3D: https://unity3d.com/ 7. Damsgaard, M., Rasmussen, J., Christensen, S.T., Surma, E., De Zee, M.: Analysis of musculoskeletal systems in the AnyBody Modeling System. Simul. Model. Pract. Theory 14, 1100–1111 (2006)
SPEXOR: Spinal Exoskeletal Robot for Low Back Pain Prevention and Vocational Reintegration Jan Babiˇc, Katja Mombaur, Dirk Lefeber, Jaap van Dieën, Bernhard Graimann, Michael Russold, Nejc Šarabon and Han Houdijk
Abstract Most assistive robotic devices are exoskeletons which assist or augment the motion of the limbs and neglect the role of the spinal column in transferring load from the upper body and arms to the legs. In the SPEXOR project we will fill this gap and design a novel spinal exoskeleton to prevent low-back pain in able bodied workers and to support workers with low-back pain in vocational rehabilitation.
J. Babiˇc (B) Department of Automation, Biocybernetics and Robotics, Jožef Stefan Institute, Ljubljana, Slovenia e-mail:
[email protected] K. Mombaur Department of Optimization in Robotics & Biomechanics, Heidelberg University, Heidelberg, Germany D. Lefeber Robotics & Multibody Mechanics Research Group, Department of Mechanical Engineering, Vrije Universiteit Brussel, Ixelles, Belgium J. van Dieën MOVE Research Institute Amsterdam, Amsterdam, The Netherlands B. Graimann Department of Translations Research and Knowledge Management, Otto Bock Healthcare GmbH, Duderstadt, Germany M. Russold Department of Translational Research, Otto Bock Healthcare Products GmbH, Wien, Austria N. Šarabon Department of Health Study, University of Primorska and with S2P Science To Practice d.o.o, Koper, Slovenia H. Houdijk Human Movement Sciences Amsterdam and with Heliomare Research and Development, Amsterdam, The Netherlands © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_51
311
312
J. Babiˇc et al.
1 Context Low-back pain (LBP) is a growing issue in our modern society. From all the diseases studied in the Global Burden of Disease Study from 2010 (Institute for Health Metrics and Evaluation), LBP was ranked number one concerning the years lived with disability in Europe [1]. Regarding the overall burden, LBP was ranked third in Central and Eastern Europe and first in Western Europe [1]. Depending on sex and region, the European point prevalence of LBP was up to 15.6 %. Prevalence and burden were found to increase with age and are even expected to further increase within the next years due to the current demographic developments [2]. From several studies reviewing the costs related to LBP in European countries, it has been calculated that the lowest cost per capita in Europe can be found in Belgium with e116 and e209 in Sweden [3]. Most of the costs of LBP are related to reduced work capacity, sick-leaves, early retirement and legal claims. To reduce those costs and the burden on the workers, work place interventions such as education, exercise and braces, are becoming widely accepted. However, most of these measures ease the consequences of LBP but don’t eliminate their causes. It has been shown that a strong correlation between physically demanding jobs and prevalence of LBP exists [4]. Movements which cause LBP in these jobs are: handling heavy loads, repeated lifting and turning, and working time spent in a flexed position. The SPEXOR project targets to design a spinal exoskeleton which assists these movements to prevent LBP for able-bodied workers and to support workers with LBP who are in vocational reintegration. Most of the existing assistive robotic devices are exoskeletons which augment the motion of the human’s limbs. However, they mostly neglect the role of the spine in transferring the load from the upper body and limbs to the lower limbs, which will be the main contribution of the SPEXOR project.
2 Objectives In the SPEXOR project, see Fig. 1, a spinal exoskeleton for working in sustained awkward postures and those involved in manual handling of heavy loads, will be designed. The exoskeleton will have several modes: The assistive mode will prevent excessive or sustained non-neutral spine postures. It will also limit the mechanical load on the spinal column. The monitoring mode will monitor spinal movements and spinal loading and provide feedback to the user. In the training mode, the device will provide active neuromuscular trunk conditioning. In this mode the exoskeleton will act as an advanced fitness or rehabilitation device to strengthen the wearer’s muscles and prevent spine disorders. Finally in the diagnostic mode the exoskeleton will provide an objective assessment of neuromuscular trunk functions. This mode will serve as a diagnostic tool to forecast possible physiological issues concerning the spine of the wearer.
SPEXOR: Spinal Exoskeletal Robot for Low Back Pain Prevention …
313
Fig. 1 A conceptual image of the SPEXOR exoskeleton [5]
The contribution of the authors of this paper will mainly be situated in the design and development of a flexible exoskeleton mechanism for spinal support. This mechanism has to be efficient as well as unobtrusive for the wearer. In this first design the emphasis will be on the requirements for LBP prevention and the use of passive viscoelastic elements for gravity compensation. The design will be made in a way to passively transfer external loads acting on the upper body and arms to the wearer’s legs. In the second phase of the project the authors are responsible for the design and development of compliant actuators for this exoskeleton. The actuation will be designed to exhibit an adaptive compliant behavior. This will allow a comfortable interaction with the user and provide different levels of mechanical effort, depending on the need of torque augmentation, rehabilitation or training activities.
3 Going Beyond the State of the Art Several power augmenting and/or assistive exoskeletons have already been developed. From a technical point of view they can be split in two categories: active and passive devices. Although active devices are more powerful, they are also heavier as they require actuators and a power source. Examples of such active exoskeletons are the power assist suit for day-care from the Kanagawa Institute of Technology, JP [6], the lightweight active power assist device HAL from the Tsukuba University, JP [7], exoskeletons for the augmentation of soldier’s combat abilities of the Defense Advanced Research Project Agency, USA, and the Lower extremity exoskeleton BLEEX, CA [8]. Recently an FP7 project named RoboMate
314
J. Babiˇc et al.
has started. The RoboMate project is also targeting the issue of workspace injuries but, in contrast with the SPEXOR project, a full body exoskeleton is developed. Passive devices are mainly used to reduce the burden of the user rather than to amplify the wearer’s forces [9]. Some examples of such devices are the Smart Suite lite, a compact lightweight robotic suit which utilizes the elastomeric force of elastic materials to assist the wearer [9] and a passive device for enhancing limb movement for people with neuromuscular disabilities [10]. A crucial element in the development of active exoskeletons are the actuators. For the safety of the wearer a compliant behavior is desired [11]. Compliant actuators possess a physical elastic element capable of storing and releasing energy. Those actuators have an equilibrium position, which is the position where they do not exert any force/torque. Deviations from this position are allowed. This allows a safer human-robot interaction, shock absorption, and can lead to a greater energy efficiency than with stiff actuators [12]. Although a lot of effort has already been spent on developing exoskeletons, only a few targeted the spine or lower back. The few which did, focused mostly on the assistance of the trunk flexion/extension [6]. To our knowledge, there is no exoskeleton which has been built yet that would provide assistance for functional trunk movements; an exoskeleton that doesn’t only take power assistance into account but also the spine motions that could result in neuromuscular injuries and subsequently in LBP. This will be done in the SPEXOR project. The exoskeleton will be effective as assistive exoskeleton for laborers which have to work in awkward postures and repeatedly lift loads, as training device and as diagnostic tool. A wide range of possibilities for exercise and measurement protocols will be provided by SPEXOR. The combination of the different modes of the SPEXOR exoskeleton will provide a powerful preventive tool against the most frequent musculoskeletal disorder of today, LBP. This will make the spinal exoskeleton developed in SPEXOR a top-end device that goes well beyond the state-of-the-art. Acknowledgments Members of SPEXOR consortium: J. Babiˇc (coordinator), T. Petriˇc, R. Goljat: Jožef Stefan Institute, Slovenia; K. Mombaur (PI), M. Sreenivasa, M. Millard, P. Manns: Heidelberg University, Germany; D. Lefeber (PI), C. Rodriguez-Guerrero, L. De Rijcke, M. Näf: Vrije Universiteit Brussel, Belgium; J. van Dieën (PI), I. Kingma, G. Faber, S. Bruin, A. Koopman: VU University Amsterdam, The Netherlands; B. Graimann (PI), M. Tüttemann, A. Kurzweg, J. Bornmann, H. Glindemann: Otto Bock Healthcare GmbH, Germany; M. Russold (PI), D. Pieringer: Otto Bock Healthcare Products GmbH, Austria; N. Šarabon (PI), A. Panjan, K. Kastelic, M. Savi´c: S2P Science to practice d.o.o., Slovenia; H. Houdijk (PI), C. van Bennekom (PI), J. Nachtegaal, S. Baltrusch: Heliomare, The Netherlands. Funding: This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 687662 - SPEXOR. Project webpage: http://www.spexor.eu.
SPEXOR: Spinal Exoskeletal Robot for Low Back Pain Prevention …
315
References 1. Buchbinder, R., Blyth, F.M., March, L.M., Brooks, P., Woolf, D.A., Hoy, D.G.: Placing the global burden of low back pain in context. Best Pract. Res.: Clin. Rheumatol. 27, 575–589 (2013) 2. Hoy, D., March, L., Brooks, P., Blyth, F., Woolf, A., Bain, C., Williams, G., Smith, E., Vos, T., Barendregt, J.: The global burden of low back pain: estimates from the Global Burden of Disease 2010 study. Ann. Rheum. Dis. (2014) 3. Lambeek, L.C., van Tulder, M.W., Swinkels, I.C.S., Koppes, L.L.J., Anema, J.R., van Mechelen, W.: The trend in total cost of back pain in The Netherlands in the period 2002 to 2007. Spine 36(13), 1050–1058 (2011) 4. Waddell, G., Burton, A.K.: Occupational health guidelines for the management of low back pain at work: evidence review. Occup. Med. 51, 124–135 (2001) 5. Spinal Exoskeletal Robot for Low Back Pain Prevention and Vocational Reintegration. http:// www.spexor.eu 6. Yamamoto, K., Ishii, M., Noborisaka, H., Hyodo, K.: Stand alone wearable power assisting suit-sensing and control systems. In: International Workshop on Robot and Human Interactive Communication, pp. 661–666 (2004) 7. Kawamoto, H., Sankai, Y.: Power assist method based on phase sequence and muscle force condition for HAL. Adv. Robot. 19, 717–734 (2005) 8. Kazerooni, H.: Hybrid control of the berkeley lower extremity exoskeleton (BLEEX). Int. J. Robot. Res. 25(5–6), 561–573 (2006) 9. Imamura, Y., Tanaka, T., Suzuki, Y., Takizawa, K., Yamanaka, M.: Motion-based design of elastic belts for passive assistive device using musculoskeletal model. In: International Conference on Robotics and Biomimetics, pp. 1343–1348 (2011) 10. Rahman, T., Sample, W., Jayakumar, S., King, M.M., Wee, J.Y., Seliktar, R., Alexander, M., Scavina, M., Clark, A.: Passive exoskeletons for assisting limb movement. J. Rehabil. Res. Dev. 43(5), 583 (2006) 11. Vallery, H., Veneman, J., van Asseldonk, E., Ekkelenkamp, R., Buss, M., van Der Kooij, H.: Compliant actuation of rehabilitation robots. IEEE Robot. Autom. Mag. 15(3), 60–69 (2008) 12. Vanderborght, B., Albu-Schaeffer, A., Bicchi, A., Burdet, E., Caldwell, D., Carloni, R., Wolf, S.: Variable impedance actuators: moving the robots of tomorrow. In: International Conference on Intelligent Robots and Systems, pp. 5454–5455 (2012)
Part XI
Posters
HeSA, Hip Exoskeleton for Superior Assistance Thomas G. Sugar, Eduardo Fernandez, Darren Kinney, Kevin W. Hollander and Sangram Redkar
Abstract A hip exoskeleton was designed that can assist hip flexion and extension. The device incorporates a motor, ball-screw, and spring in a lightweight package. The total weight including the battery is 2.95 kg. The system uses 20 W of power per leg. The system is controlled based on the phase angle of each leg and the torque is applied in synchrony with the user’s steps. The device assists walking, running, and does not interfere when going up and down stairs.
1 Introduction Wearable robotic systems are being developed to assist at the hips, knees and ankles [1–10]. These systems must be lightweight, energy efficient, and must conform to the human. For systems to assist gait, they must not interfere with the human motion. We have developed a hip exoskeleton, HeSA (Hip Exoskeleton for Superior Assistance). It was initially designed to assist soldiers walking on patrol, wearing a tactical vest that holds body armor. This tactical vest can weigh approximately 18 kg, see Fig. 1. The system was also integrated into a standard back brace to be used in a manufacturing environment, see Fig. 1. The design, control and testing of the system will be described.
T.G. Sugar (&) Technology at SpringActive, Inc., Research and Technology, Wearable Robotics Association, Arizona State University, Tempe, USA e-mail:
[email protected] E. Fernandez S. Redkar Arizona State University, Tempe, USA D. Kinney SpringActive, Inc, Tempe, USA K.W. Hollander Product Development at SpringActive, Inc, Tempe, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_52
319
320
T.G. Sugar et al.
Fig. 1 The hip exoskeleton in the left figure is integrated into a tactical vest that holds body armor. In the right figure, the HeSA device is integrated into a workers back brace. It can be used in a manufacturing environment
2 Design The HeSA device provides bi-directional torque support to each hip. The device is capable of supporting both walking and running gaits. Peak torque for hip flexion and extension Average electrical power when walking at 1.34 m/s per motor Total device weight without battery Battery weight, 24 W 72 W h Motor/ball screw/spring actuator
15 Nm 20 W 2.27 kg 0.68 kg
The hip actuator can disengage from the thigh attachment to allow for extended range of motion. The extended free motion allows the user to flex the hip to ascend stairs and get easily in and out of cars. The actuator is attached to the waist using a clamp structure. The clamp can be adjusted up and down using pins and it slides back on forth on the pins to align the motor joint with the hip joint. The structure also has a pin joint to allow hip abduction.
3 Control The system is controlled based on the phase angle of each hip. The phase angle can be calculated by knowing the angular position and velocity of the hip. Because each hip movement is unique, the phase angle is calculated independently for each leg.
HeSA, Hip Exoskeleton for Superior Assistance
321
The system uses a phase oscillator to assist hip flexion and extension. The controller has been developed and described in [11, 12]. A force/torque applied based on the sine of the phase angle has been shown to assist and maintain a limit-cycle. A force/torque applied based on the cosine of the phase angle has been shown to adjust the frequency of the oscillation. For the upper leg, the inertia of the thigh is modeled as I, the damping by b, and the stiffness of the thigh by k. A hip sensor is used to measure the angular motion of the thigh modeled by h_ and h. I €h þ bh_ þ kh ¼ c sinð;Þ þ d cosð;Þ ch_ dx2 h2 I €h þ bh_ þ kh ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h_ 2 þ x2 h2 h_ 2 þ x2 h2
ð1Þ
If c is positive, the system oscillates back and forth. The energy is always bounded. If c is negative, the energy is dissipated and the system state returns to zero. With a larger, positive value for c, more assistance is provided. If d is positive, the frequency of the oscillations are decreased, and if d is negative, the frequency increases. In our work x was fixed to a value of 2 * pi * 3.
4 Testing The device was worn and tested. Preliminary data is shown in Figs. 2 and 3. The robot angle measures the internal angle of the thigh strap and is shown in Fig. 2. This angle approximates the hip angle because there are some small angular movements of the device as the actuator attachment is fixed to a flexible waist band or tactical vest. For example, there are small oscillations as the device changes direction or when the load is applied. The applied, assistive, torque is shown in Fig. 3. It is calculated by determining the spring deflection in the mechanism. The ball screw carriage or “nut” pushes on a spring which in turn pushes on the lever arm connected to the thigh attachment. The position of the lever arm attached to the thigh is calculated from the robot angle. The motor position is adjusted based on the sine of the phase angle. After weight acceptance of the leg, the motor applies positive 9 Nm torques to extend the hip and assist in the stance phase. As the thigh is flexed during the swing phase, 8 Nm torques are applied to lift the leg against gravity. The value of c is adjusted determining the peak torques. In Fig. 4, the user can climb over obstacles and run.
322
T.G. Sugar et al.
Fig. 2 Robot angle is measured with 30° of flexion and 20° of extension
Fig. 3 The angle of the robot is converted to mm of movement. The position of the ball screw is labeled “nut pos.” The torque applied is given in Nm
Fig. 4 Walking over ground and running
5 Conclusion We have designed a lightweight hip exoskeleton that can apply 15 Nm of assistive torque in concert with the user. The user can walk, run, and climb over objects. Further developments will include the ability to assist stair climbing and sit-to-stand tasks.
HeSA, Hip Exoskeleton for Superior Assistance
323
Acknowledgments A patent application has been filed on the phase oscillator and mechanisms. This work is partially supported by the DARPA grant W911NF-15-1-0162. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the funding agency
References 1. Asbeck, A.T., De Rossi, S.M., Holt, K.G., Walsh, C.J.: A biologically inspired soft exosuit for walking assistance. Int. J. Robot. Res. 0278364914562476 (2015) 2. Asbeck, A.T., Schmidt, K., Walsh, C.J.: Soft exosuit for hip assistance. Robot. Auton. Syst. (2014) 3. Bharadwaj, K., Sugar, T.G., Koeneman, J.B., Koeneman, E.J.: Design of a robotic gait trainer using spring over muscle actuators for ankle stroke rehabilitation. ASME J. Biomech. Eng. 127, 1009–1013 (2005) 4. Blaya, J.A., Herr, H.: Adaptive control of a variable-impedance ankle-foot orthosis to assist drop-foot gait. IEEE Trans. Neural Syst. Rehabil. Eng. 12 (2004) 5. Esposito, E.R., Blanck, R.V., Harper, N.G., Hsu, J.R., Wilken, J.M.: How does ankle-foot orthosis stiffness affect gait in patients with lower limb salvage? Clin. Orthop. Relat. Res. 472, 3026–3035 (2014) 6. Giovacchini, F., Vannetti, F., Fantozzi, M., Cempini, M., Cortese, M., Parri, A., Yan, T., Lefeber, D., Vitiello, N.: A light-weight active orthosis for hip movement assistance. Robot. Auton. Syst. 73, 123–134 (2015) 7. Meijneke, C., van Dijk, W., van der Kooij, H.: Achilles: an autonomous lightweight ankle exoskeleton to provide push-off power. Presented at the 5th IEEE RAS & EMBS International Conference on Biomedical Robotics and Biomechatronics (2014) 8. Shamaei, K., Cenciarini, M., Adams, A.A., Gregorczyk, K.N., Schiffman, J.M., Dollar, A.M.: Effects of exoskeletal stiffness in parallel with the knee on the motion of the human body center of mass during walking. In: 2015 IEEE International Conference on Robotics and Automation (ICRA), pp. 5557–5564 (2015) 9. Hollander, K.W., Cahill, N., Holgate, R., Churchwell, R., Clouse, P., Kinney, D., Boehler, A.: A passive and active joint torque augmentation robot (JTAR) for hip gait assistance. In: ASME 2014 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, pp. V05AT08A079–V05AT08A079 (2014) 10. Hollander, K.W., Cahill, N., Holgate, R., Churchwell, R., Clouse, P., Kinney, D., Boehler, A., Ward, J.: A joint torque augmentation robot (JTAR) for ankle gait assistance. In: ASME 2014 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (2014) 11. Kerestes, J., Sugar, T.G., Holgate, M.: Adding and subtracting energy to body motion: phase oscillator. In: ASME 2014 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, pp. V05AT08A004–V05AT08A004 (2014) 12. Sugar, T.G., Bates, A., Holgate, M., Kerestes, J., Mignolet, M., New, P., Ramachandran, R. K., Redkar, S., Wheeler, C.: Limit cycles to enhance human performance based on phase oscillators. J. Mech. Robot. 7, 011001 (2015)
SPEXOR: Towards a Passive Spinal Exoskeleton L. De Rijcke, M. Näf, C. Rodriguez-Guerrero, B. Graimann, H. Houdijk, J. van Dieën, K. Mombaur, M. Russold, N. Sarabon, J. Babiˇc and D. Lefeber
Abstract Most assistive robotic devices are exoskeletons which assist or augment the motion of the limbs and neglect the role of the spinal column in transferring load from the upper body and arms to the legs. In this part of the SPEXOR project we will fill this gap and design a novel, passive spinal exoskeleton to prevent low-back pain in able bodied workers and to support workers with low-back pain in vocational rehabilitation.
L. De Rijcke (B) · M. Näf · C. Rodriguez-Guerrero · D. Lefeber Robotics & Multibody Mechanics Research Group, Department of Mechanical Engineering, Vrije Universiteit Brussel, Ixelles, Belgium e-mail:
[email protected] http://mech.vub.ac.be/robotics B. Graimann Department of Translations Research and Knowledge Management, Otto Bock Healthcare GmbH, Duderstadt, Germany J. van Dieën Neuromusculoskeletal Mechanics Research Program, MOVE Research Institute Amsterdam, Amsterdam, The Netherlands H. Houdijk Faculty of Human Movement Sciences, VU University Amsterdam and with Heliomare Research and Development, Amsterdam, The Netherlands K. Mombaur Department of Optimization in Robotics & Biomechanics, Heidelberg University, Heidelberg, Germany M. Russold Department of Translational Research, Otto Bock Healthcare Products GmbH, Wien, Austria N. Sarabon Department of Health Study, University of Primorska and with S2P Science To Practice d.o.o, Koper, Slovenia J. Babiˇc Department of Automation, Biocybernetics and Robotics, Jožef Stefan Intitute, Ljubljana, Slovenia © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_53
325
326
L. De Rijcke et al.
1 Introduction Low-back pain (LBP) is a growing issue in our modern society. From all the diseases studied in the Global Burden of Disease Study from 2010 (Institute for Health Metrics and Evaluation), LBP was ranked number one concerning the years lived with disability in Europe [1]. Regarding the overall burden, LBP was ranked third in Central and Eastern Europe and first in Western Europe [1]. In addition of lowering the life quality of the patient, LPB also causes an economical burden for the society. From several studies reviewing the costs related to LBP in European countries, it has been calculated that the lowest cost per capita in Europe can be found in Belgium with e116 and goes up to e209 in Sweden [2]. Most of these costs are related to reduced work capacity, sick-leaves, early retirement and legal claims. To reduce this, preventive measures such as education and training, adjustment of work stations, re-organization of the work process, the use of mechanical aids like cranes or balancers and the use of braces, are becoming widely accepted [3]. However, this may not always be possible when the load is changing constantly (like e.g. for movers) or the workplace makes these measures infeasible (e.g. with confined spaces). It has been shown that a strong correlation between physically demanding jobs and prevalence of LBP exists [4]. Movements which cause LBP in these jobs are: handling heavy loads, repeated trunk flexion and rotation, and working time spent in a flexed position. The SPEXOR project targets to design a spinal exoskeleton, illustrated in Fig. 1, which assist these movements to prevent LBP for able-bodied workers and to support workers with LBP who are in vocational reintegration. In the first part of the project, a passive, spinal orthotic module which assist the wearer with the lumbar flexion/extension will be developed.
Fig. 1 A conceptual image of the SPEXOR exoskeleton [5]
SPEXOR: Towards a Passive Spinal Exoskeleton
327
2 SOTA of Passive Exoskeletons Passive devices are mainly used to reduce the burden of the user rather than to amplify the wearer’s forces [6]. In the last years several passive spinal exoskeletons have been developed. The first one, to our knowledge, was the Personal Lift Augmentation Device (PLAD) [7], designed to prevent lower-back pain of workers. The PLAD device is an exoskeleton which is based on the concept of stored elastic energy. When the wearer bends, elastic elements are compressed and energy is stored. When the wearer goes back to an upright position, the elastic elements release the energy and the exoskeleton gives assistance to the wearer [7]. This device exchanges forces with the wearer at the spine, shoulders, pelvis and feet. For this device significant reductions of the back muscle activity have been reported during lifting and static bending [3]. The Bending Non-Demand Return (BNDR) [8] is a similar device. It also stores energy when the wearer is bending forward and releases the energy when the wearer goes upright. Its frame covers the trunk and pelvis of the wearer. It exchanges forces with the wearer at the upper legs and the chest. An important difference with the PLAD is that the BNDR reduces the torso flexion range while the PLAD doesn’t interfere with the body kinematics during flexion. The Smart Suit Lite (SSL) was developed to reduce the muscle load in the lumbar region during nursing care [9]. This passive power assist device uses elastic belts. The belts exercise an assistive torque on the lumbar joint and on the hip joint, when the wearer changes posture. An additional elastic belt is placed around the torso in order to stabilize it, similar to the working of a corset. It has been proven that by increasing the pressure surrounding the pelvis, the corset increases the lumbar support or the intra-abdominal pressure, which in turn reduces the intradiscal pressure and stabilizes posture [9]. The passive LAEVO exoskeleton [3] was designed to assist nurses for the lifting of patients. The device is connected with the wearer at the chest, the lower back and the upper legs. The exoskeleton is intended to transfer load from the lower back to the chest and to the legs by making use of circular tubes. Those tubes are situated at both sides of the body and have spring-like characteristics. For this device significant reduces of the back and leg muscles have been observed during the bending motion. Also the endurance was ameliorated. However the patients reported discomfort in the chest region.
3 Going Beyond Previous devices are all assisting the trunk flexion/extension. Interesting with the SSL is that it also stabilizes the posture by using a corset-like belt. The implementation of the range of motion reduction of the trunk flexion in the BNDR is important as it has been proven that movement in this boundary region is a risk factor for back pain [10].
328
L. De Rijcke et al.
However, these devices don’t account for rotation and/or lateral bending. The passive exoskeleton developed in the SPEXOR project will assist the trunk flexion/extension, have an adjustable trunk flexion range of motion limit, and implement adjustable rotation and lateral bending angle limits. The passive exoskeleton will interact with the wearer without interfering with walking and stair climbing, respectively. The interfaces between the wearer and the exoskeleton will be designed in a way to not cause discomfort to the wearer. Attention will be paid to apply normal forces from the exoskeleton to the wearer (avoiding shear forces). It will be designed in a way to be comfortable as well for men as for women. The main force acting on the trunk while bending is the gravitational force acting on the upper body. This force is characterized by a sinusoidal function. In order to achieve a proportional compensation a spring will be combined with a cam-profile, similar to [11].
4 Conclusion Although several passive devices have been developed to prevent lower-back pain, many aspects and functionalities have still to be improved. The criteria pointed out for the passive exoskeleton developed within the SPEXOR project will make the exoskeleton effective as assistive device for laborers which have to work in awkward postures and repeatedly lift loads, as training device and as diagnostic tool. Acknowledgments This work is funded by the European Commission under Grant agreement Nr. 687662. http://www.spexor.eu.
References 1. Buchbinder, R., Blyth, F.M., March, L.M., Brooks, P., Woolf, D.A., Hoy, D.G.: Placing the global burden of low back pain in context. Best Pract. Res.: Clin. Rheumatol. 27, 575–589 (2013) 2. Lambeek, L.C., van Tulder, M.W., Swinkels, I.C.S., Koppes, L.L.J., Anema, J.R., van Mechelen, W.: The trend in total cost of back pain in The Netherlands in the period 2002 to 2007. Spine 36(13), 1050–1058 (2011) 3. Bosch, T., van Eck, J., Knitel, K., de Looze, M.: The effects of a passive exoskeleton on muscle activity, discomfort and endurance time in forward bending work. Appl. Ergon. 54, 212–217 (2016) 4. Waddell, G., Burton, A.K.: Occupational health guidelines for the management of low back pain at work: evidence review. Occup. Med. 51, 124–135 (2001) 5. Spinal Exoskeletal Robot for Low Back Pain Prevention and Vocational Reintegration. http:// www.spexor.eu 6. Imamura, Y., Tanaka, T., Suzuki, Y., Takizawa, K., Yamanaka, M.: Motion-based design of elastic belts for passive assistive device using musculoskeletal model. In: International Conference on Robotics and Biomimetics, pp. 1343–1348 (2011)
SPEXOR: Towards a Passive Spinal Exoskeleton
329
7. Sadler, E.M., Graham, R.B., Stevenson, J.M.: The personal lift-assist device and lifting technique: a principal component analysis. Theoret. Issues Ergon. Sci. 0139, 1–16 (2011) 8. Ulrey, B.L., Fathallah, F.A.: Subject-specific, whole-body models of the stooped posture with a personal weight transfer device. J. Electromyogr. Kinesiol. 23(1), 206–215 (2013) 9. Imamura, Y., Tanaka, T., Suzuki, Y., Takizawa, K., Yamanaka, M.: Analysis of trunk stabilization effect by passive power-assist device. J. Robot. Mechatron. 26(6), 791–798 (2014) 10. Hoogendoorn, W.E., van Poppel, M.N.M., Bongers, P.M., Koes, B.W., Bouter, L.M.: Systematic review of psychosocial factors at work and private life as risk factors for back pain. Spine 25(16), 2114–2125 (2000) 11. Koser, K.: A cam mechanism for gravity-balancing. Mech. Res. Commun. 36(4), 523–530 (2009). doi:10.1016/j.mechrescom.2008.12.005
Autonomous Soft Exosuit for Hip Extension Assistance Nikos Karavas, Jinsoo Kim, Ignacio Galiana, Ye Ding, Adam Couture, Diana Wagner, Asa Eckert-Erdheim and Conor Walsh
Abstract In this abstract, we describe a mono-articular soft exosuit to assist with hip extension during overground walking. The system is comprised of a mobile Bowden-cable-driven electrical actuation unit, soft textiles, and a load cell and an inertia measurement unit per leg. The exosuit applies forces with a peak of 300 N enabled by an IMU-based iterative control algorithm. This iterative controller detects the onset timing of the hip extension assistance based on an estimation of the maximum hip flexion angle. The timing and magnitude of the applied peak force is modulated by generating step-by-step actuator position profiles based on the previously measured assistive force. Results from a human subject during overground walking at self-selected speed indicate the robustness of the system to apply effectively forces with a high consistency in terms of magnitude and timing of the peak force profile.
1 Introduction Plethora of conventional rigid exoskeletons have been developed over the past decades to assist with human locomotion [1]. These devices can support the wearer’s body weight and apply a substantial portion of the required biological torques [2]. However, such rigidly structured devices can restrict the user’s movements and apply undesired forces resulting in discomfort. In addition to
This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA), Warrior Web Program (Contract No. W911NF-14-C-0051). This work was also partially funded by the Wyss Institute for Biologically Inspired Engineering and the John A. Paulson School of Engineering and Applied Sciences at Harvard University. N. Karavas J. Kim I. Galiana Y. Ding A. Couture D. Wagner A. Eckert-Erdheim C. Walsh (&) John A. Paulson School of Engineering and Applied Science and the Wyss Institute for Biologically Inspired Engineering, Harvard University, Cambridge, MA 02138, USA e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_54
331
332
N. Karavas et al.
kinematic restrictions and joint misalignments rigid exoskeletons often have high inertias which can impede the user especially when control tracking deficiencies occur [3]. To overcome these limitations, we have recently proposed exosuits which are composed of soft materials such as textiles and elastomers to provide a compliant means to interface with the human [4–6]. Exosuits apply tensile forces to the body through load paths well defined by the textile architecture. As exosuits incorporate soft components they have extremely low inertia and eliminate the challenges associated to joint misalignments. Several research groups have investigated the benefits of powering the hip joint using a wearable robot. In particular, in [7] it is proposed that providing external power to the hip joint could yield a greater reduction in metabolic cost, than providing the same amount of power at the ankle joint. Moreover, in [8] authors reported a reduced metabolic cost of *18 % compared to walking with exoskeleton unpowered when providing an assistive torque corresponding to 100 % of the average torque of the human hip joint. To this end, we present an autonomous exosuit system to assist hip extension during overground walking especially outdoors. The system is capable of consistently delivering mechanical power to the user across subjects, varied walking speeds and step lengths. In addition, it is intuitive and does not hinder other motion tasks (uphill/downhill walking, stair climbing/descending and walking over obstacles). It is also worth to mention that the robustness of each subsystem (i.e. actuation in terms of thermo-mechanical fatigue, suit in terms of component drifting, and controller in terms of tracking performance) has been successfully evaluated by several subjects testing during 3 miles hiking on rough terrains.
2 System Description 2.1
Actuation and Suit
The Bowden-cable-driven actuation unit is a two degree-of-freedom (DoF) system that actuates each leg independently and it is designed so that it can be simply mounted on top of a MOLLE II Large Rucksack as shown in Fig. 1. Each DoF consists of a Maxon EC 4-pole, 200 W motor that is connected to a gearbox of ratio 51:1 and drives an 8 cm diameter pulley. A motor controller unit (by Elmo Motion Control Inc.) controls each motor and is integrated into a custom electronic board. In addition, an Amtel microprocessor communicates with the Elmo motor controller through the CANOpen communication protocol and provides with the required position commands. A removable battery unit containing two 6-cell Li-Po batteries (5200 mAh) is placed on the bottom of the rucksack and provides the device with an autonomy over approximately 6 miles. The overall system weight is 4.9 kg (actuation, 3.1 kg, suit, 0.4 kg, and battery, 1.4 kg).
Autonomous Soft Exosuit for Hip Extension Assistance
333
Fig. 1 A subject wearing the autonomous soft exosuit (left). Load path to actuate hip extension and components of the suit (right)
The exosuit consists of a spandex base layer, a waist belt that wraps around the subject’s pelvis, two thigh brace pieces and two IMU elastic bands (for IMU placement). The waist belt and the thigh brace provide with the two required anchor points. In particular, the Bowden cable sheath connects to the back of the waist belt and the inner cable connects to the back of the thigh brace. Therefore, when the motor retracts the cable a hip extension torque is created and when the motor feeds out cable the suit is slack.
2.2
IMU-Based Iterative Controller
The main goal of the mobile soft exosuit is to deliver consistent and robust hip extension force in sync with subject’s gait dynamics. Based on our examination of the relationship between metabolic energy reduction and net work rate provided by actuators in previous study [5], the force profile was chosen to mimic the biological joint torque, with the assumption that this will allow the muscle activation to decrease proportionally and consequently reduce the metabolic energy. From an analysis of the hip biomechanics it can be found that the onset of the hip positive power concurs with the maximum hip flexion angle, and the hip extensor muscles start to activate slightly earlier to this [4, 6]. Therefore, we select to detect the maximum hip flexion using an IMU sensor mounted on the thigh of each leg. A force-based position control is then used to create the hip extension torque with onset, peak and offset timings similar to the one of the biological hip joint torque [5]. To ensure a consistent and robust delivered force given the variability in hip kinematics, kinetics, and suit positioning, the controller adjusts the pretension level and the maximum amplitude of the motor position command based on the force profile of the previous gait step.
334
N. Karavas et al.
Fig. 2 Hip extension force (load cell) and thigh angle (IMU) based on the percentage of gait cycle (right leg)
3 Results To evaluate the performance of our proposed system and IMU-based controller, we conducted an overground walking experiment. A healthy subject (male, age 30, height 179 cm, weight 76 kg) wore the system and walked with the freedom to vary his walking speed as preferred. Our objective is to deliver constant and robust hip extension forces of 300 N to the user (which corresponds to a hip torque of approximately 45 Nm and a 60 % of the nominal biological torque) regardless of his gait pattern. Thus, the performance of the assistive force has been evaluated using the following metrics on a step by step basis: the magnitude and timing of the peak force and the offset timing of the force waveform. Figure 2 depicts the average and standard deviation of the measured forces and thigh angles for a trial of 10 min of walking, segmented based on the percentage of the gait cycle. Note that, the percentage of the gait cycle has been calculated based on the average onset timing (i.e. maximum hip flexion) and 0 % gait cycle corresponds to the heel strike. The average of the peak force was 294 ± 11 N, which results in an error of *3 %. In addition, the average of the peak and offset timing of the force profile were 16.4 ± 0.8 % and 28.2 ± 0.8 % gait cycle, respectively. These results demonstrate the ability of the system to accurately control the peak force and the peak force timing to the desired values.
4 Conclusion We presented an autonomous soft exosuit for hip extension assistance. The system is able to deliver constantly and repetitively mechanical power to the user with an error in the magnitude of the peak force of *3 % regardless of the kinematic and kinetic variations occurring during gait. In a previous study with an off-board actuation platform which applied similar force profiles with a peak of 200 N, we
Autonomous Soft Exosuit for Hip Extension Assistance
335
found that metabolic energy is reduced by an average between 5.7 and 8.5 % compared to the unpowered condition [6]. Thus, we expect that a greater reduction in metabolic energy will be achieved with a hip assistance of 300 N. Future work will include investigation of the proposed system’s benefit in energy expenditure among different subjects.
References 1. Guizzo, E., Goldstein, H.: The rise of the body bots. Spectrum, IEEE 42(10), 50–56 (2005) 2. Farris, R.J., Quintero, H.A., Goldfarb, M.: Preliminary evaluation of a powered lower limb orthosis to aid walking in paraplegic individuals. IEEE Trans. Neural Syst. Rehabil. Eng. 19(6), 652–659 (2011) 3. Browning, R.C., Modica, J.R., Kram, R., Goswami, A.: The effects of adding mass to the legs on the energetics and biomechanics of walking. Med. Sci. Sport Exerc. 39, (2007) 4. Asbeck, A.T., Schmidt, K., Galiana, I., Wagner, D., Walsh, C.: Multi-joint soft exosuit for gait assistance. In: IEEE International Conference on Robotics and Automation (ICRA), May 2015 5. Ding, Y., Galiana, I., Siviy, C., Panizzolo, F.A., Walsh, C.: IMU-based iterative control for hip extension assistance with a soft exosuit. In: IEEE International Conference on Robotics and Automation (ICRA), May 2016 (to be published) 6. Ding, Y., Panizzolo, F.A., Siviy, C., Malcolm, P., Galiana, I., Holt, K.G., Walsh, C.: Effect of timing of hip extension assistance during loaded walking with a soft exosuit. J. Neuroeng. Rehabil. (in review) 7. Sawicki, G.S., Lewis, C.L., Ferris, D.P.: It pays to have a spring in your step. Exerc. Sport Sci. Rev. 37, 130–138 (2009) 8. Ronsse, R., Koopman, B., Vitiello, N., Lenzi, T., De Rossi, S.M.M., van den Kieboom, J., van Asseldonk, E., Carrozza, M.C., van der Kooij, H., Ijspeert, A.J.: Oscillator-based walking assistance: a model-free approach. IEEE Int. Conf. Rehabil. Robot. (2011)
Comparison of Ankle Moment Inspired and Ankle Positive Power Inspired Controllers for a Multi-Articular Soft Exosuit for Walking Assistance Martin Grimmer, Brendan T. Quinlivan, Sangjun Lee, Philippe Malcolm, Denise M. Rossi, Christopher Siviy and Conor J. Walsh
Abstract Mobility can be limited due to age or impairments. Wearable robotics provide the chance to increase mobility and thus independence. A powered soft exosuit was designed that assist with both ankle plantarflexion and hip flexion through a multi-articular suit architecture. So far, the best method to reduce metabolic cost of human walking with external forces is unknown. Two basic control strategies are compared in this study: an ankle moment inspired controller (AMIC) and an ankle positive power inspired controller (APIC). Both controllers provided a similar amount of average positive exosuit power and reduced the net metabolic cost of walking by 15 %. These results suggest that average positive power could be more important than assistive moment during single stance for reducing metabolic cost. Further analysis must show if one of the approaches has advantages for wearers comfort, changes in walking kinetics and kinematics, balance related biomechanics, or electrical energy consumption.
The material is based upon the work supported by the Defense Advanced Research Projects Agency, Warrior Web Program (Contract No. W911NF-14-C-0051). This work was also funded by the Robert Bosch Stiftung (Grant No. 32.5.G412.0003.0), the National Science Foundation (Grant No. DGE1144152, CNS-1446464), the So Paulo Research Foundation (FAPESP; Grant No. 2015/02116-1) and the Samsung Scholarship This work was partially funded by the Wyss Institute for Biologically Inspired Engineering and the John A. Paulson School of Engineering and Applied Sciences at Harvard University. Martin Grimmer, Sangjun Lee and Brendan T. Quinlivan: Authors contributed equally to this work. M. Grimmer (&) Technische Universitaet Darmstadt, Darmstadt, Germany B.T. Quinlivan S. Lee (&) P. Malcolm C. Siviy C.J. Walsh (&) School of Engineering and Applied Sciences and the Wyss Institute for Biologically Inspired Engineering, Harvard University, Cambridge, MA, USA e-mail:
[email protected] D.M. Rossi University of Sao Paulo, Ribeiro Preto Medical School, Ribeiro Preto, SP, Brazil © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_55
337
338
M. Grimmer et al.
1 Introduction Mobility determines human independence and thus quality of life. It can be affected due to age and/or disease. Next to exercise, wearable robotics can provide solutions for mobility improvements. Several prototypes to assist human movement have been developed and have shown promising results [1–5]. However, to date the best configuration and assistance profile is still unknown. The majority of the first studies with exoskeletons intended for metabolic rate reduction prioritized reducing biological positive joint power over joint moment. It might be possible that assistance profiles that follow the biological moment also effectively reduce metabolic rate because it is known that biological muscles can still consume energy even when they do not produce positive joint power [6, 7]. To address this topic, we conducted a study comparing an ankle moment inspired controller (AMIC) and an ankle positive power inspired controller (APIC) using a multi-articular exosuit [5].
2 Methods The soft exosuit used in this study has a single actuator per leg which assists with ankle plantarflexion and hip flexion through the multi-articular load path specified by the textile architecture (Fig. 1) [4]. An offboard actuation system was used to generate assistive forces, and Bowden cables were used to transmit the forces to the soft exosuit local to the subject’s ankle. On each leg, two gyroscopes and a load cell were attached to measure data from the suit and the wearer. Seven healthy male adults participated in this study (26.71 ± 4.75 y; 68.43 ± 9.46 kg; 1.74 ± 0.06 m; mean ± SD). While participants walked on a treadmill at 1.50 m s−1, two different control strategies were investigated, each with Fig. 1 Experimental setup with the offboard actuation system and the multi-articular soft exosuit that assists with both ankle plantarflexion and hip flexion
Comparison of Ankle Moment Inspired and Ankle Positive Power …
339
Fig. 2 Example position trajectories of the AMIC and APIC. a and b The AMIC performs a force-based position control using a position trajectory inspired by a robotic-tendon model [8] that takes into account the suit-human series stiffness and natural ankle kinematics/kinetics in attempt to produce a biologically relevant ankle moment. c and d The APIC controller performs power-based position control using the ankle speed zero-crossing to ensure that assistance delivered by the exosuit coincides with the positive biological ankle power
their own powered-off condition for relative metabolic comparison. Metabolic rates, lower-limb kinematics, and ground reaction forces were measured (Fig. 1). Ankle moment inspired controller: As shown in Fig. 2a, b, using an approach similar to [8], the AMIC used the suit-human series stiffness of the exosuit [5] and averaged ankle kinematics and kinetics data to identify a position trajectory that would produce biologically relevant assistive ankle moments during push-off. The system performed a force-based position control on a step-by-step basis to reach a maximum peak force of 5.52 N kg−1. Ankle positive power inspired controller: As shown in Fig. 2c, d and previously described in [4], the APIC delivered assistance predominantly during the positive ankle power phase. Ankle speed zero-crossing, which corresponds to the positive power onset, was estimated using the gyroscopes and used to trigger actuation. The system performed a power-based position control on a step-by-step basis to deliver 8 W of positive exosuit power assistance in parallel to the positive biological ankle power [4]. By applying 8 W of positive power we were able to approximately match the average positive exosuit power from the AMIC.
3 Results The AMIC and APIC were able to deliver desired power and moment to the ankle joint and reduce the net metabolic cost of walking by 0.72 ± 0.21 W kg−1 (14.78 ± 3.63 %) and 0.74 ± 0.30 W kg−1 (15.36 ± 5.53 %), respectively. The average peak applied forces were 5.60 ± 0.50 and 4.99 ± 1.36 N kg−1 (Fig. 3a).
340
M. Grimmer et al.
Fig. 3 a Average exosuit force and b average exosuit power for the AMIC and APIC
The average positive exosuit power was 0.133 ± 0.028 and 0.124 ± 0.023 W kg−1 for the AMIC and APIC respectively (Fig. 3b). The average negative exosuit power was 0.026 ± 0.017 and 0.007 ± 0.002 W kg−1. Using the AMIC, peak ankle dorsiflexion during stance was significantly reduced. For both controllers, ankle peak plantarflexion angle increased and occurs earlier in the gait cycle.
4 Discussion & Conclusion APIC produces similar metabolic benefit as AMIC with much lower moment assistance during single stance and similar average positive exosuit power during push-off. This seems to suggest that positive power is more helpful. However, we cannot make strong claims about the specific influence of peak positive power since we did not test the effect of this parameter in isolation from other changes in the actuation profile. Nevertheless, the finding that positive exosuit power could be more important than assistive moment during single stance for reducing metabolic cost aligns well with previous findings [9]. In summary, in this work we showed that the metabolic cost of walking can be reduced using two different controllers, an ankle moment inspired controller and an ankle positive power inspired controller. Further analysis of the data must show if one of the approaches has advantages for wearers comfort, changes in walking kinetics/kinematics, balance related biomechanics or electrical energy consumption. These parameters might give further insight to the benefits of each approach. Acknowledgments Sponsor and financial support acknowledgments are placed in the unnumbered footnote on the first page.
Comparison of Ankle Moment Inspired and Ankle Positive Power …
341
References 1. Malcolm, P., Derave, W., Galle, S., De Clercq, D.: A simple exoskeleton that assists plantarflexion can reduce the metabolic cost of human walking. PLoS ONE 8(2), e56137 (2013) 2. Collins, S.H., Wiggin, M.B., Sawicki, G.S.: Reducing the energy cost of human walking using an unpowered exoskeleton. Nature 522, 212–215 (2015) 3. Mooney, L.M., Herr, H.M.: Biomechanical walking mechanisms underlying the metabolic reduction caused by an autonomous exoskeleton. J. NeuroEng. Rehabil. 13(1), 4–15 (2016) 4. Lee, S., Crea, S., Malcolm, P., Galiana, I., Asbeck, A., Walsh, C.: Controlling negative and positive power at the ankle with a soft exosuit. In: IEEE International Conference on Robotics and Automation (ICRA), Stockholm, 2016 (accepted) 5. Asbeck, A.T., De Rossi, S.M.M., Holt, K.G., Walsh, C.J.: A Biologically inspired soft exosuit for walking assistance. Int. J. Robot. Res. 34(6), 744–762 (2013) 6. Ma, S., Zahalak, G.I.: A distribution-moment model of energetics in skeletal muscle. J. Biomech. 24(1), 21–35 (1991) 7. Ryschon, T.W., Fowler, M.D., Wysong, R.E., Anthony, A.R., Balaban, R.S.: Efficiency of human skeletal muscle in vivo: comparison of isometric, concentric, and eccentric muscle action. J. Appl. Physiol. 83(3), 867–874 (1997) 8. Hollander, K.W., Ilg, R., Sugar, T.G., Herring, D.: An efficient robotic tendon for gait assistance. J. Biomech. Eng. 128(5), 788–791 (2006) 9. Jackson, R.W., Collins, S.H.: An experimental comparison of the relative benefits of work and torque assistance in ankle exoskeletons. J. Appl. Physiol. 119(5), 541–557 (2015)
Biomechanical Analysis and Inertial Sensing of Ankle Joint While Stepping on an Unanticipated Bump Taira Miyatake, Sangjun Lee, Ignacio Galiana, Denise M. Rossi, Christopher Siviy, Fausto A. Panizzolo and Conor J. Walsh
Abstract Walking on uneven terrain with a wearable assistive robot requires the controller to adapt to rapid changes in human’s biomechanics. To do so, the changes due to terrain should be measured using wearable sensors. We investigated human ankle joint mechanics when stepping on different small, unanticipated bumps with either the forefoot or the rearfoot. It was shown that kinematics and kinetics change significantly depending on how humans step on a bump, and that changes in kinematics could be measured by IMUs. This result could be used to inform the design of adaptive controllers for wearable robots that provide optimal assistance to the ankle joint when walking on uneven terrain.
1 Introduction Lower extremity wearable robots have been developed to help able-bodied individuals to walk or carry loads, and several recent studies have shown reductions in energy expenditure in lab-based tests [1–3]. For practical use outside of lab, it is
The material is based upon the work supported by the Defense Advanced Research Projects Agency, Warrior Web Program (Contract No. W911NF-14-C-0051). This study was also funded by Funai Overseas Scholarship (T.M.), Samsung Scholarship (S. L.), and São Paulo Research Foundation (FAPESP; Grant No. 2015/02116-1) (D. M. R.). This work was partially funded by the Wyss Institute for Biologically Inspired Engineering and the John A. Paulson School of Engineering and Applied Sciences at Harvard University. T. Miyatake S. Lee I. Galiana C. Siviy F.A. Panizzolo C.J. Walsh (&) John A. Paulson SEAS, Cambridge, USA e-mail:
[email protected] T. Miyatake S. Lee I. Galiana C. Siviy F.A. Panizzolo C.J. Walsh Wyss Institute at Harvard University, Boston, MA, USA D.M. Rossi Ribeirão Preto Medical School, University of São Paulo, Ribeiãro Preto, SP, Brazil © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_56
343
344
T. Miyatake et al.
crucial that these devices quickly adapt to variable terrain. Previous studies have shown that humans adapt their kinematics and kinetics to irregular surfaces such as wooden blocks [4] or ballast [5]. If designed purely for walking on flat surfaces, assistance from wearable devices may become less optimal or even obstructive to natural motion. However, it is still unclear how humans change their kinematics and kinetics when stepping on irregular surfaces, and how wearable sensors can measure the changes. In line with this, the aim of our study was (i) to investigate ankle joint kinematics and kinetics while performing a step on an irregular surface and (ii) to check if the kinematic variations can be sufficiently captured by inertial measurement units (IMUs) mounted on the body. IMUs have been used for gait analysis in the detection of gait events (e.g. heel-strike and toe-off) [6, 7] and joint angle measurement [8], but most studies have been conducted only on treadmills or flat ground, and the feasibility of IMU sensing to capture the kinematic changes on irregular surface has not been thoroughly studied. To address this topic, an experimental study was conducted in which subjects walked on simplified irregular surfaces, in this case a small bump, and ankle joint mechanics as well as the measurements from body-mounted IMUs were investigated.
2 Methods Nine healthy male adults (age 29.1 ± 4.8 y; mass 76.8 ± 10.2 kg; height 176.3 ± 4.7 cm; mean ± SD) participated in this study. Participants were asked to walk on a straight flat walkway during five different conditions: four conditions involving bumps of two different heights (Fig. 1) and one condition without bumps (FLAT). In the conditions with bumps, participants stepped on bumps of different sizes with their right rearfoot (High-RF, Low-RF) or with their right forefoot (HighFF, Low-FF). Participants were asked to look straight in order not to know the exact bump location before stepping on it. IMU (VN-100, VectorNav) consisting of gyroscopes, accelerometers, and magnetometers (all 3-axes) were placed at the dorsum of the foot and the front of the shank (Fig. 1). Angular velocity and orientation (Euler angles calculated by IMU’s onboard extended Kalman Filter) were collected at a frequency of 200 Hz. Ankle joint angle was calculated by subtracting the sagittal angles of the two IMUs. Ground-truth lower limb kinematics was collected by an optical motion capture system (Vicon; 120 Hz), and ground reaction forces on the stride of interest were collected with a force plate (OR-6, AMTI; 1000 Hz) to calculate reference joint torques using inverse dynamics. Positive work was calculated as the time integral of positive power. Muscle activation on eight lower limb muscles (biceps femoris,
Biomechanical Analysis and Inertial Sensing of Ankle Joint …
345
Fig. 1 a Placement of IMU on foot and shank. b Experimental setup, dimension of bumps used for experiment (right top) and foot positions on each condition
rectus femoris, vastus medialis and lateralis, tibialis anterior, soleus, gastrocnemius lateral and medial) was also collected by surface electromyography (Telemyo, Noraxon; 1500 Hz).
3 Results The ankle angle during stance phase varied greatly over the five experimental conditions (Fig. 2a). Ankle torque plantarflexion onset was found to happen significantly later in the gait cycle for High-RF (31.0 ± 3.6 %) and Low-RF (27.0 ± 3.5 %) and earlier for High-FF (9.3 ± 2.6 %) and Low-FF (13.2 ± 2.1 %), compared to FLAT (19.7 ± 2.3 %; all p < 0.01 by paired t-test) (Fig. 2b). Similarly, onset of positive joint power was found to happen later for High-RF (57.4 ± 2.0 %) and Low-RF (56.1 ± 2.0 %) and earlier for High-FF (49.4 ± 3.7 %) compared to FLAT (52.2 ± 1.9 %; all p < 0.01) (Fig. 2c). Ankle positive work was lower in both HighRF (0.17 ± 0.03 J kg−1) and Low-RF (0.20 ± 0.04 J kg−1) compared with FLAT (0.26 ± 0.03 J kg−1; both p < 0.002). Mean muscle activation of lateral gastrocnemius and soleus was higher in High-FF compared with FLAT (52 and 29 % respectively, both p < 0.01) (Fig. 2d). IMU-based joint angles showed average root mean square error of 3.2° (HighRF: 3.1°, Low-RF: 3.2°, FLAT: 2.8°, Low-FF: 3.3°, High-FF: 3.5°) compared to ground-truth (dashed lines in Fig. 2a).
346
T. Miyatake et al.
Fig. 2 Averaged ankle joint kinematics and kinetics over the five experimental conditions: a Joint angle. Ground-truth measured by optical system (solid) and IMU-based (dashed). b Joint torque. c Joint power. Joint torque and joint power are normalized to the subject’s body weight. d Normalized muscle activation of lateral gastrocnemius, a plantarflexor muscle
4 Discussion and Conclusion The present findings suggest that humans adapt differently when stepping on an unanticipated bump, depending on which part of the foot contacts the bump. There were clear changes in the kinematics and kinetics across the different conditions as measured using motion capture and a force plate. It can be noted that the joint torque profile varied significantly over the five experimental conditions, as it is one of the most important measures for designing and controlling wearable devices assisting locomotion. In addition, it was shown that body-mounted IMUs have a
Biomechanical Analysis and Inertial Sensing of Ankle Joint …
347
potential to capture the changes in kinematics on uneven terrain. These results suggest the possibility of real-time wearable assistive devices that can adapt to irregular terrain by means of body-mounted IMU sensors. Future work will include applying these findings to the soft exosuits [2, 3], making it highly adaptable to various ground conditions. Acknowledgments The authors would like to thank Philippe Malcolm and Jozefien Speeckaert for their contribution to this work.
References 1. Malcolm, P., Derave, W., Galle, S., De Clercq, D.: A simple exoskeleton that assists plantarflexion can reduce the metabolic cost of human walking. PLoS ONE 8(2), e56137 (2013) 2. Panizzolo, F.A., Galiana, I., Asbeck, A., Siviy, C., Schmidt, K., Holt, K.G., Walsh, C.J.: A biologically-inspired multi-joint soft exosuit that can reduce the energy cost of loaded walking. J. Neuroeng. Rehabil. (In press) 3. Asbeck, A.T., De Rossi, S.M.M., Holt, K.G., Walsh, C.J.: A Biologically inspired soft exosuit for walking assistance. Int. J. Robot. Res. 34(6), 744–762 (2013) 4. Voloshina, A.S., Kuo, A.D., Daley, M.A., Ferris, D.P.: Biomechanics and energetics of walking on uneven terrain. J. Exp. Biol. 216(21), 3963–3970 (2013) 5. Wade, C., Redfern, M.S., Andres, R.O., Breloff, S.P.: Joint kinetics and muscle activity while walking on ballast. Hum. Factors 52(5), 560–573 (2010) 6. Rueterbories, J., Spaich, E.G., Larsen, B., Andersen, O.K.: Methods for gait event detection and analysis in ambulatory systems. Med. Eng. Phys. 32(6), 545–552 (2010) 7. Mariani, B., Rouhani, H., Crevoisier, X., Aminian, K.: Quantitative estimation of foot-flat and stance phase of gait using foot-worn inertial sensors. Gait Posture 37(2), 229–234 (2013) 8. Seel, T., Raisch, J., Schauer, T.: IMU-Based Joint Angle Measurement for Gait Analysis. Sensors 14(4), 6891–6909 (2014)
A Novel Approach to Increase Upper Extremity Active Range of Motion for Individuals with Duchenne Muscular Dystrophy Using Admittance Control: A Preliminary Study Madeline C. Corrigan and Richard A. Foulds
Abstract Progressive muscle weakness characteristic of Duchenne muscular dystrophy (DMD) results in loss of upper extremity active range of motion (AROM) despite residual muscle strength that is insufficient to overcome gravity. Admittance control is well suited for use by individuals with DMD as it allows for the utilization of residual muscle strength to intuitively control the motion of a powerful robot without requiring strength to overcome gravity and the friction and inertia of the robot. This study examined the feasibility of using the HapticMASTER, an admittance control motorized arm support, to increase the upper extremity AROM of individuals with DMD to a greater degree than that provided by the Armon Edero, a commercially available passive arm support. The results demonstrate that the HapticMASTER robot significantly increased the reachable surface area scores compared to the Armon Edero passive arm support (paired-samples t-test, t(5) = 3.984, p = 0.010, Cohen’s d = 1.6).
1 Introduction Duchenne muscular dystrophy (DMD), a neuromuscular disease with a prevalence of 1 in 3500 male births, results in progressive muscle weakness causing loss of independence and imposing the demands of costly and intrusive assistive support
This research is funded by NJIT/Kessler Foundation Rehabilitation Engineering Research Center on Wearable Robots for Independent Living (NIDILRR-HHS 90RE5021) and Gustavus and Louise Pfeiffer Research Foundation. M.C. Corrigan (&) R.A. Foulds New Jersey Institute of Technology, Newark, NJ, USA e-mail:
[email protected] R.A. Foulds e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_57
349
350
M.C. Corrigan and R.A. Foulds
and personal care for daily living tasks [1]. Upper extremity function begins to decline while ambulation is still possible and gradually progresses with time in a proximal to distal gradient, playing a prominent role in loss of independence [2]. Importantly, upper extremity functional limitations exist despite residual muscle strength because remaining muscle strength is not sufficient to lift the arms against gravity and therefore cannot be utilized by the individual [3]. Presently, there exist a number of commercially available assistive devices aimed at augmenting upper extremity functional deficit. These devices are not widely used by individuals with DMD and have been largely unsuccessful in delivering the independence they seek to provide. Passive orthoses, the most common of these commercially available assistive devices, increase AROM by reducing friction and balancing the user’s arm against gravity using a spring or rubber bands. Limitations of passive arm supports include imperfect gravity compensation and the requirement of sufficient muscle strength to overcome inertia, which render these devices ineffective for those in the later stages of functional loss [4]. Admittance control is an inherently safe and intuitive robotic control paradigm that maps the user’s applied force to the motion of a robot. The use of admittance control provides a means to balance the arm against gravity more precisely and to minimize friction and inertia, thereby decreasing the overall force required by the user to generate a movement compared to that required by a passive arm support [5]. The objective of this study is to investigate the degree of upper extremity AROM associated with use of an admittance control motorized arm support compared to that associated with the use of a passive arm support by individuals with DMD.
2 Materials and Methods The upper extremity AROM of 6 individuals with DMD was evaluated while unsupported, while supported by the Armon Edero passive arm support, and while supported by the HapticMASTER admittance control robot. Figure 1 shows the Armon Edero arm support and the HapticMASTER robot. The maximum workspace of both arm supports is about 0.08 m3, with the HapticMASTER having a slightly smaller workspace compared to the Armon Edero. Inclusion criteria required individuals to be nonambulatory with limited upper extremity function (a Brooke scale score of 4 or greater). The reachable workspace protocol was administered for each of the arm support conditions and the reachable surface area scores were used as the primary outcome measure to evaluate AROM [6]. This study was approved by NJIT’s Institutional Review Board.
A Novel Approach to Increase Upper Extremity Active Range …
Fig. 1 a The Armon Edero passive arm support and b the HapticMASTER robot
Fig. 2 The reachable surface area (RSA) scores for a unsupported movements, b Armon Edero supported movements, and c HapticMASTER supported movements for one subject
351
352
M.C. Corrigan and R.A. Foulds
3 Results Figure 2 shows an example of the reachable workspace evaluation output for one subject for unsupported movements, Armon Edero supported movements, and HapticMASTER supported movements. Figure 3 shows the mean reachable surface area scores for all 6 subjects for Armon Edero and HapticMASTER supported movements. The HapticMASTER robot significantly increased the reachable surface area scores compared to the Armon Edero passive arm support (paired-samples t-test, t(5) = 3.984, p = 0.010, Cohen’s d = 1.6). An a priori power analysis, conducted using G*Power 3.1 software, suggested that a sample size of n = 6 would be sufficient to obtain a minimum of 95 % power (a = 0.05, paired-samples t-test) to detect an effect, given an effect size of Cohen’s d = 1.6. Five out of 6 subjects showed increased vertical range of motion for HapticMASTER supported movements compared to Armon Edero supported movements. All subjects reported that less exertion was required to generate the reachable workspace evaluation movements while supported by the HapticMASTER robot compared to the Armon Edero passive arm support.
Fig. 3 Mean reachable surface area scores for Armon Edero and for HapticMASTER supported movements. Error bars show SEM
A Novel Approach to Increase Upper Extremity Active Range …
353
4 Discussion and Conclusion The results of this study support the hypothesis that an admittance control arm support will provide individuals with DMD greater AROM than that provided by a passive arm support. Ongoing work involves the development of a wheelchairmountable admittance control arm support that aims to increase independence in activities of daily living for individuals with DMD. Acknowledgments The authors would like to acknowledge the developers of the reachable workspace evaluation software (G. Kurillo and J. Han, et al.).
References 1. Jung, I.Y., Chae, J.H., Park, S.K., Kim, J.H., Kim, J.Y., Kim, S.J., Bang, M.S.: The correlation analysis of functional factors and age with duchenne muscular dystrophy. Ann. Rehabil. Med. 36, 22–32 (2012) 2. Bartels, B., Pangelila, R.F., Bergen, M.P., Cobben, N.A.M., Stam, H.J., Roebroeck, M.E.: Upper limb function in adults with duchenne muscular dystrophy. J. Rehabil. Med. 43, 770– 775 (2011) 3. Mazzone, E.S., Vasco, G., Palermo, C., Bianco, F., Galluccio, C., Ricotti, V., Castronovo, A. D., DiMAuro, M.S., Pane, M., Mayhew, A., Mercuri, E.: A critical review of functional assessment tools for upper limbs in Duchenne muscular dystrophy. Dev. Med. Child Neurol. I 54, 879–885 (2012) 4. Rahman, R., Sample, W., Seliktar, R., Alexander, M., Scavina, M.: A body-powered functional upper limb orthosis. J. Rehabil. Res. Dev. 37(6) (2000) 5. Corrigan, M., Foulds, R.: Admittance control of the intelligent assistive robotic manipulator for individuals with duchenne muscular dystrophy: a proof-of-concept design. J. Rehabil. Rob. 3, 1–5 (2015) 6. Han, J.J., Kurillo, G., Abresch, R.T., Nicorici, A., Bajcsy, R.: Validity, reliability, and sensitivity of a 3D vision sensor-based upper extremity reachable workspace evaluation in neuromuscular diseases. PLOS Currents Muscular Dystrophy (2013)
Modulation of Knee Range of Motion and Time to Rest in Cerebral Palsy Using Two Forms of Mechanical Stimulation Peter A. Michael, Ghaith J. Androwis and Richard A. Foulds
Abstract Spasticity and dystonia are challenging motor impairments that may interfere with the use of exoskeleton-based therapy. We suggest that two mechanical stimulation techniques that target and remediate these manifestations in cerebral palsy, will allow exoskeletons become a much more feasible rehabilitation technique. This will improve function and, importantly, safety. Our previous studies have shown the positive outcomes of vestibular stimulation, and published literature proposes the possible advantages of whole body vibration. Our current approach utilizes both techniques to define a rehabilitation method specific to the subject’s diagnosis of spasticity and dystonia. Our recent pilot data shows great potential in temporarily eliminating/reducing both spasticity and dystonia in a subject with CP. Results suggest a reduction in tone and possible improvements to mobility after a single session of stimulation. Therefore, this paper serves to propose the use of this approach to enable and enhance the benefits of robotic therapy.
This work is supported by the NJIT/Kessler Rehabilitation Engineering Research Center on Wearable Robots for Independent Living, (NIDILRR-HSS grant 90RE5921). P.A. Michael (&) G.J. Androwis R.A. Foulds New Jersey Institute of Technology, Newark, NJ, USA e-mail:
[email protected] G.J. Androwis e-mail:
[email protected] R.A. Foulds e-mail:
[email protected] G.J. Androwis Kessler Foundation, West Orange, NJ, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_58
355
356
P.A. Michael et al.
1 Introduction The recent increase in exoskeleton use and development encourages the exploration of methods to enable access by users who otherwise would not have be able to take advantage of this technology. Robot-assisted rehabilitation for individuals who exhibit high muscle tone is extremely problematic for the device and potentially dangerous for the user. Studies often exclude subjects with spasticity to avoid complication [1–3]. By reducing the excessive tone, not only will the therapy function as intended, but the user is less likely to sustain injury or experience fatigue [1]. Neurological disabilities such as cerebral palsy (CP) are usually accompanied by loss of motor function, spasticity, and/or dystonia [4, 5]. Cerebral palsy poses a challenge in that subjects not only have excessive tone due to spasticity, but may also exhibit dystonia in the form of unwanted co-contraction of muscles. Many individuals with CP actually have a combination of spasticity and dystonia [4], but the distinction is often clinically blurred between the two classifications. Spasticity is defined as a hypersensitivity to joint movement above a certain velocity threshold. Dystonia on the other hand is sudden and uncontrollable co-contraction of muscles that usually leads to excessively stiff joints and bad posture [6]. Previous work from our laboratory shows that vestibular stimulation of the otolith organs is able to significantly reduce muscle tone in individuals who are purely spastic [7]. This effect lasts for 20 min, during which robotic therapy may be particularly effective. However, we discovered that if any dystonia were present, our stimulation had a much less dramatic effect since motor function was dominated by the dystonia [8], which caused joints to appear locked. Dystonia can be triggered by a startle or sudden movement. Some of our dystonic subjects have actually exhibited more dystonic symptoms after receiving vestibular stimulation [8]. Figure 1 utilizes the Pendulum Knee Drop (PKD) to demonstrate the leg trajectory of a spastic/dystonic subject before and after vestibular intervention. Fig. 1 Representative shank trajectory of a dystonic subject before and after vestibular stimulation. In dystonia, the limb appears to ‘hang’ at a non-zero angle, and slowly relax to a vertical (0 rad) position. The relaxation (time to rest) takes longer after vestibular stimulation
Modulation of Knee Range of Motion and Time …
357
It is important to note the amount of time it takes for the leg to relax and return to zero (vertical). Our current recent case study suggests that a different form of stimulation may temporarily overcome dystonic co-contraction. Several studies have used whole body vibration to reducing muscle tone, referred to as spasticity [5, 9]. However, our review of the data led us to believe that subjects in these studies also exhibit a high degree of dystonia. We have conducted a single subject trial to explore the use of whole body vibration followed by vestibular stimulation to remediate bit dystonia and spasticity [10].
2 Materials and Methods 2.1
Whole Body Vibration (WBV)
Whole body vibration consists of applying high frequency low amplitude vertical oscillations to the feet of a standing subject. The DKN XG10 vibrating platform was used at a frequency of 35 Hz. The subjects was instructed to assume a shallow squat position to target the legs and dampen the vibrations to the torso and head. The subject received 1 min stimulation bursts with seated rest in between; total stimulation time was 5 min.
2.2
Vestibular Stimulation (VS)
The vestibular stimulation apparatus used in this study was developed and tested extensively in our lab [7]. It is an IRB-approved chair that is translated vertically by four pneumatic actuators. Valve timing is programed in MATLAB and is used to vary the speed and amplitude of movement. Our previous work has optimized the parameters to be 7.5 cm excursion at 2 Hz. This phase of stimulation lasts for 15 min.
2.3
Assessment Technique
Quantitative measurements are based on the PKD test which is administered in three sets throughout the stimulation session; baseline, post WBV and post VS. An Ascension Technologies TrakSTAR was used to collect position data of the subject’s shank. Data were recorded with a sampling rate of 100 Hz.
358
P.A. Michael et al.
Fig. 2 Unfiltered PKD trajectory for testing phase: (i) baseline, (ii) post whole body vibration and (iii) post vestibular stimulation Table 1 Knee range of motion and time to rest of the shank (A) (B)
Range of motion (% of lift angle) Time to rest (s)
Pre
Post WBV
Post VS
74.32 5.90
86.50 3.62
101.82 3.25
3 Results See Figure 2 and Table 1.
4 Discussion The above data are promising in remediating tone in preparation for robotic therapy. The baseline data is clearly dominated by the subject’s dystonia, which is nearly eliminated by the application of whole body vibration, revealing the underlying spasticity. Following subsequent vestibular stimulation, the spasticity is also reduced as indicated by improvement in range of motion, decreased hyper-reflexia and reduced joint stiffness. The points marked “A” signify the range of motion of the initial swing. A larger swing is indicative of reduced tone. The locations marked “B” are the time for the shank to come to rest in the vertical position. These results highlight the benefits of the combination of stimulation techniques, as well as suggest their beneficial use prior to robotic therapy. By increasing the compliance of the body, the individual is less likely to sustain injuries from interacting with a robot.
Modulation of Knee Range of Motion and Time …
359
5 Conclusion Utilizing our ability to customize a stimulation routine based on the subject’s CP classification (i.e. spastic, dystonic or spastic/dystonic) gives us a unique advantage in modulating their tone and co-contraction. The noted retention is arguably the most crucial outcome, whereby the subject is now more likely a candidate to operate and benefit from an exoskeleton. A larger study will be conducted to verify these outcomes.
References 1. Evans, N., et al.: Acute cardiorespiratory and metabolic responses during exoskeleton-assisted walking overground among persons with chronic spinal cord injury. Top. Spinal Cord Inj. Rehabil. 21(2), 122–132 (2015) 2. Frisoli, A., et al.: Positive effects of robotic exoskeleton training of upper limb reaching movements after stroke. J. NeuroEng. Rehabil. 9(1), 1 (2012) 3. Louie, D., et al.: Gait speed using powered robotic exoskeletons after spinal cord injury: a systematic review and correlational study. J. NeuroEng. Rehabil. 12(1), 1 (2015) 4. O’Dwyer, N., Ada, L.: Reflex hyperexcitability and muscle contraction in relation to spastic hypertonia. Curr. Opin. Neurol. 9(6), 451–455 (1996) 5. Ahlborg, L., et al.: Whole-body vibration training compared with resistance training: effect on spasticity, muscle strength and motor performance in adults with cerebral palsy. J. Rehabil. Med. Off. J. Eur. Board Phys. Rehabil. Med. 38(5), 302–308 (2006) 6. Sanger, T., et al.: Classification and definition of disorders causing hypertonia in childhood. Pediatrics 111(1), e89–e97 (2003) 7. Androwis, G., et al.: Alterations of neuromuscular signals as a result of vestibular stimulation. In: 6th International IEEE/EMBS Conference on Neural Engineering (NER), pp. 1238–1241. IEEE (2013) 8. Androwis, G., et al.: Spasticity and dystonia differentiated via the equilibrium point hypothesis. In: 40th Annual Northeast Bioengineering Conference. IEEE (2014) 9. van Nes, I., et al.: Short-term effects of whole-body vibration on postural control in unilateral chronic stroke patients: preliminary evidence. Am. J. Phys. Med. Rehabil. 83, 867–873 (2004) 10. Michael, P., et al.: Non-invasive mechanical stimulation for reduction of dystonia and spasticity in cerebral palsy: a case study. In: EMBC. IEEE (2016)
Training Response to Longitudinal Powered Exoskeleton Training for SCI A. Ramanujam, A. Spungen, P. Asselin, E. Garbarini, J. Augustine, S. Canton, P. Barrance and G.F. Forrest
Abstract The goal of this research is to assess how powered exoskeleton-training for 5 h per week over 20 weeks can change gait parameters to increase walking speed for chronic SCI. Gait parameters include Center of Mass (CoM) excursions, walking velocity, initial double stance time (IDS), single stance time (SS), terminal double stance time (TDS), swing time (SW), and spatial parameters such as step length, step width and stride length. Exoskeleton training had a significant effect on walking velocity due to specific temporal spatial gait parameters (IDS, TDS, Step and Stride Length) and increased stability (CoM).
1 Introduction Spinal cord injury (SCI) can be characterized by the degree of loss of motor and sensory function an individual sustains that limits or restricts their walking ability overground. Powered exoskeletons are robotic devices that are intended for rehabilitation, mobility and walking overground in those with SCI who have limited or no ability to walk. The powered exoskeleton is also envisioned as a device that will potentially improve mobility and independence in the home and community. Currently, there are publications emerging that reflect training adaption in walking velocity as an indicator of improved mobility within the device [1, 2]. However, the underlying motor control strategies for gains in walking speed have not been presented for either spinal cord populations or ablebodied (AB) controls. Understanding acute and chronic training adaptions for gains in mobility (such as walking velocity) using different devices is critical to development of powered robotic exoskeletons, rehabilitation, motor control and neuromuscular control adaption. The objective of A. Ramanujam E. Garbarini J. Augustine S. Canton P. Barrance G.F. Forrest (&) Kessler Foundation, West Orange, NJ, USA e-mail:
[email protected] A. Spungen P. Asselin National Center of Excellence for the Medical Consequences of SCI, James J. Peters VAMC, Bronx, NY 10468, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_59
361
362
A. Ramanujam et al.
this paper is to explore how individuals with motor incomplete SCI who train for 100 h adapt in the device to increase walking velocity. A secondary analysis explores how increased voluntary control can affect walking velocity in ablebodied controls.
2 Materials and Methods 2.1
Experimental Test Conditions
Five participants with SCI and two AB controls were recruited for the study (Table 1), of which 3 SCI and 2 AB trained using EksoGT™ (Ekso Bionics, Richmond, CA) and 2 SCI participants trained using ReWalk™ (ReWalk Robotics, Inc., Marlborough, MA). All participants in the EksoGT™ were tested using the Max Assist condition. For each group a standardized training paradigm was implemented. The participants were tested pre- and post-training under set walking conditions with their respective exoskeleton. The AB control participants were asked to ambulate under 5 experimental conditions in the same session with and without the Ekso™. For conditions one, two, and three, the AB group walked overground without the Ekso™: at a SLOW, SELF-SELECTED, and FAST pace. For conditions four and five, the AB control group ambulated with the Ekso™ with minimal or no volitional control (EXO-WOVC) and with increased active volitional control of the lower limbs (EXO-WVC). All trials captured 6–8 steps while walking along a 10-m walkway and were repeated 2–4 times as needed within a one hour session. For all individuals, walking with the exoskeleton occurred with close supervision (no physical contact) by a skilled physical therapist. All procedures performed were approved by the Institutional Review Board with prior informed consent of participants.
Table 1 Demographics Group/participant
Age (yrs) SCI group RW1 29 RW2 31 EK1 54 EK2 42 EK3 52 AB group AB1 26 AB2 26 *TPI time post injury; LOI level of
TPI* (mths) 21.2 36.7 112.8 13.9 32.2
injury
LOI*/ AIS T7/C T4/C C6/C T12/C C2/C
Training (hours) 100 80 60 100 100 10 10
Testing conditions ReWalk™ EksoGT™ Max Assist With & without EksoGT™
Training Response to Longitudinal Powered Exoskeleton …
2.2
363
Data Collection and Analysis
Kinematic data were collected at 60 Hz using a Motion Analysis Corporation (Santa Rosa, CA) motion capture system. Reflective markers were placed on anatomical landmarks on the whole body according to the Helen-Hayes marker sets using adhesive Velcro tape. Kinematic data were filtered (2nd order low pass Butterworth filter, cut-off frequency of 6 Hz) and normalized to percent gait cycle (foot-strike to foot-strike). Temporal measures analyzed included time (in secs and %) during Initial Double Stance (IDS), Single Stance (SS), Terminal Double Stance (TDS), and Swing (SW). Spatial measures comprised of Step Length (StepL), Step Width (StepW), and Stride Length (StrideL). Kinematic variables included Anterior-Posterior CoM (normalized to %StepL), Medial-Lateral CoM (normalized to %Base of Support) and walking velocity. Primary statistical analyses included a one-way ANCOVA with device as the covariate to evaluate training responses for velocity, temporal, and spatial variables.
3 Results 3.1
Demographics
Participant demographics are presented in Table 1. The maximum training time for all SCI participants was 100 h with mean number of steps = 170,000 (±79,708).
3.2
Spatial Temporal Parameters
Temporal parameters showed a significant decrease in IDS and TDS (p < 0.05) post-training, Fig. 1a. Spatial parameters showed a significant increase in StepL and StrideL (p < 0.05), Fig. 1b. After training, there was also a significant increase in walking velocity (p < 0.05), Fig. 2a. For percent of gait cycle, there was a significant greater percent swing time for ReWalk™. The ReWalk™ post-training SW is similar to AB control in EXO-WVC, Fig. 2b.
3.3
Correlation of Temporal-Spatial Measures to Velocity
Correlation data showed a strong significant positive correlation between walking velocity and StepL (q = 0.9010, p < 0.001) and StrideL (q = 0.9027, p < 0.001) and a strong negative correlation between walking velocity and IDS (q = −0.9644, p < 0.001) and TDS (q = −0.9604, p < 0.001).
364
A. Ramanujam et al.
Fig. 1 a Temporal (in seconds) and b spatial measures (in meters) with EksoGT™ and ReWalk™
Fig. 2 a (left). Walking velocity and b (right). Temporal distribution as % Gait cycle with EksoGT™ and ReWalk™
3.4
CoM
Post-training in the EksoGT™, the MLCoM excursion increased for all SCI subjects and was more centered towards midline of 50 % (particularly EK3, Fig. 3). For AB controls the MLCoM excursion decreased with volitional control (EXO-WVC) in the EksoGT™ and with increasing speed.
3.5
Able Bodied with EksoGT™
For AB controls the walking velocity for EXO-WVC was significantly greater than EXO-WOVC (p < 0.05). Temporal parameters (IDS, TDS) significantly decreased while spatial parameters for StrideL and StepL increased.
Training Response to Longitudinal Powered Exoskeleton …
365
Fig. 3 Medial-lateral (X-axis) versus anterior-posterior (AP) (Y-axis) normalized CoM excursions with respect to left leg
4 Discussion After 100 h of training, walking velocity for both devices positively correlated with StrideL/StepL and negatively correlated with IDS/TDS. Greater MLCoM excursion (in EksoGT™) combined with reduced IDS/TDS time post-training provides evidence of a more efficient weight transfer strategy onto the stance leg thereby initiating swing on the opposite leg. The increase in velocity for SCI can be attributed to the changes in these characteristics as well as the improved dynamic stability (measured by the MLCoM). For ReWalk™ the significant increase in %SW combined with increase in StepL contributed towards greater walking velocity. Notably, after training there was a significant increase in walking velocity for EksoGT™, even though absolute walking velocity was greater for ReWalk™. For AB controls in EXO-WVC, there was a decrease in IDS/TDS and a significant increase in StepL/StrideL that resulted in increased walking velocity. The increase in active voluntary control also reflected an increase in muscle activation of the lower limbs as reported elsewhere [3].
5 Conclusion This study provides a potential rationale for the increase in robotic exoskeleton velocity for SCI who have completed longitudinal training and AB controls who train using an increase in active voluntary control. Acknowledgments CSCR13IRG013 New Jersey Commission on Spinal Cord Research.
366
A. Ramanujam et al.
References 1. Esquenazi, A.: The ReWalk powered exoskeleton to restore ambulatory function to individuals with thoracic-level motor-complete spinal cord injury. Am. J. Phys. Med. Rehabil. 91(11), 911–921 (2013) 2. Kressler, J.: Understanding therapeutic benefits of overground bionic ambulation: exploratory case series in persons with chronic, complete spinal cord injury. Arch. Phys. Med. Rehabil. 95(10), 1878 (2014) 3. Ramanujam, A., Cirnigliaro, C., Pilkar, R., Forrest, G.F.: Muscle activation during robotic powered exoskeleton assisted walking: case series. Front. Neurosci. (submitted for publication)
Adaptive Classification of Arbitrary Activities Through Hidden Markov Modeling with Automated Optimal Initialization Chris T.M. Baten, Thijs Tromper and Leonie Zeune
Abstract An adaptive method for classification of arbitrary activities is presented that assesses continuously the activity in which a subject is engaged, thus providing contextual information facilitating the interpretation of any continuous data gathered from an (unsupervised) applied wearable robotics device and its bearer. Specifically the effect of a novel adaptive and fully automated initialization method using Potts energy functionals is discussed. Exemplary data suggests that this method very likely improves overall performance equally or better than more traditional methods. This includes state of the art methods based on segmental k-means initialization that do require substantial recurrent manual intervention.
1 Introduction To interpret all detailed kinematic, kinetic and muscle activation and system control data gathered from a wearable robotic device and its wearer in a real life unsupervised application, contextual information is very valuable. Specifically there is a need for continuous contextual information on the activity in which the subject is engaged. The ideal method for contextual data assessment recognizes any activity in which a person is engaged in from data that is already gathered by the wearable robotic device. As there are many activities in which a subject could be engaged and given the many specific ways in which a given activity can be performed by a subject, a method that can adapt to the situation is required.
C.T.M. Baten (&) T. Tromper ‘Ambulatory Analysis of 3D Human Movement’ Group of Roessingh Research and Development (RRD), Enschede, Netherlands e-mail:
[email protected] L. Zeune University of Twente, Enschede, Netherlands © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_60
367
368
C.T.M. Baten et al.
One such an adaptive method is presented earlier by authors and is built around Hidden Markov Modeling, This method is capable of adapting to the activity classification challenge at hand by learning to recognize a given set of activities from a given set of relevant data channels performed in a specific way [3]. This paper discusses how to improve application of this method by optimizing the initialization of the used HMMs with a newly developed adaptive method.
2 Methods A. Data preparation From 3 IMMUs (Xsens Mtw) placed on pelvis, lumbar back section and sternum only 3D linear acceleration and 3D angular velocity data were used. To make data more independent from the occasion, subject and sensor placement, all sensor data was transformed to body segment through a body segment calibration procedure [1]. Transformation of the data to a more mutually independent vector base was done by applying Principal Component Analysis (PCA). B. HMM initialization 3 different types of HMM initialization were tested. ‘Flat start’ (or no) initialization, an initialization using k-means clustering and an initialization through a custom algorithm developed using Potts energy functionals: 1. ‘Flat start initialization’ (standard approach) • All activities are represented by 8 states representing segments of equal duration • Each state has a normal data distribution with the same overall mean and variance for the whole activity as initial mean and variance 2. ‘Segmental K-means initialization’ • Each activity has a manually chosen number of states representing segments of equal duration • Each state has an estimate for initial mean and average determined through a K-means clustering procedure [2]. 3. ‘Potts Initialization’ • For each activity the number of states and the duration of each state was estimated by a variation on Potts algorithm and each state was then given the mean and variance of the data in the interval of each state The classical Potts problem to be solved to find the optimal number and duration of states is formulated as:
Adaptive Classification of Arbitrary Activities Through Hidden Markov Modeling…
369
Pc ¼ ckrtk0 þ kt f k22 ! min: c2R where f(t) = u(t) + noise(t) is the measured signal where u(t) is the state function. Here a successful optimization run delivers u(t) and the number of jump discontinuities in it: the initial number of states. Gamma controls the sensitivity of this process in adding extra states. The Hidden Markov Models for activity k = (A, B, p) were optimized in a training sequence with a the Baum-Welch algorithm [3], i.e. solving: ki ¼ arg maxðki ÞðPðActivityi jki ÞÞ Here A is a transition probability matrix, B the signal probability distribution matrix and p the initial state distribution.
3 Results Figure 1 shows the HMM training results for the 3 methods of initialization applied to an artificial signal. It is clear that flat initialization performs badly and the other 2 perform quite good, where the Potts has a more correct segment start, end and duration estimate. In another example Fig. 2 shows in the left graph the original set of the trunk acceleration z-channel instances for one activity and their average (Black). In the right graph the mean and the resulting state function u(t) with estimated state segment durations, means and variances. Note the different durations of the state segments. Finally Fig. 3 shows in the top row classification results for a set of typical activities for an order picker for applying the full method with all three initialization methods. Clearly the K-means method performed better than the flat initialization method. The Potts method failed badly. It turned out that this was caused by the
Fig. 1 Results after HMM training for three methods of initialization applied on an artificial one dimensional signal with 4 states with duration average 25 and variance of 1 and amplitude average (6, 4, 6, 4), with variance (1, 1, 1, 1). Left Flat, middle K-means and right Potts initialization
370
C.T.M. Baten et al.
Fig. 2 Results of Potts method of initialization applied on actual thorax z channel acceleration signal of a squat lift activity. Left Original signal, right depiction of states with duration, amplitude mean and variance as estimated by the Potts initialization method. (arbitrary units on both axes)
Fig. 3 Results when applying to set of activities. Top classification results for all three initialization methods, bottom classification table with predicted versus original activities segmental k-means initialization
HMM software not being adapted for the new nuance in initialization, introduced by the Potts method, of state segments possibly having different initial lengths. At the time of writing this abstract it is not clear how the Potts method will perform when this ‘software bug’ is corrected.
4 Discussion and Conclusion Although statistically more powerful analysis results are not yet available, the exemplary data presented in this paper strongly suggests that optimizing the HMM initialization with the k-means method improves the systems classification capabilities. As the Potts method performed n better than the k-means method in estimating initial state parameters it is expected that this method will perform similar or better in classification, specifically in situations of states with different segment durations.
Adaptive Classification of Arbitrary Activities Through Hidden Markov Modeling…
371
If this indeed proves true the fully automated Potts method is strongly preferred over the k-means method that requires manual human intervention for each activity to be trained every time it is applied.
References 1. Baten, C.T.M., Luinge, H.J., Moerkerk, Hv: Estimating body segment orientation applying inertial sensing. Neural Syst. Rehabil. Eng. 15(3), 469–471 (2000) 2. Juang, B.-H., Rabiner, L.R.: The segmental K-means algorithm for estimating parameters of hidden Markov models. IEEE Trans. Acoust. Speech Signal Process. 38.9, 1639–1641 (1990) 3. Wassink, R.G.V., Baten, C.T.M., Veltink, P.H.: Classifying human lifting activities automatically by applying hidden markov modeling technology. J. Biomech. 40, S428 (2007)
Design and Motion Analysis of a Wearable and Portable Hand Exoskeleton Matteo Bianchi, Francesco Fanelli, Roberto Conti, Lapo Governi, Enrico Meli, Alessandro Ridolfi, Andrea Rindi, Federica Vannetti and Benedetto Allotta
Abstract The design of an aid for the hand function based on exoskeleton technologies for patients who have lost or injured hand skills, e.g. because of neuromuscular or aging diseases, is one of the most influential challenge in modern robotics to assure them an independent and healthy life. This research activity is focused on the design and development of a low-cost Hand Exoskeleton System (HES) for supporting patients affected by hand opening disabilities during the Activities of Daily Living (ADLs). In addition, the device, able to exert suitable forces on the hand, can be used during the rehabilitative sessions to implement specific tasks useful to restore the dexterity of the user’s hand. The validating and testing phase, conducted in collaboration with the Don Carlo Gnocchi Foundation, showed satisfying results both in terms of portability and wearability which are fundamental requirements for assistance during the Activities of Daily Living (ADLs) and for rehabilitation of people with hand impairments.
1 Introduction Nowadays, the first cause of adult disability in Europe is the Cerebral Vascular Accident (CVA) [1]. Up to 80 % of post-stroke patients suffer from hemiparesis of the upper arm. Moreover, the number of patients with a disability in the upper part of the body is rising together with the number of elderly [2], who will constitute, in a A special thanks goes to the Don Carlo Gnocchi Foundation, where the validation phase of the research activity has been carried out and the testing phase is currently ongoing. M. Bianchi (B) · F. Fanelli · L. Governi · E. Meli · A. Ridolfi · A. Rindi · B. Allotta Department of Industrial Engineering, University of Florence, 50139 Florence, Italy e-mail:
[email protected] R. Conti IUVO S.r.l. Company, Pontedera, Italy F. Vannetti Don Carlo Gnocchi Foundation, Florence, Italy © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_61
373
374
M. Bianchi et al.
Fig. 1 The low-cost hand exoskeleton developed by the MDM Lab
few years, more than the 30 % of the total population. Post stroke survivors, genetic disease patients and elderly with hand disease need timely and persistent rehabilitative training to regain previously dexterity and an assistance during the Activities of Daily Living (ADLs). An effective and specific solution for the aforementioned patients may be provided by the use of robotic devices. In fact, robotic systems allow to provide prolonged and higher-intensity rehabilitation treatments, with a reduction of costs and burden for the therapists [3]. Since such devices [4] are designed to be used during rehabilitative sessions, their portability requirement is not mandatory, as they are employed in hospitals, rehabilitation centers or at home in a specific location dedicated to their utilization [5, 6]. In case the hand functions may not be totally restored even after an intense rehabilitation process, hand exoskeletons can be used to support the user in ADLs assisting the hand performance by amplifying the hand gripping force [7] or automating the motion [8]. In this case, the wearability of the device is essential as it must be worn for a long time during the day. Basing on strict requirements of wearability, portability, cheapness and modularity, the researchers of the Mechatronics and Dynamic Modeling Laboratory (MDM Lab) of the Department of Industrial Engineering of the University of Florence (DIEF) have developed and tested an assistive and rehabilitative device for hand disabilities [9] (Fig. 1). This robotic orthosis is designed to be a low-cost and portable hand exoskeleton for patient’s assistance and for rehabilitative use. If compared to the current state of the art [10, 11], the proposed novel mechanism results easily adaptable to different hand sizes by modifying only a few geometrical parameters. In addition, the device is specific for the hand anatomy of the user and is able to closely follow the trajectories of all the long fingers.
2 Design Phase The design of the mechanism and the choice of the actuation system have been performed in parallel with the development of the 3D multibody model of the exoskeleton (which includes the hand model): this model-based approach allows the authors to optimize the wearability of the system.
Design and Motion Analysis of a Wearable and Portable Hand Exoskeleton
375
Fig. 2 Exoskeleton mechanism parts
Fig. 3 Closed-loop angular control of the device
A. Mechanical Part Starting from the 3D multibody model, a novel exoskeleton kinematic chain, based on a 1 Degree Of Freedom (DOF) mechanism (Fig. 2), has been designed to reproduce the desired trajectories of the fingers. The hand exoskeleton mechanism has then been optimized to reduce both the lateral and the height encumbrances and, finally, all the parts have been built and assembled. The device is mainly realized in Acrylonitrile Butadiene Styrene (ABS) and all the structural components have been built by means the 3D printer of the MDM Lab. B. Electronics The electronics consists of four independent servomotors (one for each finger, placed on the back of the hand), a control unit and a battery pack. The characteristics of the servomotor, in terms of torque, size and weight, have been chosen according to the results of numerical simulations. Using a 15-bit magnetic encoder placed on the joint A of the Fig. 2, the value of the angle α2 , which identifies the single DOF of the mechanism, can be measured. Using this angular information, a closedloop angular control of the mechanism is realized by means of an Arduino singleboard microcontroller (see Fig. 3 for the control loops architecture). The opening and closing gestures are commanded using buttons.
3 Results This phase of the research activity consists in the evaluation of the transparency of the HES (the capability of the device in reproducing the real trajectories of the hand phalanges) for a particular test case (a patient suffering from Spinal Muscular
376
M. Bianchi et al.
Fig. 4 Comparison between the finger (blue) and exoskeleton (green) trajectories
Atrophy) and it has been executed using a Motion Capture (MoCap) system. The motion analysis was carried out by means of the optoelectronic system made up of 3 cameras. The system (BTS SMART-Suite Motion Capture System by BTS Bioengineering S.p.A., Milano, Italy) is able to automatically record 3D trajectories of passive markers Fig. 4 reports the comparison between the trajectories of the contact point between the hand and the end effector of the device for the index finger. In particular, the blue line is the trajectory of the index finger acquired through the MoCap system, while the green one is the trajectory of the mechanism end effector (also this trajectory is acquired by means the SMART-DX system) when the device is not worn by the user. The kinematic behavior of the device is very close to the natural one; in this case, the maximum error is 0.57 mm (red line in Fig. 4).
4 Conclusion In this work the development and the testing phase of a portable HES for assisting people with hand opening disabilities have been presented. The comparison between the real phalanx and the mechanism trajectories provides very satisfying results in terms of trajectory agreement, highlighting the transparency of the mechanism with respect to the users general movements. At the time of writing, several exoskeleton devices are involved in the testing phase on patients at the Don Carlo Gnocchi Foundation Rehabilitation Center, Florence, Italy. These exoskeletons present a reduced weight, less than 500 g, and they proved to be very comfortable for the users.
References 1. Truelsen, T., Piechowski-Jzwiak, B., Bonita, R., Bogousslavsky, J., Boysen, G.: Stroke incidence and prevalence in Europe: a review of available data. Eur. J. Neurol. 13, 581–598 (2006) 2. United Nation Department of Economic and Social Affairs Population Division, World population ageing 2013, United Nations publication, 2013, ST/SEA/SER.A/348
Design and Motion Analysis of a Wearable and Portable Hand Exoskeleton
377
3. Rosenstein, L., Ridgel, A.L., Thota, A., Samame, B., Alberts, J.L.: Effects of combined robotic therapy and repetitive-task practice on upper-extremity function in a patient with chronic stroke. Am. J. Occup. Ther. Official Publ. Am. Occup. Ther. Assoc. 1(62), 28–35 (2008) 4. Idrogenet srl.: Gloreha(R) (2014). http://www.gloreha.com/ 5. Schabowsky, C., Godfrey, S., Holley, R., Lum, P.: Development and pilot testing of HEXORR: Hand EXOskeleton Rehabilitation Robot. J. NeuroEng. Rehabil. 7(1), 36–52 (2010) 6. Takahashi, C.D., Der-Yeghiaian, L., Le, V., Motiwala, R.R., Cramer, S.C.: Robot-based hand motor therapy after stroke. Brain 131(2), 425–437 (2008) 7. Toya, K., Miyagawa, T., Kubota, Y.: Power-assist glove operated by predicting the grasping mode. J. Syst. Design Dyn. 5(1), 94–108 (2011) 8. Lucas, L., DiCicco, M., Matsuoka, Y.: An EMGcontrolled hand exoskeleton for natural pinching. J. Robot. Mechatron. 16(5), 482–488 (2004) 9. Conti, R., Meli, E., Ridolfi, A.: A novel kinematic architecture for portable hand exoskeletons. Mechatronics 35, 192–207 (2016) 10. Maciejasz, P., et al.: A survey on robotic devices for upper limb rehabilitation. J. NeuroEng. Rehabil. 11(1) (2014) 11. Heo, P.: Current hand exoskeleton technologies for rehabilitation and assistive engineering. Int. J. Precis. Eng. Manuf. 13(5), 807–824 (2012)
Nitiglove: Nitinol-Driven Robotic Glove Used to Assist Therapy for Hand Mobility Recovery C. Jiménez, R. Mora, J. Pérez and K. Quirós
Abstract Physical therapy is an important resource for the recovery process of several medical conditions. Hand mobility impairment, for example, affects a patients life quality, making it a need to develop aid devices that improve the results of hand therapy, quickening the recovery process. The engineering design process of a wearable and portable rehabilitation glove, was based on the use of muscle wires or Nitinol and specially designed flex sensors. The automated control of this device is performed based on Pulse Width Modulation (PWM), its working cycle, and the feedback provided by the flex sensors, which allow the controlled movement of the different joints in each finger through the use of an interactive graphical user interface, simplifying the phases of measuring the bending angles of each joint before and after each session.
1 Introduction Different medical conditions such as carpal tunnel syndrome, arthritis and neurological disorders compromise the ability and strength to properly control hand movement [2]. In these cases physical therapy becomes an important resource in achieving recovery. The device aids specifically in repetitive joint movement which helps the recovery process. Hand therapy includes repeating a sequence of movements to allow the muscles to develop the strength and coordination to perform simple tasks. Patients that suffered C. Jiménez (B) · R. Mora · J. Pérez · K. Quirós Instituto Tecnológico de Costa Rica, Cartago, Costa Rica e-mail:
[email protected] R. Mora e-mail:
[email protected] J. Pérez e-mail:
[email protected] K. Quirós e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_62
379
380
C. Jiménez et al.
a stroke, for example, need to be treated in accordance to both condition severity and stage of recovery (in this case, specifically flaccidity or spasticity). One of the main treatments involves repeatedly opening one’s hand completely and then making a fist [3]. Given the exercises included in a hand therapy session, a wearable, portable and low-cost rehabilitation glove was designed using shape memory wires to execute the programmed movement. Specially designed flex sensors work alongside a user interface (UI), allowing the patient and therapist to monitor the session and visualize the movement of each phalange. The following sections will detail the design, operational principle and specifications of this device.
2 Engineering Design Process 2.1 Muscle Wires Nitinol is a shape memory alloy composed from nickel and titanium, that has the distinction of being able to react to environmental stimuli (such as electrical current and heat), allowing the development of more advanced and inexpensive actuators with a significant reduction in mechanical complexity and size [4]. To achieve the desired movement the material was preformed by a heat treatment that forces it to transition from martensitic to austenitic crystalline structure. Setting the material’s shape before this treatment allows it to regain said form on the application of stimuli, therefore, once the material’s temperature rises it will reshape to a desired configuration [1]. The main objective of the preforming process is to obtain wires bent on 90◦ or "L" shaped. This simulates a joint while completely bent, as shown in Fig. 1. The preformed wire was then sewn to a glove on the joints both lateral and palmar sides to handle the contraction (as shown in Fig. 2) while straight ones were set in the palmar side only, to handle the finger stretching. The proposed design contemplates the independent movement of each joint, allowing the glove to perform independent contractions. This enables the user to
Fig. 1 Hand model showing 90◦ bent on joints
Nitiglove: Nitinol-Driven Robotic Glove …
381
Fig. 2 Hand prototype showing 90◦ bent on joints
plan different exercises to work on. In order to generate the required force to move the joints, three 0.015 in. caliber muscle wires are used on the contraction, while two are used to stretch a single finger. Finally, each joint movement is controlled by applying Pulse Width Modulation (PWM) signals to a wire array using a microcontroller and a basic circuit with a power MOSFET. The waves frequency as well as its duty cycle are calibrated in values of 25 Hz and 50 % respectively, supplied with 6 V and 1.5 A peak power per wire. Wire arrays handle a single joint movement. This means that bending a joint at 90◦ is done by a set, while another one handles the stretching, though not simultaneously. This means that the average power consumption for the whole system equals the specifications required for the biggest wire set.
2.2 Flex Sensors Custom sensors were built with a pressure-sensitive conductive sheet, named Velostat or Linqstat, which is a versatile component that is used to create different kinds of flex and pressure sensors. The final prototype was built with three 2 × 0.5 cm Velostat rectangles, and two copper tape layers. Each one has an electrical resistance range of [1.5, 30] k approximately. This range is converted into bending angles through the reading of the analog pins of an Arduino. To be able to do this conversion and obtain appropriate angle values its necessary to perform a basic calibration with the sensors secured to each intended joint. These sensors are sewn opposite to the wire array used to stretch joints, as shown on Fig. 3.
382
C. Jiménez et al.
Fig. 3 Flex sensors sewn onto the glove in order to measure joint approximate position
2.3 Results This design process achieves complete control over joints. Changes in frequency showed changes in the wires behavior while contracting, but due to the requirement of the users comfort, frequency was adjusted to the given value of 25 Hz. Heating the wire is needed for it to change its shape, so through an average temperature at a 50 % duty cycle PWM, the wire exerted enough force to both contract and expand the joint used to experiment. Given that three sensors per finger are required, except for the thumbs, which requires two, its advisable to use the same dimensions for all the used sensors, obtaining a value within a specific range and reducing measurement errors. Fixing the sensors directly to the glove increased their precision and stability. Tables 1 and 2 show joint angles obtained before and after applying the controlled signal to stretching and bending wire arrays respectively. Data shows effectiveness on stretching and bending motions in a single joint. Variation on results goes up to 4 4 % error on position 4◦ along 90◦ movements, meaning it represents a 4/90 90 obtained. Given the system’s specifications this error can be dismissed. The wire contracts on stimuli due to the material’s properties. This contraction, however was observed to be 1 mm on both the 46 and 26 mm wires. The relative Table 1 Experimental angles achieved by the stretching wire set
Initial angle (◦ )
Final angle (◦ )
90 90 90
9 9 7
Table 2 Experimental angles achieved by the bending wire set
Initial angle (◦ )
Final angles (◦ )
0 0 0
94 92 90
Nitiglove: Nitinol-Driven Robotic Glove …
383
contraction is irrelevant given its dimension and the fact it doesn’t affect the glove’s functionality.
3 Conclusion Functionality was achieved because of the adaptability and integrated work of the nitinol and the flex sensors, making the glove wearable and portable, helping therapies be more practical. The modification and programming of routines, makes the device adaptable for the treatment of different conditions, and helps acquire data for research and feedback of the patients progress. The results also show the repetitiveness of the wire performance in both cases, stretch and bend, which supports the reliability of the design device. Futures work involves the combination between the different actuators in order to achieve mid bending angles. Acknowledgments The authors gratefully acknowledge the intellectual contributions of Licda. Alejandra González C., Msc. Marta E. Vílchez M., and Ing. Jorge Bolaños, for this project’s research and development.
References 1. Gao, F., Deng, H., Zhang, Y.: Hybrid actuator combining shape memory alloy with dc motor for prosthetic fingers. Sensors and Actuators A: Physical, vol. 223, pp. 40–48 (2015). ISSN: 0924-4247, http://dx.doi.org/10.1016/j.sna.2014.11.025 2. Gloreha: Hand Rehabilitation device. http://www.gloreha.com/index.php/en/home-eng 3. Craig, D., Lucy Der-Yeghiaian, Vu Le, Rehan R. Motiwala, Steven, C.: Robot-based hand motor therapy after stroke. Brain 131(2), 425–437 (2008). doi:10.1093/brain/awm311 4. Jani J.M., Leary, M., Subic, A., Gibson, M.A.: A review of shape memory alloy research, applications and opportunities. Materials & Design, vol. 56, pp. 1078–1113 (2014). ISSN: 02613069, http://dx.doi.org/10.1016/j.matdes.2013.11.084
3D Printed Arm Exoskeleton for Teleoperation and Manipulation Applications Pierre Letier, Gonzalo Rodriguez, Guillaume Fau, Shashank Govindaraj, Torsten Siedel, Jeremi Gancet and Michel Ilzkovitz Abstract This paper presents the recent activities of Space Applications Services for the development of force feedback arm exoskeletons based on 3D printing technologies. The paper describes the design of the exoskeleton system and illustrates its application through two different projects of slave robotic arm teleoperation, ICARUS and DEXROV, where the concept is used.
1 Introduction Since several years, Space Applications Services develops exoskeletons technology to enhance advanced space and ground robotics arm teleoperation applications. The conventional approach and interfaces for e.g. tele-operating a robotic manipulator can be ineffective for operators. Arm exoskeletons are intuitive interfaces allowing to overcome such limitations, capturing human arm movements and translating them to robot motion commands in a convenient and effective manner. Assets include a large workspace and force feedback to the operator during manipulation tasks, resulting in highly intuitive operations. This is especially important for activities requiring human skills and expertise, in e.g. hostile, non-reachable or de/un/structured environment. The first exoskeleton arm system we have been working on was a demonstrator for robotic teleoperation in space [1, 2]. This paper presents our experience with the next generation of exoskeleton we have been developing afterward, in the context of the EC funded projects ICARUS (and more recently DexROV). As a key characteristics, this 2nd generation is almost fully built relying on 3D printing rapid prototyping techniques, instead of regular aluminum structural components. 3D printing allows completely re-considering the approach to the mechanical design as it offers a lot more freedom than more conventional approaches. This technique has been used to reduce the complexity of the overall mechanical design (e.g. number P. Letier (&) G. Rodriguez G. Fau S. Govindaraj T. Siedel J. Gancet M. Ilzkovitz Robotics Group, Space Applications Services, Zaventem, Belgium e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_63
385
386
P. Letier et al.
of components, optimized components regarding efforts and functionality), and to reduce the weight and the cost of manufacturing. In the following section, this paper describes the mechatronic design of the exoskeleton. As an application case, the EC FP7 ICARUS project is further introduced, as well as the EC H2020 DexROV project where similar approach is being applied in a large extent.
2 Exoskeleton Design 2.1
Mechanical Design
The developed exoskeleton arm has an anthropometric kinematic configuration that follows the structure of the human arm, with seven degrees of freedom, from the shoulder to the wrist. The upper arm rotation is ensured by an open rail guiding system that allows to enter easily within the exoskeleton and reduces collisions with user’s body (in contrast to the previous generation implementing a full circular guide). Apart from the rail support, all the structural mechanical components are built from Alumide (mix powder of plastic and aluminum) and 3D printed with a Selective Laser Sintering process (SLA). This approach offers an excellent trade-off between mechanical design freedom, weight and stiffness (required for haptic interfaces). Moreover, this material can be easily reworked with conventional techniques, ensuring good integration with COTS elements as bearings (Fig. 1).
Fig. 1 3D printed exoskeleton
3D Printed Arm Exoskeleton …
387
The exoskeleton has five adjustment links - between the back, the arm and the hand, that can be easily adapted, without tooling, to the size of the operator. Polypropylene orthopedic orthoses ensure a good, safe and comfortable fixation and force transmission between the exoskeleton and the operator.
2.2
Mechatronics
Each joint of the exoskeleton is actuated by the coupling of a geared brushless DC motor with a capstan reducer (cable based reducer frequently used for haptic interface due to its limited friction and absence of backlash). The specificity of this design is the implementation of multiple cables on the most powerful joint to improve to robustness and reliability of this technology. Each joint is also equipped with position sensors and local joint electronics (Synapticon boards) that perform electrical interfaces with the sensors and the motors, control locally the joint (in position or torque) and communicate with the other joint and the main controller through EtherCAT communication bus. The main controller is responsible for the global management of the system and is based on the OROCOS framework with the implementation of an EtherCAT Master interface.
3 Application 1: ICARUS ICARUS is an EC FP7 funded project aiming developing and demonstrating mobile robotic capabilities (aerial, ground and see) in support to Search and Rescue (SAR) operations. In the context of Urban Search and Rescue (USAR) scenarios, the exoskeleton has been interfaced with 2 different types of robotic arms: a lightweight 6 dof electrical manipulator mounted on a 50 kg tracks UGV, and large 5 dof hydraulic manipulator mounted on a 4 tons tracks UGV. Relying on the developed exoskeleton system, we could experiment complex manipulation operations such as clearing debris, opening doors, inspection, deploying first aid kit and assisting robot motion in uneven terrain. Windowing position control (to scale the position control of the robotic arm) and force-feedback based on position-position algorithms (no force sensors or estimations on the slave robots) have been implemented and tested. During field trials, the effectiveness of the exoskeleton has been analyzed and the intuitiveness of the system proven (Fig. 2).
388
P. Letier et al.
Fig. 2 Excavator control with exoskeleton during ICARUS field trial
4 Application 2: DEXROV DexROV is an ongoing EC Horizon 2020 funded project addressing the development of under-sea robotic intervention capabilities, with a focus on (1) far distance teleoperation—involving variable communication latencies to mitigate, and (2) dexterous manipulation capabilities benefiting from context specific human skills and know-how. DexROV intends to develop cost-effective technologies and methods that will enable subsea operations with fewer off-shore personnel while increasing the range, flexibility and complexity of operations that are possible. In DexROV, a double exoskeleton arms system will be developed and interfaced with a virtual reality environment (rendering models built from data perceived undersea). The operator will have the possibility to demonstrate the dexterous manipulation actions to be done. These will be translated into primitives and passed to the ROV controller (offshore side), so that to execute corresponding manipulation actions. In this project, the former exoskeleton concept will be enhanced with a second arm and with exoskeleton hands to match all degrees of freedom available with 2 purposely developed dexterous manipulators to be mounted on the ROV. The exoskeleton system will feature a modular gravity compensation system to reduce the constraints on the actuation chain. Novel mechanical design technics based on shell structure are under investigation in complement to the previous 3D printing based approach, for improved performance and robustness (Fig. 3).
3D Printed Arm Exoskeleton …
389
Fig. 3 DexROV operation concept
Acknowledgments ICARUS is founded from the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement n285417 -project ICARUS (Integrated Components for Assisted Rescue and Unmanned Search operations). DexROV contributes in the «Blue Growth» long term European strategy to support sustainable growth in the marine and maritime sectors. DexROV is funded through the EC Horizon 2020 programme (Grant #635491).
References 1. Letier, P., Motard, E., Verschueren, J.-P.: EXOSTATION: haptic exoskeleton based control station. In: 2010 IEEE International Conference on Robotics and Automation (ICRA), 1840– 1845 (2010). doi:10.1109/ROBOT.2010.5509423 2. Letier, P., Preumont, A.: Portable haptic arm exoskeleton. In: Sobh, T., Xiong, X. (eds.) Prototyping of Robotic Systems: Applications of Design and Implementation, pp. 122–145. Information Science Reference, IGI Global (2012). Chapter 5
Musculoskeletal Simulation of SOLEUS Ankle Exoskeleton for Countermeasure Exercise in Space Moonki Jung, Guillaume Fau, Pierre Letier, Uwe Mittag, Jochen Zange, Jörn Rittweger and Arnaud Runge
Abstract SOLEUS system aims at providing efficient countermeasure exercises focused on the lower legs. The final product is foreseen to be based on an orthotics structure allowing exercising the ankle joint and muscle groups, even in microgravity. In order to test the pedal-pulling scenario for the operation of the system, multibody dynamics based musculoskeletal simulation has been performed. The result of the simulation shows the profile of the ankle plantar flexion torque by muscles with the given exoskeleton’s actuator force and motion conditions. Also the muscle activation patterns could be retrieved.
M. Jung (&) AnyBody Technology, Aalborg, Denmark e-mail:
[email protected] G. Fau P. Letier Space Applications Services, Zaventem, Belgium e-mail:
[email protected] P. Letier e-mail:
[email protected] U. Mittag J. Zange J. Rittweger German Aerospace Center (DLR), Köln, Germany e-mail:
[email protected] J. Zange e-mail:
[email protected] J. Rittweger e-mail:
[email protected] A. Runge European Space Agency, Paris, France e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_64
391
392
M. Jung et al.
1 Introduction Current exercise devices on-board the ISS include e.g. a treadmill device (Treadmill with vibration isolation system, TVIS), and a device allowing a training analogue for weight lifting with barbells (Advanced Resistive Exercise Device, ARED). Astronauts use only about 45 min of their recreation time for exercise and the loads used for running or resistive strength training (on IRED, the precursor of ARED) were lower than the virtual body weight related to current body mass. These circumstances may explain the partial efficiency of exercise countermeasures in space. But neither in a treadmill where the trainee performs the typical locomotion activities like walking or running, nor in the resistance devices the lower leg enters the energy storing spring like mode. This can be only effectively achieved by a jumping or hopping-like motion. SOLEUS system aims at providing efficient countermeasure exercises focused on the lower legs by allowing customized and targeted training of the body parts mainly afflicted by bone and muscle degradation. In order to test the pedal-pulling scenario as a possible exercise protocol of the system, inverse dynamics based musculoskeletal simulation has been performed so that the profile of ankle plantar flexion torque by muscles could be calculated with given exoskeleton’s actuator force and motion conditions. Also the relevant muscle activations could be simulated as well so that the system’s performance can be estimated.
2 Methods AnyBody Modeling System [1] has been used in order to perform the inverse dynamics analysis on the human-exoskeleton combined model.
2.1
Musculoskeletal Human Model
The lower body musculoskeletal model including head, thorax, pelvis and leg segments from AnyBody Managed Model Repository (AMMR v1.6) was used for this study [2]. This human model is scaled by its default setting so that the height is 1.8 m and the mass is 75.6 kg. There are 22 degrees of freedom in this lower body model. Muscles only exist in the lower body segments whereas there are some joint torque actuators in the upper body joints.
Musculoskeletal Simulation of SOLEUS Ankle Exoskeleton … Table 1 Simulatoin parameters
2.2
393
Parameters
Values
Dorsiflexion max angle (deg.) Plantarflexion max angle (deg.) Plantarflexion speed (deg./sec.) Medial actuator force (N) Lateral actuator force (N)
20 40 60 −400 −400
Definition of Pedal-Pulling Motion
User of this system is supposed to only move his ankle joint except anything else. During 1 s of the simulation period, then ankle plantar flexion angle will change from 20° in dorsiflexion to 40° in plantarflexion. So the speed of ankle plantarflexion angle is 60° per second.
2.3
Conditions of the Linear Actuators
There are two linear actuators (medial and lateral) which are involved in both dorsiflexion/plantarflexion and pronation/supination degrees of freedom. For the simplicity of simulations, both medial and lateral linear actuators have the same force profile of constant linear force as -400 N. Negative force means that there will be a pulling force inside the linear actuator instead of a pushing force (Table 1). In terms of the kinematics of these actuators, these two actuators are aimed to be driven in order to always have the same length each other. So the ankle pronation/supination motion of the human model will be determined by the kinematics of these two actuators.
2.4
Interactions Between Human and SOLEUS System
There are several contact points defined between SOLEUS and human segments such as thigh, shank and foot in order to predict the interactions between these segments [3]. Also the existence of tension forces in the straps is assumed. And the shank part of SOLUES system is assumed to be fully connected to the ground by using 6 reactional forces and moments. Finally, pelvis is also fully fixed to the virtual chair by using 6 reactional forces and moments (Fig. 1).
394
M. Jung et al.
Fig. 1 Interactions between human, SOLEUS and environment
2.5
Inverse Dynamics of Musculoskeletal System
The system of equilibrium equations in inverse dynamics for a musculoskeletal system can be organized on the following form [1]: Cf ¼ r
ð1Þ
where f is a vector of muscle and joint forces, r is a vector of representing the external forces and inertia forces, and C is the matrix of equation coefficients. In general, there is a known issue for muscle redundancy due to the fact that there are more muscles and the degrees of the freedom of the entire system. So the equilibrium equation has more unknowns than the number of its equations and therefore there are infinitely many solutions. So this muscle recruitment can be solved by using optimization and the problem can be formulated in the following way: Minimize : Gðf ðMÞ Þ ¼
X fi 3 Ni
ð2Þ
And there are some constraints regarding the directions of the muscles force because muscles are unilateral element that can only generate the pulling force [1]: ðMÞ
fi
0;
i ¼ 1; . . .nðMÞ
ð3Þ
3 Results Inverse dynamics analysis of human-SOLEUS combined model could generate various human-related biomechanical values such as muscle activations and joint torques by muscles. Individual muscle forces and activations could be determined
Musculoskeletal Simulation of SOLEUS Ankle Exoskeleton …
395
Fig. 2 Muscle activations while using SOLEUS system
Fig. 3 Ankle plantarflexion torque generated by muscles
by muscle recruitment optimization under the condition that those muscle forces should generate the necessary amount of human joint torques. And especially the ankle plantar flexion torque by muscles should overcome the torque that is generated by two linear actuators (Figs. 2, 3).
4 Conclusion A musculoskeletal simulation model that includes both human and exoskeleton could be developed in order to perform the inverse dynamics analysis so that various musculoskeletal values such as muscle activations and joint torques could be calculated. Also the interaction forces between human and exoskeleton are available from the same simulation process. All of these simulations values can be utilized in order to estimate the performance of a new exoskeleton system which is supposed to provide a certain amount of countermeasure exercise effect on its users.
396
M. Jung et al.
Acknowledgments This study is funded by ESA in the framework of a Technology Research Program (contract no. 4000112181/14/NL/RA) entitled “Integrated Countermeasures with Biofeedback and Actuators”.
References 1. Damsgaard, M., Rasmussen, J., Christensen, S.T., Surma, E., De Zee, M.: Analysis of musculoskeletal systems in the AnyBody Modeling System. Simul. Model. Pract. Theory 14, 1100–1111 (2006) 2. AnyBody Technology A/S. http://www.anybodytech.com/ 3. Jung, Y., Jung, M., Lee, K., Koo, S.: Ground reaction force estimation using an insole-type pressure mat and joint kinematics during walking. J. Biomech. 47(11), 2693–2699 (2014)
Human Gait Feature Detection Using Inertial Sensors Wavelets S. Glowinski, A. Blazejewski and T. Krzyzynski
Abstract The human gait analysis by using wavelets transform of signal obtained from six inertial ProMove mini sensors is proposed in this work. The angular velocity data measured by the gyro sensors is used to estimate the translational acceleration in the gait analysis. As a result, the flexion–extension, the adduction– abduction joint angles of the hips, flexion–extension of the knees and dorsi and plantar flexion of the ankle are calculated. After measurements we propose to use one of wavelet transform (wavelet type) in order to analyze the signals, indicate a characteristic feature and compare them.
1 Introduction This paper is a proposal of a new analysis and experimental tools application in a scope of human gait. Gait analysis is a clinical tool for obtaining quantitative information of the gait of a person to diagnose walking disabilities [1, 2]. Common methods of gait analysis include cameras application to track the position of body by using reflective markers. An alternative is to use acceleration and angular velocity data, which are measured from inertial sensors attached to the body. Miniature inertial sensors are steadily gaining interest because of their limited power consumption, low cost and good user compliance, when they are embedded in wearable sensor systems or portable devices [3]. Currently several applications in human motion analysis may benefit from miniature inertial sensors [4]. Human body motions are captured, measured and next inertial data from sensors are transmitted to computer. Then data have to be processed because this method does S. Glowinski (&) A. Blazejewski T. Krzyzynski Koszalin University of Technology, Koszalin, Poland e-mail:
[email protected] A. Blazejewski e-mail:
[email protected] T. Krzyzynski e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_65
397
398
S. Glowinski et al.
not directly measure position. But a major challenge is to translate these data into meaningful information.
2 Wireless Sensing System The sensing system is composed of six body wireless inertial sensor nodes based on ProMove mini platform [5]. The measurement range of the sensor is 10 degreesof-freedom: ±2G to 16G (accelerometer), ±250 to 2000◦/s (gyroscope with resolution 0.007◦/s) and ±4912 lT magnetic field intensity (compass). The maximum sampling rate is 1000 Hz, and the measured data are transferred wirelessly by using Inertia Gateway as a central hub. The sampling rate of all measured signals equals 200 Hz, because it is expected a significant meaning of low frequencies the signals’ components. The Inertia Studio software enables real-time visualization of sensor data, as well as over-the-air reconfiguration of the sensors and wireless parameters. All data retrieved by the Inertia Studio software is logged for post-analysis. As shown in the Fig. 1, the ProMove mini sensors are placed on the part of body along the right and left leg on the thigh, shank and foot. In global coordinate system (green color) the x-axis is the walking direction, y-axis lateral direction and z-axis opposite direction of gravity. For each sensor the orientation relative to Earth reference frame is calculated in terms of roll, pitch and yaw angles. By combining the orientation of each node, we obtain the joint motions of the three segments of the lower limbs. In our experiments the kinematic parameters of interest are the knee and ankle joint angles. By using four quaternion parameters q0, q1, q2, q3 obtained from survey, we calculate the Euler angles h of each node as follows: h ¼ arcsinð2ðq0 q2 q3 q1 ÞÞ
Fig. 1 The human model and coordinate systems. The x, y, z coordinates represent the global coordinate system, the xs, ys, zs represents the sensor coordinate system
ð1Þ
Human Gait Feature Detection Using Inertial Sensors and Wavelets
399
Fig. 2 Knee angles during gait
where it means rotation about y-axis. In the Fig. 2, the knee flexion angle of left and right leg in sagittal plane, averaged over whole gait period, is shown. The angles are calculated by divided data into separated steps. Each step is interpolated, because of different number of samples. The mean angle of the right and left knee is transformed to the cycle. For example, in this case it is noted a slight difference between angles in loading response and swing (approx. 3°). In the left leg in stance phase, there is noticed the characteristic change in flexion angle in 13–15 % of cycle, whereas in the right leg this phenomenon does not occur.
3 Wavelet Analysis The sensors transfer an acceleration signals, among others. The sum of the components is taken into consideration as analysed signal, which describes the movement of the part of a leg. The signal includes acceleration of gravity. We propose to use algorithm combining discrete Fourier transform (DFT) and continuous wavelet transform (CWT). It applies DFT of the signal in first step and next the same transform of the analyzing wavelet at the appropriate angular frequencies in order to obtain directly comparable scales. In the next step the algorithm takes the product of the signal DFT and the wavelet DFT over all scales found. Eventually, it inverts the DFT to obtain the CWT coefficients. This procedure easy allows to convert wavelets’ scales to frequencies, precisely to so-called pseudo-frequencies. These pseudo-frequencies represent, not the exact frequencies, but some frequency ranges. The Morlet wavelet is the analyzing wavelet in the using algorithm [6, 7]. The conducted analysis shows some particular features of the human gait, which are not possible to observed in raw signal. Look out for important symptoms of human gait may be difficult as well. The example of analysis of 4 steps gait, for one sensor placed on human foots (right and left leg), is shown in Fig. 3. The Fig. 3a, b show the analyzed signal (red curve) and the signal after reconstruction (blue curve) by using inverse wavelet transform for specific, chosen range of scales. Both signals should be the same for whole range of scales, which indicates that the wavelet transform does not cause any loss of information.
400
S. Glowinski et al.
Fig. 3 The analyzed (red) and reconstructed (blue) signals after inverse wavelets transform (aright, b-left foot). Continuous wavelet transform of the analyzed signals obtained by using Morlet wavelet of parameter 4, where c, d is modulus of wavelet coefficients vs. scales and e, f versus pseudo-frequency
In the Fig. 3c, d there are absolute scales values of wavelet coefficients shown for each leg. This kind of analysis may deliver details about components of a gait characterized by constant frequencies, having variation similar to harmonic functions. The area approx. above 4.59 scales values, along whole time axis, indicates constant pace of human gait. The character and variation of this pace, confirm the harmonic changes. The other area is in the range of higher frequencies (Fig. 3c, d), at scale level approx. 1.09. It is related to steps periods. In these figures the areas on appropriate scales level (approx. 0.59) show, when the human foot, at first moment a heel and next a mid-foot, take a contact with a floor and next it is lifted above. Below, in the highest frequency range, there is another region that shows moments, when a body weight is removed from the mid-foot on toes. These parts of signals are more clearly seen after to pseudo-frequency domain recalculation, which is shown in Fig. 3e, f. Comparing the both set of wavelets coefficients values, it is possible to observe the symmetry or asymmetry of a gait and other significant features.
4 Conclusion The many more analyses are possible using wavelets. In the paper we propose the methodology, which helps to recognize specific feature of human gait that can be useful in health diagnostic or sport individual training programs creation. The
Human Gait Feature Detection Using Inertial Sensors and Wavelets
401
advantage afforded by wavelets is the ability to perform local analysis. Because wavelets are localized in time and scale, wavelet coefficients are able to localize characteristic changes or differences in analysed signals [8]. By shifting parameters of wavelets, they can be applied as a focus directed to interesting signal area described by time and scale related to frequencies.
References 1. Glowinski, S., Krzyzynski, T., Pecolt, S., Maciejewski, I.: Design of motion trajectory of an arm exoskeleton. Arch. Appl. Mech. 85, 75–87 (2015) 2. Glowinski, S., Krzyzynski, T.: An Inverse Kinematic Algorithm for Human leg. J. Theor. Appl. Mech. 54(1), 53–61 (2016) 3. Chen, X.: Human Motion Analysis with Wearable Inertial Sensors. University of Tennessee, Knoxville, Doctoral Dissertation (2013) 4. Zhu, R., Zhou, Z.A.: A real-time articulated human motion tracking using tri-axis inertial/magnetic sensors package. IEEE Trans. Neural Syst. Rehabil. Eng. 12, 295–302 (2004) 5. ProMove wireless inertial sensing platform. http://www.inertia-technology.com/promove-mini 6. Mallat, S.: A Wavelet Tour of Signal Processing. Academic Press, San Diego, CA (1998) 7. Torrence, C., Compo, G.P.: A practical guide to wavelet analysis. Bull. Am. Meteorol. Soc. 79, 61–78 (1998) 8. Daubechies, I.: Ten Lectures on Wavelets. Society for Industrial and Applied Mathematics (SIAM), Philadelphia (1992)
On the Importance of a Motor Model for the Optimization of SEA-driven Prosthetic Ankles Tom Verstraten, Glenn Mathijssen, Joost Geeroms, Louis Flynn, Bram Vanderborght and Dirk Lefeber
Abstract Several examples in literature demonstrate the potential impact of motor inertia on the electrical energy consumption of actuators. Nevertheless, optimizations of actuated prosthetics are often based on the mechanical energy consumption, disregarding the potential effects of motor inertia. In this short abstract, we simulate the electrical energy consumption of a powered prosthetic ankle actuated by a Series Elastic Actuator. Its compliant element is optimized for mechanical energy consumption, a typical strategy in state-of-the-art prosthetics. Our results confirm the importance of motor inertia. Due to the resulting changes in the operating points of the motor, the average motor efficiency is lowered by 17 %.
1 Introduction In a previous study on a pendulum, we concluded that motor inertia plays an important role in the electrical energy consumption of a stiff actuator [1]. In such actuators, the operating point corresponding to minimal mechanical energy consumption does not coincide with the operating point of minimal electrical energy consumption. The difference between both can be explained by the power needed to accelerate the motor’s inertia, leading to an electrical power profile which is dinstinctly different from the mechanical power profile. In compliant actuators, motor inertia also has an impact on the dynamics of the actuator. While a Series Elastic Actuator (SEA) with zero motor inertia would have a single resonance frequency, a nonzero motor inertia introduces a second resonance frequency and an antiresonance frequency [2, 3]. It is the latter which is of importance for the minimization of electrical peak power and energy consumption [4]. The first and second author are Ph.D fellows of the Research Foundation Flanders (FWO). The third author is funded by a Ph.D grant of the Agency for Innovation by Science and Technology Flanders (IWT). T. Verstraten (B) · G. Mathijssen · J. Geeroms · L. Flynn · B. Vanderborght · D. Lefeber Robotics and Multibody Mechanics Research Group, Vrije Universiteit Brussel, Ixelles, Belgium e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_66
403
404
T. Verstraten et al.
Nevertheless, most papers on the optimization of SEAs for prosthetics have focused on mechanical output energy or peak power, not including the motor inertia, motor and gearbox efficiency into their models [5–7]. In this extended abstract, we briefly discuss the consequences of such a simplified approach for the power and energetics of actuated prosthetic ankles.
2 Materials and Methods In our analysis, we will study a prosthetic ankle actuated by a SEA. We consider a conceptual SEA consisting of a torsional spring and a 200 W Maxon RE50 DC motor with a gearbox (gear ratio n = 319, efficiency ηtr = 68 %). The spring stiffness is will be the output of an optimization with the goal of minimizing the absolute mechanical energy consumption E mech =
|Pmech |dt
(1)
This is the approach followed in most papers. Pmech is calculated from the torque (Tm ) and speed (θ˙m ) at the motor shaft, Pmech = Tm θ˙m
(2)
Tm and θm can be calculated from the desired output torque and position T and θ , which are obtained from gait data [8]: Tm = C T /n θm = n · (T /ks + θ )
(3) (4)
where ks is the stiffness of the series spring and C is the gearbox efficiency function C=
1/ηtr (Pmech > 0) (Pmech < 0) ηtr
(5)
The electrical power consumption Pelec is then calculated as the product of motor current (I ) and voltage (U ). These are obtained by applying the motor model
I = k1t Jm θ¨m + Tm + νm θ˙m U = L I˙ + R I + kb θ˙m
(6)
which requires knowledge of the motor parameters listed in Table 1. A more detailed discussion of this model can be found in [1].
On the Importance of a Motor Model for the Optimization … Table 1 Motor parameters for 200W RE50 Maxon motor
405
Symbol Name
Value
kt kb νm R L Jm
38.5 mNm/A 248 rpm/V 1.46e-5 Nms/rad 0.103 0.0717 mH 536 gcm2
Torque constant Speed constant Friction coefficient Terminal resistance Terminal inductance Motor inertia
3 Results and Discussion In this section, we will present a motor efficiency map constructed from Eq. (6). It represents the motor efficiency as a function of the motor speed θ˙m and the dynamic motor torque Tm + Jm θ¨m . This is the torque which is relevant for the energy consumption of the motor, since it is this torque which needs to be delivered by the motor windings. By plotting the operating points of the motor on such an efficiency map, one can obtain a fairly good idea of how efficiently the motor is used for a specific motion. As such, efficiency maps can be a useful tool for tuning the imposed motion or the actuator design in order to obtain a minimal energy consumption of the actuator [1]. In the case of an SEA-driven ankle prosthesis, the motion profile of the motor is determined by the choice of spring stiffness ks through (4). As mentioned earlier, stiffness is found by minimizing (1), yielding ks = 375 Nm/rad. This is only slightly different from the value of ks = 368 Nm/rad found by Grimmer et al. [5], who applied this exact method for their optimization. Fig. 1 Efficiency map for the 200 W Maxon RE50 motor, which includes the Joule losses R I 2 and the friction losses νm θ˙m2 . The green trajectory corresponds to the operating points (θ˙m , Tm ), the blue trajectory to (θ˙m , Tm + Jm θ¨m ). For reference, the trajectory of a stiff actuator (motor inertia included) is shown in black. The white lines denote the maximum permissible motor torque
406
T. Verstraten et al.
Table 2 Effect of motor inertia Jm on the energy consumption of the Series Elastic Actuator Actuator SEA (Jm = 0) SEA (Jm = 536 gcm2 ) Stiff (Jm = 536 gcm2 ) E mech (J) 21.4 E elec (J) 27.7 Average motor eff (%) 77.1
21.4 35.8 59.7
21.4 46.4 46.1
To illustrate the importance of motor inertia, two sets of operating points for the prosthetic ankle motor are plotted on the motor’s efficiency map (Fig. 1). The first set (green) is based on the motor shaft torque Tm , which can actually be measured by a torque transducer. The second set (blue) is based on the motor torque with motor inertia included, Tm + Jm θ¨m . While the optimization is based on mechanical power, thus on Tm , it is the latter set of operating points which influences the instantaneous motor efficiency. However, we can see that both sets of operating points follow completely different trajectories. The operating points (θ˙m , Tm ) are confined within the recommended operating region of the motor. They do not exceed the maximum motor speed (9500 rpm) or maximum motor torque (700 mNm), as required for a good design. Looking at (θ˙m , Tm + Jm θ¨m ), however, we find that the operating points exceed this recommended region by far, yielding peak torques of up to 1.28 Nm. Furthermore, commanding high torques at relatively low speeds leads to poor efficiencies [1]. This is exactly the region in which the SEA is operating. We can therefore expect the actual efficiency of the motor to be a lot lower than the catalog efficiency, with potentially detrimental consequences for the overall efficiency of the actuator. This is confirmed by the calculated energy consumption, presented in Table 2. With a massless rotor, the calculated electrical energy consumption would be 27.7J. With motor inertia, the energy consumption rises to 35.8 J. Still, the SEA clearly outperforms a stiff actuator, which would consume no less than 46.4 J. Note that, in a cyclic motion, inertias do not contribute directly to energy consumption. However, like springs, their ability to store energy affects the system’s power flows. This can have a positive or negative influence on the losses in other components. As seen in Table 2, the motor inertia causes a decrease in average motor efficiency of approximately 17 %. This demonstrates that the SEA’s trajectory (θ˙m , Tm + Jm θ¨m ) indeed passes through regions of lower motor efficiency.
4 Conclusion In this short abstract, we have shown that motor inertia may contribute significantly to the total torque experienced by the motor. As such, it has a considerable impact on the energy consumption of the actuator. This could also lead to differences in the optimized spring stiffness, which is very relevant for compliant designs. Future research will therefore focus on quantifying the impact of the motor model on the optimization of compliant actuators for wearable robotics.
On the Importance of a Motor Model for the Optimization …
407
References 1. Verstraten, T., Mathijssen, G., Furnémont, R., Vanderborght, B., Lefeber, D.: Modeling and design of geared DC motors for energy efficiency: comparison between theory and experiments. Mechatronics 30, 198–213 (2015) 2. Sensinger, J.W., Burkart, L.E., Pratt, G.A., et al.: Effect of compliance location in series elastic actuators. Robotica 1–6 (2013) 3. Beckerle, P., Wojtusch, J., Rinderknecht, S., von Stryk, O.: Analysis of system dynamic influences in robotic actuators with variable stiffness. Smart Struct. Syst. 13(4), 711–730 (2014) 4. Verstraten, T., Beckerle, P., Furnémont, R., Mathijssen, G., Vanderborght, B., Lefeber, D.: Series and parallel elastic actuation: impact of natural dynamics on power and energy consumption. Mech. Mach. Theory 102, 232–246 (2016) 5. Grimmer, M., Eslamy, M., Seyfarth, A.: Energetic and peak power advantages of series elastic actuators in an actuated prosthetic leg for walking and running. Actuators 3(1), 1 (2014) 6. Paluska, D., Herr, H.: The effect of series elasticity on actuator power and work output: implications for robotic and prosthetic joint design. Robot. Auton. Syst. 54(8), 667–673 (2006) 7. Flynn, L., Geeroms, J., Jimenez-Fabian, R., Vanderborght, B., Vitiello, N., Lefeber, D.: Ankleknee prosthesis with active ankle and energy transfer: development of the CYBERLEGs alphaprosthesis. Robot. Auton. Syst. 73, 4–15 (2015) 8. Winter D.A.: Biomechanics and Motor Control of Human Movement. 4th edn. Wiley (2009)
Assessment of a 7-DOF Hand Exoskeleton for Neurorehabilitation Kevin Abbruzzese and Richard Foulds
Abstract Stroke is the leading cause of disability in the United States with approximately 800,000 cases per year. This cerebral vascular accident results in neurological impairments that reduce limb function and limit the daily independence of the individual. Robotic rehabilitation may present an exercise intervention that can improve training and induce motor plasticity in individuals with stroke. A motorized hand exoskeleton that operates under admittance control provides support for wrist flexion/extension, abduction/adduction, pronation/supination, and finger pinch has been integrated with a pre-existing 3-Degree of Freedom (DOF) haptic robot (Haptic Master, FCS Moog) to determine the efficacy of increased DOF during proximal and distal training for neurorehabilitation.
1 Introduction Stroke is the leading cause of disability in the United States with approximately 800,000 cases per year [1]. Despite current technology and interventions, 30–66 % of hemiplegic stroke patients remain without arm function after 6 months recovery; while only 5–20 % demonstrate complete functional recovery of the paretic arm [2]. Neurological impairments affecting the Upper Extremities (UE) significantly limit the independence of the affected subjects. Strong evidence suggests that therapeutic interventions that introduce task specific training in the presence of motor practice can enhance recovery [2–4]. However, optimal therapeutic interventions that promote recovery of arm and hand function remain unclear due to discrepancies in This research is funded by Rehabilitation Engineering Research Center on Wearable Robots for Independent Living (NIDILRR-HSS 90RE5021). K. Abbruzzese (&) R. Foulds NJIT, Newark, NJ, USA e-mail:
[email protected] R. Foulds e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_67
409
410
K. Abbruzzese and R. Foulds
current rehabilitation. Robotic rehabilitation may present a complete exercise therapy that can enhance training and standardize clinical practice. Most robotic systems are designed for elbow and shoulder rehabilitation. However, there is a necessity to develop upper extremity devices that focus on hand rehabilitation. We propose that to improve neuroplasticity and establish appropriate cortical representations, robot mediated therapy should provide a robust approach to incorporate coordination of arm and hand movements. An active 4-DOF hand exoskeleton will be integrated with a pre-existing 3-DOF haptic robot (Haptic Master, FCS Moog) to determine the efficacy of proximal and distal training in UE rehabilitation. The functional system will allow the user to move their arm in space, while obtaining robotic assistance for wrist movements such as flexion/extension, abduction/adduction, and supination/pronation. Pinch assistance will rely on an active gripper with rotation about the metacarpal joints (MCP) joint of the index finger and rotation of the thumb MCP to provide flexion/extension of each digit. The NJIT HandsOn (7-DOF) will support distal hand movements and proximal arm movements in 3-D space in a Virtual Environment (VE). The system operates under admittance control and will allow patients with residual muscle function to perform the desired movements during training. The training system provides a consistent environment to facilitate repetitive practice with haptic assistance. Upper extremity movements are biologically encoded for simultaneous control of arm and hand movement. The motor control response utilizes all DOF to perform sequential movement during a task. Training with more DOF should provide a more natural response and greater coordination. UE arm movements for rehabilitation should reflect the encoded movement and offer training with both gross and fine motor control.
2 Design Components 2.1
Admittance Control Paradigm
Admittance control (AC) allows the user to input a force, translating the force into motion. Under AC, the user senses only the inertia of the small virtual mass which can be very small compared to the inertia and load capabilities of the actual robot actuators. This makes it useful for individuals with significant muscle weakness to employ their residual motor control to effectively drive a powerful robot [5]. The positive features associated with an AC system for UE rehabilitation include: force amplification, back-drivability, and intuitive volitional control (Fig. 1). Users unable to modulate force must rely on preprogrammed movements. Newton’s Second Law can be expanded to include variable system damping where: F(x) = Force (N), M = Mass (kg), B = damping (Ns/m), x_ (t) = velocity (m/s), €x(t) = acceleration (m/s2 ). This second order differential equation (Eq. 1) serves as the basis for the admittance control algorithm.
Assessment of a 7-DOF Hand Exoskeleton for Neurorehabilitation
411
Fig. 1 Schematic representation of admittance control paradigm
€x ¼
2.2
FðxÞ B x_ M M
ð1Þ
Wrist End Effector
A 3-DOF wrist exoskeleton has been designed to mimic the range of motion of the wrist and serve as a distal attachment for the Haptic Master. The device incorporates 3-Dynamixel Motors (Robotis) to provide roll, pitch, and yaw (Fig. 2 Top). The motors are coupled to 3-D printed constructs to support the wrist. The system provides assistance for flexion/extension, abduction/adduction, and pronation/ supination. It should be noted that intended movements are assisted regardless of strength limitations.
2.3
Modular Gripper
A modular gripper can be added to the active exoskeleton to provide the ability to manipulate objects in space and provide haptic sensation within the VE. A single degree of freedom gripper allows rotation of the MCP of the thumb and index
Fig. 2 NJIT HandsOn components: 3-DOF distal end effector with Haptic Master (top). Modular Gripper (left) with VE (right)
412
K. Abbruzzese and R. Foulds
finger. The MCP rotation is generated at the index joint using a Dynamixel MX-28 motor and is coupled to the thumb MCP to provide rotation (Fig. 2). This movement provides tip pinch which is critical to hand function for daily life. Impaired pinch skill significantly affects dexterity function for stroke survivors, leading to higher dependency of daily activities and poor quality of life [6]. It is important that the patient is able to see the hand on the mechanism, especially the thumb to facilitate grasp pre-shaping, maintain grasp and release [3]. The integrated exoskeleton grants the user an increase in range of motion and the ability to manipulate and interact with objects.
2.4
Virtual Environment
Exercise environments can be rendered virtually in order to create a systematic therapy where the intensity of practice as well as the rewarding feedback can be consistently monitored and manipulated to aid in motor learning tasks. Subjects who train in VE show accelerated learning rates as compared to control subjects [7]. Virtual environments possess the ability to provide a more appealing and motivating environment for therapy, offering realistic scenarios with challenging tasks that can better prepare the patient for daily activities. Unity3D is a cross platform 3-D engine that can be used to develop powerful games and applications for therapeutic interventions. A VE was created using Unity3D to develop an interface for the hand exoskeleton. The trajectories of the wrist and finger are mapped to the virtual representation of the hand, Fig. 2.
3 Methods Healthy subjects with no visual or motor impairments that are right hand dominate will be recruited to participate in the study. Subjects will be seated with their left forearm in an arm orthosis attached to the Haptic Master. The left hand will be placed in the wrist exoskeleton mounted to the end effector of the Haptic Master. Subjects will be randomly assigned into 4 groups: Haptic Master control group (HM), Haptic Master with Gripper (HMG), Haptic Master with Wrist (HMW), and Haptic Master with Gripper and Wrist (HMGW). Subjects will be instructed to perform the Pick and Place Task and pick up a virtual cube of varied thickness (0.45, 0.35, and 0.25 in) and place it on the specified target as many times as possible during the 120 s trial. Each subject will participate in 6 training sessions, where each session will consist of 8 trials each. Subjects will perform two sessions per cube size for 8 trials each.
Assessment of a 7-DOF Hand Exoskeleton for Neurorehabilitation
413
4 Conclusion The purpose of this study is to establish the effectiveness of the NJIT HandsOn (7-DOF active system) and the ability to augment learning with haptic feedback in a virtual environment. This research observes the effect of an admittance controlled end effector that provides increased DOF at the wrist and hand. The wrist exoskeleton and modular gripper are able to provide haptic feedback to augment reaching and grasping mechanics in a VE. The virtual Pick and Place Task will assess the efficacy of increased DOF at the hand during training. Increased DOF should provide more biologically appropriate excursions and yield greater success compared to movements without the appropriate DOF. Feasibility for stroke or populations with neurological impairment affecting motor control will be assessed based upon the outcome measures of this study.
References 1. Mozzafarian, D., Benjamin, E.J., Go, A.S., et al.: Heart disease and stroke statistics—2015 update: a report from the American Heart Association. Circulation 131(4), e29–322 (2015) 2. Kwakkel, G., Kollen, B.J., Krebs, H.I.: Effects of robot-assisted therapy on upper limb recovery after stroke: a systematic review. Neurorehabil. Neural Repair 22(2), 111–121 (2008) 3. Loureiro, R., Harwin, W., Nagai, K., Johnson, M.: Advances in upper limb stroke rehabilitation: a technology push. Med. Biol. Eng. Comput. 49(10), 1103–1118 4. Maciejasz, P., Eschweiler, J., Gerlach-Hahn, K., Jansen-Troy, A., Leonhardt, S.: A survey on robotic devices for upper limb rehabilitation. J. Neuroeng. Rehabil. 11, 3 (2014). doi:10.1186/ 1743-0003-11-3 5. Hu, X.: A novel continuous intention-driven rehabilitation robot and its training effectiveness. Biomech. Med. 6. Faria-Fortini, I., et al.: Upper extremity function in stroke subjects: relationships between the international classification of functioning, disability, and health domains. J. Hand Ther. 24(3), 257–264 (2011) 7. Todorov, E., Shadmehr, R., Bizzi, E.: Augmented feedback presented in a virtual environment accelerates learning of a difficult motor task. J. Mot. Behav. 29(2), 147–158 (1997)
Improving the Standing Balance of People with Spinal Cord Injury Through the Use of a Powered Ankle-Foot Orthosis Amber Emmens, Iolanda Pisotta, Marcella Masciullo, Edwin van Asseldonk and Herman van der Kooij
Abstract In this study, our goal was to improve the standing balance of people with a Spinal Cord Injury (SCI) by using a powered Ankle-Foot orthosis acting in the sagittal plane. We tested four different controllers on two SCI subjects that have a lesion at a low level. In the experiments the subjects repeatedly had to recover from pelvis perturbations, while receiving ankle assistive torques from the orthosis. We found that the controllers that use centroidal dynamics as input parameters were able to provide proper support to the subjects after a perturbation had been applied, even though they worked against the subjects after they had recovered from the perturbation. These preliminary results show the potential of balancing controllers that operate in Center of Mass-space.
1 Introduction For people with a spinal cord injury (SCI) who lack ankle motor control, maintaining balance during standing can be difficult, if not impossible. When properly controlled, exoskeletons could help to improve the standing balance of these paraplegics. For people that have an injury at a low level, but who still have hip function, an ankle-foot orthosis (AFO) may already provide sufficient support. SYMBITRON is supported by EU research program FP7-ICT-2013-10 (contract #611626). SYMBITRON is coordinated by University of Twente. A. Emmens (B) · E. van Asseldonk · H. van der Kooij Department of Biomechanical Engineering, University of Twente, Enschede, The Netherlands e-mail:
[email protected] I. Pisotta Laboratory of Robotic Neurorehabilitation, Neurological and Spinal Cord Injury Rehabilitation Department A, Fondazione Santa Lucia, Rome, Italy M. Masciullo Laboratory of Spinal Rehabilitation, Neurological and Spinal Cord Injury Rehabilitation Department A, Fondazione Santa Lucia, Rome, Italy © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_68
415
416
A. Emmens et al.
Our goal is to improve the standing balance of people with a SCI by using a powered AFO. In this case study, we test standing balance in the sagittal plane using various balance controllers, that operate in Center of Mass (CoM) space or joint space. We have implemented these controllers on the Achilles AFO [1] and compared their balancing performance.
2 Materials and Methods Two subjects that have a SCI participated in the balance experiments. Their characteristics are shown in Table 1. The experimental setup consisted of the Achilles exoskeleton; a robotic pusher able to provide systematic perturbations on the trunk [2]; three inertial measurement units (IMUs) placed at the lower leg, the upper leg and the back of the subject; force plates; and a safety harness. Figure 1 shows an overview of the setup. Ethical approval for the experimental protocol was given by the ethical board of Fondazione Santa Lucia, Italy. In the experiments, subjects had to maintain their standing balance, without stepping, while receiving pushes on the pelvis from the pusher. Each subject tested the following controllers: • a Zero-Impedance controller (ZI) (tested twice) • a fixed ankle stiffness (Pankle) Table 1 Test pilot characteristics
Fig. 1 Experimental setup
Subject Sex code
Mass (kg)
Height (m)
Lesion level
ASIA
S02 S03
71 80
1.65 1.78
C7 L3
D D
M M
Improving the Standing Balance of People with Spinal Cord Injury ...
417
• a PD-controller on the CoM (PDCoM) that controls the CoM to a reference location. • a Momentum-based Controller (MBC) that tries to find joint torques such that a certain desired centroidal momentum is obtained [3, 4]. By optimization, torques are found that satisfy constraints on the center of pressure (CoP), joint angle limits and limit torques. In each trial, one of the controllers was tested and three different perturbation sizes were applied by the pusher: 0.02Mg (N), 0.08Mg (N) and 0.14Mg (N), where M is the mass of the subject and g the gravitational constant. Each perturbation was repeated seven times in one trial, or less when subjects got tired before all perturbations were applied. Using the force plates, the torque generated by the subject was estimated and compared to the torque delivered by the Achilles, to evaluate the supportive effect of the Achilles. Furthermore we estimated the CoM based on the IMU data to check when the subjects had restored from a perturbation, that is, when the CoM returned to a steady state.
3 Results Figure 2 shows that in the PDCoM trial, and to lesser extent in the MBC trial, an assistive torque is applied by the Achilles after the onset of a perturbation. The torque induced by the subject in the PDCoM trial is then reduced compared to the ZI case, while the total ankle torque is similar to that in the ZI trial. This means that the subject could supply less ankle torque to maintain balance, because the Achilles was helping. Figure 2a also shows that the Pankle controller only gives a small support torque shortly after a perturbation is applied and the subject needs to provide most of the balancing torque himself. The time it takes for CoM to return to a steady-state value after a perturbation is similar for all three controllers in Fig. 2a, indicating that the balancing performance is also similar. Strikingly, Fig. 2b shows that before the perturbation is applied and after the subject has recovered from a perturbation (stationary state), the torques delivered by the Achilles and the subject have opposite sign, which means that they work against each other.
4 Discussion We found that the PDCoM and MBC could provide a support torque to the subjects, so subjects did not have to apply all the ankle torques necessary for balancing themselves. Although we could not make a clear distinction between the balancing performances of the different controllers, this is a promising results, because these controllers could be beneficial for e.g. people with a SCI that are not able to supply
418
A. Emmens et al.
Fig. 2 Left: torque delivered by Achilles vs. the torque delivered by the subject and right: corresponding CoM trajectories. Each line in the figure represents a response to the largest perturbation size. The perturbation time is indicated with the grey box. a Responses of subject S02 in case of the ZI, Pankle and PDCoM trial. Not enough data was available of the large perturbation responses in the MBC trial. b Responses of subject S03 in case of the MBC trial
all the balancing torques themselves. Ideally, the Achilles complements the torque provided by the subject, but we found that the Achilles in some cases works against the subject. This occurs because the PDCoM and MBC try to bring back the CoM to a certain desired location (measured in a static pose). When this location is different
Improving the Standing Balance of People with Spinal Cord Injury ...
419
from the instantaneous desired CoM location of the subject, as is the case in Fig. 2b shown by the difference between the stationary CoM value and zero, he needs to counteract the Achilles torque. It may not be obvious to subjects that changing their CoM location can reduce the opposing torque. In that sense, it would be better to let subjects practice with the controllers first. It is the first time that these controllers are implemented on an AFO and we expect that the problem of the undesired counteraction can easily be solved in next iterations by not defining a desired CoM location, but a desired range within which the CoM must return and by resetting this range when a new steady-state value is measured.
5 Conclusions We implemented various controllers on the Achilles AFO to improve the standing balance of subjects with a SCI. We found that the controllers working in CoM-space were able to provide support torques to help subjects balancing. This could particularly be useful for paraplegics that can generate little ankle torques themselves. In future work we will improve the balance controllers, based on a centroidal dynamics analysis of standing balance in healthy subjects and extend the controllers to an exoskeleton with more actuated degrees of freedom.
References 1. Meijneke, C., van Dijk, W., van der Kooij, H.: Achilles: an autonomous lightweight ankle exoskeleton to provide push-off power. Biomed. Robot. Biomech. 918–923, Aug. 2014 2. Vlutters, M., Van Asseldonk, E.H.F., Van der Kooij, H.: Center of mass velocity based predictions in balance recovery following pelvis perturbations during human walking. J. Exp. Biol. (to be published) 3. Herzog, A., Righetti, L., Grimminger, F., Pastor, P., Schaal, S.: Momentum-based balance control for torque controlled humanoids. http://arxiv.org/abs/1305.2042v1 (2013) 4. Lee, S.H., Goswami, A.: A momentum-based balance controller for humanoid robots on nonlevel and non-stationary ground. Auton. Robot. 33, 399–414 (2012)
Transparent Mode for Lower Limb Exoskeleton R. Mendoza-Crespo, R. Soto and J.L. Pons
Abstract In this paper a method to reduce the mechanical impedance of the joints of a lower limb exoskeleton is presented. When user is in charge of the motion the exoskeleton mimic its movements. Gravity, Friction and Interaction Force compensators are designed in order to reduce the force necessary to move the exoskeleton joints. Gravity compensation is used to mitigate the effect of the exoskeleton’s weight. This weight adds a force component when the orientation of the limb is different to the gravity vector. The added Friction compensation effect reduces the frictional phenomena of the joints gearboxes. The Interaction Force is calculated from the measured strain of the segments of the exoskeleton. User intention is also detected using the Interaction Force. The gain block adjusts the weights of the friction and interaction compensators depending on the joint velocity. First in this work the context is shown. Followed by the experimental set-up, several compensators and its effects and the control algorithm.
1 Introduction There are several control systems for robotic exoskeletons [1, 2]. Impedance control is one of the most used approach to control robotic systems that interact with humans [3–5]. Other works take into account user intention. Measuring EMG signals the exoskeleton is able to assist walking. This system calculates virtual torques from EMG signals [6]. There are other patient-cooperative approaches. The system detects
R. Mendoza-Crespo (B) · R. Soto Biomechatronics Group, Tecnológico de Monterrey, Monterrey, NL, Mexico e-mail:
[email protected] R. Soto e-mail:
[email protected] J.L. Pons Bioengineering Group, Spanish National Research Council, Madrid, Spain e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_69
421
422
R. Mendoza-Crespo et al.
user’s intention and voluntary efforts and calculates the correspondent commands to the actuators instead of repeating predefined sequences or other rigid strategies [7]. In most impedance control methods there is the fundamental difficulty of finding the appropriate impedance parameters. This is because the physical interaction forces generated by the user to move the exoskeleton changes from one to another. Even it changes with the same user between days [8]. Other implementations use the minimal force interaction approach to make the exoskeleton more transparent to the user [9]. This means that the main goal of the system is to reduce the sensation of forcing the exoskeleton to move the way the user wants to. Making the exoskeleton system is an interesting problem to solve. The dynamics of the system should be canceled in order to have a transparent behavior. Because of the high reduction ratio of the gear boxes it could be difficult to reduce the inertia of the assembly.
2 Experimental Set-Up The H2 robotic exoskeleton has six actuated joints, 3 degrees of freedom per leg in sagittal plane. Flexion and extension movements are taken into account for Hip and Knee joints, and dorsi and plantarflexion for the Ankle joint. The exoskeleton has 2 braces per segment to attach it to the user legs. Each joint has its own low level control and power electronics. The joints have a brushless motor and a harmonic drive to generate each movement. Each segment that couples a joint to the next has a pair of strain gauges to measure the link deformation. Using these, the interaction forces are estimated [10].
3 Gravity Compensation This compensation is explained here as the joints commands necessary to compensate the gravity effect on the exoskeleton. Using direct kinematics with all 6 joint angles known. We are able to calculate the magnitude of the compensation. If velocities and accelerations are zero it can be expressed as following. Ti = G (θi )
(1)
In (1), T is the magnitude of the torque. G is a function that depends on the angular position of the joints θ . i represent the iterator index of each joint. Tgc H i p = g [(m H l H + m K l K + m A l A ) s H ] + g [(m K l K + m A l A ) s H K + m A l A s H K A ]
(2)
Transparent Mode for Lower Limb Exoskeleton
423
Tgc K nee = g [(m K l K + m A l A ) s H K + m A l A s H K A ]
(3)
Tgc Ankle = gm A l A s H K A
(4)
From (2) to (4), T gc is the compensation torque for Hip, Knee and Ankle. g is the gravitational acceleration. m H , m K , m A are the mass of the Hip, Knee and Ankle segments respectively. l is the length of the center mass. And s H , s H K , s H K A are sin function of Hip joint angle, Hip plus Knee angles and the 3 joint angles. The same calculations are made for both legs.
4 Friction Compensation If two surfaces are in contact, friction forces may arise when one or both surfaces are moved. This is a consequence of the irregularities at micro scale. The effects of the friction depends on many factors, like temperature, properties of the material that the surface is made of, presence of lubrication and many others [11]. Friction could be modeled in many different ways. It could be taken into account the static and dynamic effects. The most basic friction model is the so-called Coulomb Friction Model. It greatly simplifies the frictional phenomena. But it does not include dynamic effects. Anyway it is part of more complex models even dynamic ones. It is modeled as a constant force that opposites to motion. Another term that can be added to the model is the Viscous Friction. It models the friction force as a force that is directly proportional displacement velocity. FFC = sign (v) (FCoulomb + FV iscous |v|)
(5)
where FFC is the resultant magnitude of Coulomb and Viscous friction models. FCoulomb Force of coulomb friction. FV iscous Viscous friction force and v is the relative velocity of the surfaces involved.
5 Interaction Force Each exoskeleton joint by itself present mechanical impedance. That is resistance to movement, but it depends on the speed at which the joint is moved. The interaction force is defined as the reaction force between the user’s limb and the exoskeleton braces. It is produced when the user is trying to move its legs with the exoskeleton attached. The magnitude of this force does not produce painful movements but it could be uncomfortable. In consequence the user can not move (walk, sit down, stand up, or climb stairs) freely.
424
R. Mendoza-Crespo et al.
Fig. 1 Gravity TGC , friction TFC and interaction force TI F compensation. External forces are modeled as the torque generated by the user Tuser
The proposed control algorithm minimizes the interaction force between the user’s legs and the robotic exoskeleton straps. This mode is called Transparent Mode. It means that the user perceive the exoskeleton like only carrying a lightweight device attached to its legs. Except for the braces that has to be adjusted to the user’s legs.
6 Control System The proposed control system integrates three terms: Gravity, Friction and Interaction Force compensation in order to reduce the dynamics of the exoskeleton. In consequence the sensation of wearing the exoskeleton is reduced. The gain block adjusts the weights of the friction and interaction compensators depending on the joint velocity. When the joint velocity is zero the ratio between gains is 9:1. With the Torque controller gain set higher. This is because strain gauges array are more sensitive to small movements. In addition even if the joint is not moving at all, it is still possible to calculate the interaction force. When the velocity is different to zero the ratio will be 1:1 for both signals. This means that the friction compensation has bigger gain if the joint is moving. When a change of direction of movement is detected, the ratio changes again. The torque controller gain is 3 times higher than the friction compensation gain. This way the effect of inertia of the motor and the joint segment is reduced. The gains where fine tuned heuristically.
7 Conclusion The compensation strategy has been able to reduce the feeling of the mechanical impedance of the exoskeleton joints. Friction and Interaction force compensation can work alone by itself. But the added effects reflects better behaviors and sensation
Transparent Mode for Lower Limb Exoskeleton
425
when wearing the exoskeleton to the user. The Gain adjust block gave us the opportunity to fine tune the behaviour of the joints depending on the velocity and direction. Using the transparent mode we were able to capture the walking pattern of healthy subjects wearing the lower limb exoskeleton.
References 1. Kazerooni, H., Racine, J.L., et al.: On the control of the Berkeley lower extremity exoskeleton (BLEEX). In: IEEE Int. Conf. on Robotics and Automation, Barcelona, Spain, pp. 4353–4360 (2005) 2. Mendoza-Crespo, R., Gordillo, J.L., Soto, R.: Wearable human lower limb prototype exoskeleton: an operative approach. In: 11th IEEE International Conference on Control and Automation ICCA, Taichung, Taiwan (2014) 3. Lee, S., Sankai, Y.: Virtual impedance adjustment in unconstrained motion for an exoskeletal robot assisting the lower limb. Adv. Robot. 19(7), 773–795 (2005) 4. Aguirre-Ollinger, G., Colgate, J.E., et al.: Active-impedance control of a lower-limb assistive exoskeleton. In: IEEE 10th International Conference on Rehabilitation Robotics, Noordwijk, pp. 188–195 (2007) 5. Unluhisarcikli, O., Pietrusinski, M., et al.: Design and control of a robotic lower extremity exoskeleton for gait rehabilitation. In: IEEE/RSJ Int. Conf. on Intel. Rob. and Sys., San Fran., USA, pp. 25–30, Sept. 2011 6. Lee, S., Sankai, Y.: Power assist control for walking aid with HAL- 3 based on EMG and impedance adjustment around knee joint. In: IEEE Int. Conf. on Intel. Rob. and Sys., Switz., vol. 2, pp. 1499–1504 (2002) 7. Riener, R., Lunenburger, L., et al.: Patient-cooperative strategies for robot-aided treadmill training: first experimental result. IEEE Trans. Neural Syst. Rehabil. Eng. 13(3), 380–394 (2005) 8. Tran, H.-T., Cheng, H., Duong, M.-K., Zheng, H.: Fuzzy-based impedance regulation for control of the coupled human-exoskeleton system. In: Proceedings of the 2014 IEEE International Conference on Robotics and Biomimetics, Bali, Indonesia (2014) 9. Claros, M.J., Soto, R., Gordillo, J.L., Pons, J.L., Contreras-Vidal, J.L.: Robotic assistance of human motion using active-backdrivability on a geared electromagnetic motor. Int. J. Adv. Robot. Syst. (2016) 10. Bortole, M., Venkatakrishnan, A., Zhu, F., Moreno, J., Francisco, G., Pons, J.L., Contreras-Vidal, J.L.: The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study. J. NeuroEng. Rehabil. (2015) 11. Bona, B., Indri, M.: Friction compensation in robotics: an overview. In: IEEE Conference on Decision and Control, and the European Control Conference, Seville, Spain, December 12–15, 2005
Human-Robot Mutual Force Borrowing and Seamless Leader-Follower Role Switching by Learning and Coordination of Interactive Impedance Ker-Jiun Wang, Mingui Sun and Zhi-Hong Mao
Abstract This research developed a bilateral human-robot mutual adaptive impedance control strategy. The developed interactive impedance coordination methods can let human and robot arbitrarily switch the role between leader and follower seamlessly. Also, through iteratively increasing the impedance in the intended moving direction, human and robot can mutually borrow the force from each other to facilitate the task execution.
1 Introduction In recent years, with the growing number of stroke, neurologically-disordered patients, as well as the more and more aging population, scientists are constantly looking for the development of wearable exoskeletons to compensate and improve the human capabilities. In a foreseeable future, human-robot symbiosis on exoskeleton devices will be a common scenario in our society. In designing and controlling the exoskeletons, some critical issues need to be considered. First of all, since the exoskeleton is tightly coupled with the human body, a seamless physical Human-Robot Interaction (pHRI) strategy, which lets human and robot jointly perform motor skills and dynamically react to the world, has to be properly designed. The controller should be a two-way bilateral structure, where both human and robot controllers are in parallel with the system plant, treating each other’s feedback gains as part of their own control plants. Human and robot are constantly learning and adaptively stabilizing the system until reaching a mutually agreed control protocol [1]. Secondly, the controller should be able to negotiate both human and robot intentions in order to decide a consensus assistive modality. In such a way, human and robot can arbitrarily switch the role between leader (active role) and follower (passive role) in necessary circumstances. When either human or robot is weak (or has erroneous motions), the other one can actively K.-J. Wang (&) M. Sun Z.-H. Mao Department of Bioengineering, University of Pittsburgh, Pittsburgh, USA e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_70
427
428
K.-J. Wang et al.
take the lead to guide the overall movements. This strategy is extremely useful when the patients receive active training in the rehabilitation procedures. Finally, to make the controller work, one grand challenge is the human-robot mutual understanding. A biomimic human-like interaction skill should be established into the exoskeleton. As a consequence, human can easily understand exoskeleton’s behavior and interprets its intention. Here, we propose a biomimic human-robot symbiosis framework to address the above issues. We have developed a human-robot co-adaptive optimal controller to realize the bilateral cooperation behavior [1]. This method can find out the cooperative impedance both of the human and the robot. Another important contribution of this paper is that we developed a human-robot mutual impedance coordination strategy from observing the behaviors of human and human playing Chinese Tai Chi Chuan [2], an ancient martial art with abundant physical interactive tactics. It exemplifies how to use minimum efforts to steer the opponent’s limbs by exploiting the intrinsic biomechanical properties while the limb is moving. Under the impedance coordination strategies of Chinese Tai Chi Chuan, human and robot can arbitrarily switch the roles between leader and follower seamlessly. In addition, through iteratively increasing the impedance in the intended moving direction, human and robot can mutually borrow the force from each other to facilitate the task execution.
2 Human-Robot Mutual Force Borrowing and Seamless Leader-Follower Role Switching Human impedance characteristics when interacting with the environment have been investigated intensively for a long time [3, 4]. Different human impedance in static or dynamic environment has been measured or tested [5] to understand how human control their movement to stabilize the interaction. Until recently, many researchers start to teach the robot human-like impedance in order to make the interaction more effective [6, 7]. However, these approaches still lack the fundamental knowledge of how and when human generates different impedance coordination strategies to react to different interaction situations. Here, in this research, we observe the behavior of human-human playing Chinese Tai Chi Chuan and generalize the rules of impedance coordination strategies. In Fig. 1, we have illustrated the human arm endpoint impedance profile. One interesting finding is that the human stiffness ellipse is proximately aligned with the direction of arm’s movement, meaning that the human has large impedance in the direction of motion, but has low impedance in the lateral direction perpendicular to that motion. This finding is exactly how human utilizes his interactive impedance when playing Chinese Tai Chi Chuan. The players are constantly borrowing force from each other in their common intended direction, and change the interactive
Human-Robot Mutual Force Borrowing …
429
Fig. 1 Illustraion of human arm endpoint impedance in different motion directions and under different external disturbances. a Stiffness ellipse profile when subject moves in X-Axis direction [5]. b Stiffness ellipse when subject maintain a constant force against perturbations with different magnitudes in different directions [5]. c Stiffness geometry interprets a correlation with the direction of motion. The major principle axis (i.e., maximum stiffness) direction of the ellipse is roughly aligned with motion direction. d Stiffness ellipse profile when moving in Y-Axis direction [5]
behavior in the lateral direction by tweaking the traits of human impedance characteristics. On the intended motion trajectory, human will use his arm stiffness ellipse to stabilize each transient goal at every time instance. The attractive force generated from the stiffness ellipse of the next transient goal will drive his arm moving forward. The force borrowing strategy is achieved by generating additional attractive force by his opponent in the next-to-the-next transient goal according to the prediction of his opponent. Human and robot can iteratively execute this technique to realize the mutual force borrowing strategy. In the mean while, when human and robot have different intended motion trajectories with each other, the Leader-Follower role switching condition can be triggered by generating an attractive force in the direction of minor principle axis of the estimated stiffness ellipses of human or robot, where it has the lowest stiffness to resist the motion. To be noticed here, since the Leader-Follower role switching also involves cognitive level interaction. The switching condition we provided here is simply for the stable physical interaction. After the motion direction has been changed, the human or robot could still decide whether they will keep the same role
430
K.-J. Wang et al.
Fig. 2 Human-robot impedance interaction strategy to achieve force borrowing and Leader-Follower role switching. Base on the predicted human trajectory and the estimated human stiffness ellipses, robot can decide how to interact with human according to its own synthesized stiffness ellipses, and vice versa for human to interact with robot. Leader-Follower role switching condition is valid only when one is moving in another one’s weakest point (i.e., minor principle axis of the ellipse)
or not. The constant role switching at this Leader-Follower role switching condition will provide a seamless interaction behavior. The overall strategy can be found in Fig. 2.
3 Co-Adaptive Optimal Control Framework To realize our impedance interaction strategy, we build up a human-robot co-adaptive optimal controller [1]. Human and robot are modeled as two adaptive optimal controllers in parallel with the system plant, i.e., coupled human-robot dynamics. The objective of the two controllers is to negotiate their intentions and cooperatively stabilize the system to achieve a common goal. At here, in order to reduce the online co-adaptation time required for iteratively adjusting the feedback gains of both human and robot, we build up a virtual physical human-robot interaction model inside the robot brain (see Fig. 3). This virtual model can virtually interact with each other and directly predict the future coordinated consensus ^h; K ^ r . These matrices behavior, interpreted as cooperative stiffness matrices, K can be used to cooperate with the real entity, or generate the desired impedance
Human-Robot Mutual Force Borrowing …
431
Fig. 3 Human-robot co-adaptive optimal control framework with build-in virtual pHRI model. We assume that robot has all the necessary sensors to collect data. Through the data collections of fur ; uh ; xg, we can use the illustrated method to learn the virtual plant dynamics x_ ¼ ^f ður ; uh ; xÞ, ^ ;R ^ . The the prediction of human intention ^ vih , and the human optimal control cost values Q h h ^ h; R ^ h , we cost values are estimated using inverse optimal control method [8]. After we obtain Q can reconstruct the virtual human optimal controller to interact with the virtual plant as well as the robot optimal controller. The virtual interaction will eventually reach to a stable consensus behavior that we can calculate by solving iterative Linear Quadratic Regulator (iLQR). The consensus cooperative behavior is calculated as feedback gain or interpreted as cooperative ^r . ^h; K stiffness matrices K
interaction behavior discussed in Sect. 2. How to learn this virtual interaction model can be find in Fig. 3. Here, we use Extended Kalman Filter (EKF) to predict the human intended trajectory.
References 1. Wang, K.J., Sun, M., Xia, R., Mao, Z.H.: Human-robot symbiosis framework on exoskeleton devices. In: 2016 IEEE International Conference on Industrial Technology (ICIT), Taipei, pp. 1500–1506 (2016) 2. Cheng, M.C.: Master Cheng’s new method of Tai Chi Chuan self-cultivation (translated by Mark Hennessy). North Atlantic Books, Berkeley, CA (1999) 3. Hogan, N.: Impedance control: an approach to manipulation: Part i, ii, iii. J. Dyn. Syst. Meas. Control 107(1), 1–24 (1985) 4. Burdet, E., Osu, R., Franklin, D.W., Milner, T.E., Kawato, M.: The central nervous system stabilizes unstable dynamics by learning optimal impedance. Nature 414(6862), 446–449 (2001)
432
K.-J. Wang et al.
5. Tee, K.P., Burdet, E., Chew, C.M., Milner, T.E.: A model of force and impedance in human arm movements. Biol. Cybern. 90(5), 368–375 (2004) 6. Rozo, L., Calinon, S., Caldwell, D., Jimenez, P., Torras, C.: Learning collaborative impedance-based robot behaviors. In: AAAI Conference on Artificial Intelligence, North America, June 2013 7. Howard, M., Mitrovic, D., Vijayakumar, S.: Transferring impedance control strategies between heterogeneous systems via apprenticeship learning. In: Proc. 2010 IEEE-RAS International Conference on Humanoid Robots (2010) 8. Anderson, B.D.: The inverse problem of optimal control. DTIC Document, Fort Belvoir, VA, USA, Tech. Rep. 6560-3 (1966)
Upper Limb Exoskeleton Control for Isotropic Sensitivity of Human Arm Rok Goljat, Tadej Petriˇc and Jan Babiˇc
Abstract Most of today’s assistive devices are controlled to provide uniform assistance irrespectively from the configuration of the human arm and the direction of the movement. We propose an innovative control method for arm exoskeletons that takes into account both of these parameters and compensates the anisotropic property of the force manipulability measure, intrinsic to the biomechanics of the human arm. To test our controller we designed a set of reaching tasks where the subjects had to carry two different loads to targets at five different locations and of two different sizes. Reaching times and trajectories were analysed for the evaluation of the controller. Through the analysis of the average reaching times we found that our method successfully enhances the motion while the analysis of the average maximal deviation from the ideal trajectories showed that our method does not induce any additional dynamic behaviour to the user.
1 Introduction Robotic wearable devices such as exoskeletons are being developed to either augment the abilities of healthy individuals or to improve the condition of those with impaired physical abilities [1, 2]. A common approach to control such devices is to measure interaction dynamics and use the inverse dynamics models to either directly amplify the forces produced by the human muscles or to generate task dependent The work presented in this paper was supported by the European Unions’ Horizon 2020 research and innovation programme under grant agreement No. 687662—SPEXOR. R. Goljat (B) · T. Petriˇc · J. Babiˇc Automation, Biocybernetics and Robotics Department, Jozef Stefan Institute, Ljubljana, Slovenia e-mail:
[email protected] T. Petriˇc e-mail:
[email protected] J. Babiˇc e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_71
433
434
R. Goljat et al.
trajectories [3]. Although this approach efficiently augments human motion, it does not take into account the mechanical characteristics such as the highly anisotropic manipulability of the human arm. The aim of this paper is to propose a feed-forward control approach that augments the motion of the human arm by transforming the anisotropy of the arm manipulability [4, 5] in such a way that the axes become equal. Instead of amplifying the forces exerted by the human hand equally in all directions, our control approach produces forces in such a way that the manipulability ellipsoid becomes a circle.
2 Materials and Methods For the design of our control approach we considered two biomechanical parameters of the human arm: the manipulability measure which is based purely on the arm kinematics, and the mobility measure which takes into account both the kinematics and the dynamics of the arm.
2.1 Manipulability The manipulability measure takes the uncertainty of joint angles and transforms it to the uncertainty in end-point position. By assuming that the joint sensors are noisy with variance σ 2 and independent from each other then the covariance of uncertainty is transformed from joint space to covariance of uncertainty at the end-point: Cov(x) = JCov(q)JT = σ 2 M.
(1)
Here, x and q are the end-point and joint uncertainties and J is the Jacobian of the current arm configuration. In (1), the matrix M = JJT
(2)
represents the manipulability matrix [6] that shapes the independent joint noise into Cartesian end-point noise.
2.2 Mobility The mobility measure is based on the instantaneous response of the arm to dynamic perturbations [7]. If the inertial matrix H of the arm is known, we can then define the mobility measure as (3) W = JH−1 JT ,
Upper Limb Exoskeleton Control for Isotropic Sensitivity of Human Arm
435
where W represents the end-point mobility matrix [8]. Because mobility also includes the inertia of the arm, it can be considered as a more precise measure of arm sensitivity.
2.3 Control Method The eigenvectors of mobility and manipulability matrices represent directional sensitivity of the arm [7]. Specifically, it is the easiest to control the end-effector position and sense perturbations along the axis of the minor eigenvector and the hardest in the direction of the major eigenvector. If we invert the sensitivity matrices, than they relate to the joint torque and end-effector force. In this case, the major eigenvector points in the direction where it is the easiest to exert the end-effector force and the minor eigenvector points in the direction where it is the hardest to exert the endeffector force. To counteract these anisotropic properties we propose a novel control method for arm exoskeletons that transforms the anisotropic sensitivity of the arm to the isotropic sensitivity. Using singular value decomposition, the sensitivity matrices can be represented as ellipses with the major and minor axes representing the major and minor eigenvectors. To augment the user exerted force and obtain isotropic sensitivity, the assistive force Fa needs to equal the user exerted force Fu amplified by the ratio between ellipse major axis Maxis and the projection of the user force Faxis on the ellipse (4). Fa =
||Maxis || − 1 Fu ||Faxis ||
(4)
3 Experiments and Results We implemented the proposed controller in a simulation environment and compared its performance for both mobility and manipulability measures. For the manipulability measure, the controller was also experimentally evaluated on 5 subjects. Figure 1 shows the forces generated by the controller if the user exerted 1 N forces in three different positions of the arm. Due to the thin shape of the ellipse on the edges of the workspace, larger assistive forces were exerted there than in the central region of the workspace. For the safety reasons, the maximum assistive force was limited to 9 N. Subjects performed unassisted and assisted load carrying motions to five different targets of two sizes (20 mm, 5 mm). The motions were performed on a transverse plane at the shoulder height. The experiment consisted of three consecutive sessions. In the first session, the subjects were unassisted and had to move a 3 kg object (baseline session). In the second session, the subjects were also unassisted but had to move a heavier object weighting 30 kg (heavy session). In the final session the
436
R. Goljat et al.
Arm position
0.5 -0.5
0 -9 9
Fy [N]
1.5
y [m]
Controller
9
Fy [N]
y [m]
1.5
0.5 -0.5
0 -9 9
Fy [N]
1.5
y [m]
Fig. 1 Plots in the left column show the inverted manipulability (blue) and mobility (red) ellipsoids for tree different positions of the arm inside the workspace (black). The dotted lines show the normalized ellipsoids. Plots in the right column show the assistive forces of the proposed controller for the corresponding configurations in the left column. The arrows represent the assistive forces for the manipulability measure (blue) and the mobility measure (red)
0.5 -0.5 -1
0
0 -9
1
-9
0
9
Fx [N]
x [m]
subjects had to move the 30 kg object but were assisted by the proposed controller (assisted session). Table 1 shows the average reaching times and average maximum deviations of motions performed by subjects during the three sessions and for both target sizes. When the subjects were asked to move a heavy load, their average reaching times were larger than during the baseline session. When they were asked to move a heavy load while being assisted by the controller, their average reaching times were close to the times during the baseline session. Decreasing the target size increased the overall time of the movements but did not have any notable effect on the motion trajectory.
Table 1 Experimental results Trial (target size) Baseline (large) Heavy (large) Assisted (large) Baseline (small) Heavy (small) Assisted (small)
Average time (s)
Average deviation (m)
0.51 0.76 0.58 0.98 1.26 1.05
0.024 0.028 0.018 0.030 0.028 0.025
Upper Limb Exoskeleton Control for Isotropic Sensitivity of Human Arm
437
4 Conclusions We developed and evaluated a novel controller to augment the motion of the human arm based on its manipulability and mobility measures. By simulations and an experimental study we demonstrated that the proposed control method successfully augments human capabilities without introducing any additional disturbances to the human motion. In the future we plan to carry out an extensive experimental study involving both manipulability and mobility measures and further expand the sensitivity model to generalize it for more complex tasks.
References 1. Dollar, A.M., Herr, H.: Lower extremity exoskeletons and active orthoses: challenges and stateof-the-art. IEEE Trans. Robot. 24, 144–158 (2008) 2. Peternel, L., Noda, T., Petriˇc, T., Ude, A., Morimoto, J., Babiˇc, J.: Adaptive control of exoskeleton robots for periodic assistive behaviours based on EMG feedback minimisation. PLoS ONE 11(2), 02 (2016) 3. Peternel, L., Petriˇc, T., Oztop, E., Babiˇc, J.: Teaching robots to cooperate with humans in dynamic manipulation tasks based on multi-modal human-in-the-loop approach. Auton Rob 36(1), 123– 136 (2013) 4. Cos, I., Belanger, N., Cisek, P.: The influence of predicted arm biomechanics on decision making. J. Neurophysiol. 105(6), 3022–3033, Jun 2011 5. Yamashita, M.: Robotic rehabilitation system for human upper limbs using guide control and manipulability ellipsoid prediction. Procedia Technol. 15, 559–565 (2014) 6. Yoshikawa, T.: Foundations of robotics: analysis and control. MIT Press, Cambridge, MA (1990) 7. Sabes, P.N., Jordan, M.I.: Obstacle avoidance and a perturbation sensitivity model for motor planning. J. Neurosci.: Off. J. Soc. Neurosci. 17(18), 7119–7128, Sept. 1997 8. Hogan, N.: Impedance control: an approach to manipulation: part ii implementation. J. Dyn. Syst. Meas. Control 107(1), 8–16 (1985)
AUTONOMYO: Design Challenges of Lower Limb Assistive Device for Elderly People, Multiple Sclerosis and Neuromuscular Diseases A. Ortlieb, M. Bouri and H. Bleuler
Abstract Lower limb wearable robotics also known as exoskeleton or power suit is a booming field of research. Potential medical applications cover a large range of gait disorders from rehabilitation to assistance in daily mobility. Surprisingly, or not, paraplegia seems to be the first target of all commercialized exoskeleton. In this paper we will try to understand this choice and look at other disorders leading to the inability to walk. Neuromuscular, autoimmune or neurological diseases such as muscular dystrophy, multiple sclerosis or stroke, can lead to similar gait disorders and are mostly incurable today. SCI (Spinal Cord Injury) symptoms are quite dissimilar from theirs and reveal specific design challenges. Existing devices’ architecture and human-robot interaction are presented and discussed in terms of adaptation toward non-SCI disorders.
1 Introduction Trends in lower limb assistive devices currently target mostly SCI (Spinal Cord Injury) patients. Over the four commercialized products (Ekso from Eksobionics, ReWalk from ReWalk robotics, REX from REX bionics and HAL from Cyberdyne) targeted users are unanimously SCI patients [1–4]. However, based on
This study is co-financed by ASRIMM (Association Suisse Romande Intervenant contre les Maladies neuro-Musculaires) and FSRMM (Fondation Suisse de Recherche sur les Maladies Musculaires). A. Ortlieb (&) M. Bouri H. Bleuler Laboratory of Robotic Systems, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland e-mail:
[email protected] M. Bouri e-mail:
[email protected] H. Bleuler e-mail:
[email protected] © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_72
439
440
A. Ortlieb et al.
the National Health Interview Survey from 2012, about 7 % of the adult US population (17 million people) reported not to be able to walk a quarter of a mile (*400 m) or with major difficulties [5]. The same year, the number of SCI patients was estimated to be about 270,000 [6], which represents 1.59 % of the population under the walking autonomy of 400 m. The two questions that come out from this observation are: why does the industry of medical exoskeleton target mostly SCI patients while the market seems 60 times bigger and what are the other disorders that lead to a poor walking autonomy?
2 Walking Impairments Various effects can severely affect the walking ability. First, Blackwell et al. [5] reports that the percentage of people with walking disability largely increases with age, where the median is about 64 years old. The elderly population has thus the larger representation in gait impairments. Ma et al. [7] reported that in rehabilitation perspectives there exist 8 dominant cause of disability in the US: back pain, osteoarthritis, rheumatoid arthritis, stroke, traumatic brain injury, amputation, multiple sclerosis and SCI. In terms of prevalence (in the US) [7], osteoarthritis is the main cause of disability for 8.6 million adults (2012), rheumatoid arthritis affected 1.3 million adults in 2005, stroke 6.8 million people (2013), traumatic brain injury 3.32 million (2005), lower limb amputation 1.6 million (2005), multiple
Fig. 1 Illustration of the different systems of the body involved in the walking ability process
AUTONOMYO: Design Challenges …
441
sclerosis 400,000 (2009) and SCI *270,000 (2012). From this non-exhaustive list, neuromuscular diseases (NMD) such as muscular dystrophy are not mentioned. Based on the incidence in Northern Ireland [8] extrapolated to the US adult population in 2012, the number of NMD patients is estimated to be about 80,000. The nature of the disability is essential for the design of the assistance. Figure 1 below maps the different origins of the walking impairment in the human body.
3 Trends of Existing Medical Devices As described in the introduction, actual medical exoskeletons focus mainly on SCI patients with applications as for personal use at home or for rehabilitation in clinics. These devices basically propose two designs and different but comparable strategies of interaction.
3.1
Human-Robot Interaction
Current strategies in medical exoskeletons are mostly called mobilization strategies. What underlies this term is that the motion trajectories made by the device follow a given pattern independently from the user will. However, the user can smartly decide on the activity and the timing of the motion based on different approaches such as a hand controlled joystick, the trunk motion or the position of the center of mass (CoM). The current trend in human-robot interaction follows a high level control where the interactions between the user’s legs and the exoskeleton are unidirectional.
3.2
Design Architecture
Two architectures are currently observed among existing devices: a quasi-full actuation architecture with more than five actuated degrees of freedom (DOFs) on each side of the device and the minimal actuation architecture with two actuators each side. The minimal actuation architecture locks all DOFs at the ankle joint, while the hip and knee flexion/extension are motorized. For the quasi-full actuation architecture, only the internal/external rotations can be non-actuated. As a consequence, the minimal architecture cannot be self-stabilizing and the user needs an external support (crutches or a rollator) to maintain balance. If the quasi-full architecture allows balance control and free hands, it is heavier, bulkier and slower up to a factor 10.
442
A. Ortlieb et al.
Another aspect of current devices is the poor back drivability of the actuation-transmission design which is however coherent with the “unidirectional” interaction presented above.
4 Challenges Toward Assistive Devices As described in section II, SCI represents a small part of walking impairments. However frail elderly people, multiple sclerosis and neuromuscular diseases share several similarities: A need for (assisted) daily activity to prevent the deterioration of the walking ability, residual but localized and variable strength in the lower limbs, weak upper limb muscles and a high risk of falls due to low balance control. To address these disorders, a new type of exoskeleton needs to be developed in accordance with the following criteria.
4.1
Human-Robot Interaction
In opposition with SCI patients, targeted populations need a strong bidirectional interaction between the user’s legs and the device. This interaction will strongly depend on the residual control and strength at each joint of the user. The interaction should be at a lower control level to allow the real time synchronization between the user will and the exoskeleton motion. At the same time, the device should be able to evaluate and predict the risk of falls and react over the user in case of emergency.
4.2
Design Architecture
The first element is that to be able to have a bidirectional interaction, the actuation and transmission system requires a high transparency or back drivability. The other aspect is that the device should be very flexible to be able to adapt each joint power to the impairment of the user in order to keep a lightweight and allow a high degree of back drivability. Finally, the exoskeleton’s architecture should be complete enough to augment the balance of the user in the frontal and lateral directions while keeping the number of actuators as low as possible to reduce the weight of the device. The proposed design in the frame of the AUTONOMYO project comprises high torque, efficient and low inertia motors combined with planetary gearboxes with an efficiency about 80 %. Each motor-gearbox chain is standard so that it can easily be changed to better suits the user’s need for each joint. All actuators and batteries which represents about 70 % of the mass of the device are placed, thanks to cable
AUTONOMYO: Design Challenges …
443
transmission, about the pelvis which makes the robotic lower limbs very lightweight and allows high dynamics.
5 Conclusion While commercialized medical exoskeletons currently target SCI patients, a wide range of other walking impairments could benefit from different and more adapted assistive exoskeletons. The solution proposed allows an important level of back drivability while affording high dynamics and torque. The possibility to easily adapt the hardware to different joints requirements largely increases the scope of user.
References 1. Ekso Bionics - An exoskeleton bionic suit or a wearable robot that helps people walk again (Online). http://intl.eksobionics.com/. Accessed 30 Mar 2016 2. ReWalk 6.0 – Home: ReWalk – More Than Walking (Online). http://rewalk.com/. Accessed 30 Mar 2016 3. REX Bionics - Step into the Future: Rex Bionics (Online). http://www.rexbionics.com/. Accessed 30 Mar 2016 4. CYBERDYNE (Online). http://www.cyberdyne.jp/. Accessed 30 Mar 2016 5. Blackwell, D., Lucas, J., Clarke, T.: Summary health statistics for U.S. adults: National Health Interview Survey, 2012. National Center for Health Statistics (2014) 6. 2012 NSCISC Annual Statistical Report Complete Public Version: National Spinal Cord Injury Statistical Center (2012) 7. Ma, V.Y., Chan, L., Carruthers, K.J.: Incidence, prevalence, costs, and impact on disability of common conditions requiring rehabilitation in the United States: stroke, spinal cord injury, traumatic brain injury, multiple sclerosis, osteoarthritis, rheumatoid arthritis, limb loss, and back pain. Arch. Phys. Med. Rehabil. 95(5), 986–995.e1, May 2014 8. Hughes, M.I., Hicks, E.M., Nevin, N.C., Patterson, V.H.: The prevalence of inherited neuromuscular disease in Northern Ireland. Neuromuscul. Disord. 6(1), 69–73 (1996)
Passive Lower Back Moment Support in a Wearable Lifting Aid: Counterweight Versus Springs Ralph J.L. Macke, Arvid Q.L. Keemink and Arno H.A. Stienen
Abstract Passive wearable lifting aids support workers by applying gravity force compensation at the arms. In this study we investigated the feasibility of a compensatory lower back moment, generated by a practically constant spring force (38.5 Nm), extending the lower back by pushing on the upper leg. This design is proposed as a light-weight solution to generate lower back moments. The method is compared to using counterweights at a different distances. We recorded EMG activity of the erector spinae longissimus (ES) muscle, the perceived workload (NASA TLX) and the preference of 12 subjects. Results showed no significant difference in ES peak EMG activity during the task, and no significant difference between perceived workload between conditions, as we expected. However, 10 out of 12 subjects indicated preferring the spring mechanism over both counterweights. The main reason of preference was the reduction of weight and inertia of the system. Therefore, the proposed constant spring force mechanism is a feasible alternative to counterweights.
1 Introduction Heavy lifting is a strong contributor to low back injuries, in part due to the spinal loading [4]. Part of this load stems from trunk muscles with a short moment arm that need to provide high forces to stabilize the spine. The authors would like to thank Hankamp Rehab for providing the SaeboMAS and the SaeboMAS Mini passive lifting devices. This research is supported by the Dutch Technology Foundation STW, which is part of the Netherlands Organisation for Scientific Research (NWO), and which is partly funded by the Ministry of Economic Affairs. Project Number: 12162. R.J.L. Macke · A.Q.L. Keemink (B) · A.H.A. Stienen Department of Biomechanical Engineering of the University of Twente, Enschede, Netherlands e-mail:
[email protected] A.H.A. Stienen Department of Physical Therapy and Human Movement Sciences, Northwestern University, Chicago, USA © Springer International Publishing AG 2017 J. González-Vargas et al. (eds.), Wearable Robotics: Challenges and Trends, Biosystems & Biorobotics 16, DOI 10.1007/978-3-319-46532-6_73
445
446
R.J.L. Macke et al.
Passive wearble lifting devices to reduce this burden on its user can therefore be very valuable for the health of the user. The passive FORTIS (Lockheed Martin Corporation, Bethesda, MD, USA) uses a counterweight to balance the moment induced by the supported load carried by the user. Devices that use a spring mechanism only provide a supportive force when the user is flexing their trunk (much like a torsion spring) such as the Personal Lift Assist Device (Queen’s University, Kingston, ON, Canada) [1, 3] or the Laevo (Laevo B.V., Delft, Netherlands). This is a major difference compared to using a counterweight, which continuously provides compensation, but also makes the device heavier. In this study we investigated if springs are a feasible alternative to counterweights, when providing an almost constant moment compensation, which different from torsion spring behavior. We measured electromyography (EMG) activity on the lower back, specifically the erector spinae longissimus (ES), during a lifting and lowering task. This study also investigated subject preference for the various moment compensation methods using a questionnaire and the NASA Task Load Index (TLX).
2 Materials and Methods A balanced repeated measures design was used where subjects performed a symmetric lifting task in three varying conditions. Subjects were asked to lift an 11.2 kg load while being assisted in gravity compensation force for 50 % by a passive lifting assistive device (SaeboMAS Mini, Saebo, Charlotte-NC, USA). During lifting, subjects were supported and the lower back with a counter moment of 38.5 Nm. In this experiment three different ‘moment compensation’ conditions were compared: a constant spring force mechanism, a light counterweight of 6 kg at a distance of 0.54 m, and a heavy counterweight of 12 kg at a distance of 0.27 m. The lifting aids are shown schematically in Fig. 1 and the real setup is shown in Fig. 2. Subjects were instructed to lift once per minute the load from a table onto a plateau and after 30 s lower the load back onto the table, for a total of 10 repetitions per condition. The table had a height of 79.5 cm and the plateau had a height of 120.7 cm; both measured from the ground. Twelve healthy subjects (11 male, 1 female) were recruited from the university population. Average subject characteristics were 23.3 ± 1.8 years of age and 185 ± 7 cm in length. Subjects were aware of the goal of the study and blinding was not possible. We met institutional requirements and informed consent was obtained from all subjects. EMG of the ES was measured using the Delsys Trigno wireless EMG system (Delsys Inc., Boston, MA). Electrodes were placed 3 cm lateral from the spinous process of the L3 [2]. An online tool was used to fill in the NASA-TLX. Subjects were asked after the experiment which compensation method they preferred, and why. From the EMG data, peak muscle activity was investigated, because this would relate to peak spinal compression. The envelope of the EMG signal was calculated by phase-lag free (Matlab filter filtfilt) high-pass filtering (10 Hz 4th order
Passive Lower Back Moment Support in a Wearable …
447
Fig. 1 Schematic view of the assistive device with only relevant forces to show the contribution to moments around and forces in the lower back. A load Fload is held at distance dload . A gravity compensator (its weight and reaction force omitted for clarity, see Fig. 2) assists with force Fgc . The human generates a residual lifting force Fhand at the hand and an extension moment Mlb at the lower back. The moment in shoulder is omitted for clarity. a The spring mechanism provides an almost constant force at the leg-interface Fsp,nor mal , over distance d, assisting the extension moment, irrespective of the angle made with the rest of the suit. b The counterweight generates a gravity force Fcw , at a distance d, assisting the extension moment. The downward force experienced at the hip is higher than in (a) due to the counterweright Fcw Fig. 2 The gravity compensation device on the upper back assists in lifting 50 % of the weight. a The spring mechanism pushes on the upper legs to generate an assistive moment in the lower back. b a heavy counterweight of 12 kg delivers an assistive moment in the lower back (a) Spring Mechanism
(b) Counterweight
Butterworth), full wave rectification and phase-free low-pass filtering (3 Hz 2nd order Butterworth). Peak muscle activity for the left ES was calculated per lift or lowering as the highest peak of the EMG envelope. Due to the lack of space, the right ES data is omitted. EMG data was normalized to a % maximum voluntary isometric contraction (MVIC), which was determined before the experiment. From the 10 task repetitions, the compound mean peak activity was determined. This measure, as well as the overall workload score from the NASA-TLX, were compared in a repeated measures ANOVA.
448
R.J.L. Macke et al.
Fig. 3 Peak activity for the left erector spinae longissimus during lifts (a) and lowers (b) and the overall score for the NASA Task Load Index (TLX) (c). The mean is indicated by the red circle, the median by the red line. No statistical difference was found between the different moment compensation methods for either lifts, lowers, or the workload. CW: Counterweight
3 Results Repeated measures ANOVA showed that the peak EMG activity was not significantly affected by the type of moment compensation provided. For the left ES during the lifts F(2, 22) = 2.78, p > 0.05. For the left ES during the lowers F(2, 22) = 1.21, p > 0.05. The results show that the overall TLX score was not significantly affected by the type of moment compensation provided, V = 0.17, F(2, 10) = 1.01, p > 0.05. On the question which of the three methods was most preferred, 10 out of 12 subjects answered the springs mechanism. One subject prefered the heavy, and one prefered the light counterweigt. The main motivation for preferring the spring mechanism was its low weight, compared to the counterweights and its reduced inertia during (especially rotational) movements.
4 Discussion Results show no obvious differences between the mean peak EMG activity for the type of moment compensation provided; this holds for both the lifts and the lowers. Mean peak EMG activity is somewhat lower for the lowers than for the lifts. This is likely due to subjects first pulling the load towards themselves before lifting the load from the plateau and placing it down on the table. The workload, quantified by the NASA TLX, also shows very little difference between the different moment compensation methods. There was no significant difference found for the workload. This does not correlate to the subjective preference of the subjects, since 10 out of 12 subjects indicated
Passive Lower Back Moment Support in a Wearable …
449
to prefer the spring mechanism over either of the counterweights. This, and the fact that no statistically significant difference in peak EMG was found, shows that a constant spring force mechanism is a feasible and prefered alternative to heavy counter-weights.
References 1. Abdoli, E.M., Agnew, M.J., Stevenson, J.M.: An on-body personal lift augmentation device (PLAD) reduces EMG amplitude of erector spinae during lifting tasks. Clin. Biomech. 21(5), 456–465 (2006) 2. van Dien, J.H., de Looze, M.P., Hermans, V.: Effects of dynamic office chairs on trunk kinematics, trunk extensor EMG and spinal shrinkage. Ergonomics 44(7), 739–750 (2001) 3. Sadler, E.M., Graham, R.B., Stevenson, J.M.: The personal lift-assist device and lifting technique: a principal component analysis. Ergonomics 54(4), 392–402 (2011) 4. Waters, T.R., Dick, R.B., Krieg, E.F.: Trends in work-related musculoskeletal disorders: A comparison of risk factors for symptoms using quality of work life data from the 2002 and 2006 general social survey. J. Occup. Environ. Med. 53(9), 1013–1024 (2011)
