System-Level Simulations of Selected Aspects of 5G Cellular Networks Chair of Wireless Communications Poznan University of Technology {krzysztof.bakowski, marcin.rodziewicz, pawel.sroka}@put.poznan.pl Abstract — This paper presents a system-level simulation approach developed at the Poznan University of Technology for the evaluation evaluation of selected selected aspects aspects of next generation generation wireless wireless systems. The details of the considered environment and propagation models are given, as well as several important aspects of the simulator implementation methodology. Selected use cases of the presented simulation platform are demonstrated in relation relation to differen differentt research research topics topics in the area area of next generation wireless systems. The achieved simulation results show sho w tha thatt the pro problem blemss related related to the the develop developmen mentt of modern modern wireless systems require detailed and precise modeling of the environment in order to accurately accurately validate system performance. Keywords—system-level simulations, 5G, interference, D2D, relaying
I.
I NTRODUCTION
Societal Socie tal developm development ent will will lead lead to a chan change ge in the requirements for a modern wireless system. An avalanche of mobile and wireless traffic volume is predicted, with a thousand-fold increase over the next decade [1]. The fact that the current 4G technology will not be capable of carrying this increase in data consumption imposes the necessity to develop develop new solutions in wireless communication technologies. Moreover, Moreov er, today’s dominant dominant scen scenarios arios of human-centric human-centric communication will be complemented by a tremendous increase in the numbers of communicating machines leading to the so-called Internet-of-Things (IoT). Very diverse requirements on mobile 5G wireless communication systems are expected, including: stringent latency and reliability requirements, ultra-high data rates (up to multiple Gbps), as well as network scalability and flexibility. Therefore, it is presumed that a single new radio access technology will not be able to satisfy all these requirements and replace today’s networks, netw orks, but as shown shown in Fig. 1, a flexible combinati combination on of evolved evolv ed existing technolog technologies ies and new radio concepts concepts will have to be used instead [1]. In order to assess whether the new radio technologies fulfill the 5G system requirements, system capabilities have to be evaluated with respect to several aspects. These may include the cell spectral efficiency, bandwidth scalability, cell edge user spectral efficiency, latency, mobility, and many other parameters. A tool commonly used in such an evaluation 978-1-4673-6540-6/15/$31.00 ©2015 IEEE
is performing system-level simulations, where where simplified link models are usually applied for reducing both complexity and simulation time. Simultaneously, these models should still be accurate enough to allow capturing the influence of link characteristics on system performance metrics. Therefore, there is a need to find a tradeoff between excessive simulation run time (complexity) (complexity) and the model's imperfections. This paper paper presents presents a 5G system-level system-level evaluati evaluation on tool developed at the Poznan University of Technology in the framework of the METIS project and its possible applications in the assessment assessment of performance of different aspects aspects of 5G systems. The rest of the paper is organized as follows: Section II presents the employed environmental model with the main focus on propagation characteristics characteristics and mobility. In Section III, the details details of system-leve system-levell evaluation evaluation tools are given, given, describing a model of the foreseen 5G system system and radio link representation. Section IV outlines examples of the evaluation evaluation of selected aspects of 5G systems performed using the simulation tool. Finally, Section V concludes the paper. Wireless backhaul D2D V2V/M2M
Wireless backhaul
Moving Relay
Micro BS
Macro BS
Macro BS Fixed Relay
Fig. 1.
Wireless Wirel ess communi communicatio cation n systems systems beyo beyond nd 2020 2020
II .
E NVIRONMENT AND CHANNEL MODELING
Emerging 5G technologies set numerous new challenges to the evaluation framework. One of the projects strongly focused on 5G technologies and their evaluation was the METIS project, project, where where a cons consortium ortium defined defined a detailed detailed 5G evaluation eval uation framew framework ork which is capture captured d in [2, 3]. In this framework, the evaluation of wireless technologies is scenario-driven. The foreseen scenarios are described in [2], with accompanying test cases. To enable the evaluation of technologies investigated in specific test cases, simulation guidelines were formulated [3]. These guidelines contain, among other things, the details of channel, environment and
mobility modeling, which are described in the following paragraphs. A. The METIS channel models
The METIS channel models are defined as the so-called propagation scenarios (PS). These propagation scenarios can be divided into three sub-groups based on transmitter and receiver locations. These comprise: outdoor–to-outdoor (O2O) propagation, outdoor–to-indoor (O2I) propagation indoor–to-indoor (I2I) propagation. The sub-groups of propagation scenarios can be further divided depending on the type of the communication link, i.e., macro base station (MaBS) to user equipment (UE) links, micro base station (MiBS) to UE links, femto base station (FeBS) to UE links and UE to UE links. The MaBS to UE link propagation scenarios correspond to a situation where the BS is located over a building rooftop. The MiBS to UE link propagation scenarios are defined for BSs located below building rooftops where the dominant part of the propagation is due to the reflections between buildings. The FeBS to UE scenarios capture the indoor propagation characteristics, whereas the UE to UE link is defined to model the propagation for direct device-to-device (D2D) communication. Taking into account the aforementioned division, eight propagation scenarios have been identified that represent different communication paths [3]. The propagation scenarios have been defined with two important assumptions in mind. Firstly, a realistic and not synthetic scenario is required. Secondly, a 3D propagation model should be utilized in system evaluations. In the currently commonly used propagation models, LoS or NLoS conditions are randomly selected. However, for scenarios that are realistic, the sight conditions between transmitter and receiver should be evaluated on a real-time basis.
B. Environment model
The propagation scenarios could not exist without a suitable environment model. For this purpose, several models have been defined within METIS [3]. The most detailed one is the Madrid Grid Model (MGM). The MGM is a realistic urban environment model that can be used particularly in the evaluation of technologies related to Ultra Dense Networks (UDN). Its realism is achieved by considering different environments of buildings, roads, parks, bus stops, subway entrances, sidewalks and crossing lanes, etc., all in three dimensions. The model is based on the structure of the city of Madrid and captures typical European city characteristics.
also an enabler to model user clustering (e.g., a group of pedestrians waiting for a green light) in the network [3]. III.
SIMULATION TOOLS
In order to be able to perform a comprehensive evaluation of 5G technology proposals, a set of simulation tools is required. Typically, simulation tools can be divided into linklevel simulators and system-level simulators encompassed by an appropriate environment and mobile radio channel emulators. Therefore, we have developed a set of software simulation tools that facilitate the evaluation of 4G and 5G radio access technologies [4, 5]. The MACHINE (Mobile networks evaluation tools with Advanced CHannel and INterference modEling) software simulation package consists of link-level and system-level simulators together with a linkto-system interface utilizing the Mutual Information-based Exponential SNR Mapping (MIESM) methodology. Moreover, two standardized evaluation frameworks considering the orthogonal frequency division multiple access (OFDMA) technique are implemented: IMT-Advanced and METIS-MGM for 4G and 5G systems evaluations, respectively. In this paper, we focus on the system-level simulator that can be utilized to model heterogeneous networks, including relay or D2D links. Support for simultaneous uplink and downlink simulations is provided as well. Moreover, the tool is capable of simulating different bandwidth and carrier frequency setups in a single network deployment. The simulation tool realistically models the overheads of the physical (PHY) layer and allows for a flexible specification of the PHY layer parameters, such as subcarrier spacing, the number of subcarriers per resource block, etc.. The implemented schedulers utilize modulation and coding schemes (MCSs) and associated transport block size (TBS) tables generated automatically for the specified PHY layer parameters (also for multilayer transmissions). Finally, the simulator can be configured to be compliant with the LTE system specification. It can utilize both hexagonal grid- or map-based deployment scenarios. For the simulation of possible solutions for 5G cellular networks described in the subsequent paragraph, the set of parameters presented in Tab. 1 has been used. Fig. 2 presents a sample deployment in a simplified METIS-MGM scenario.
C. Mobility models
The environment model is complemented with a mobility model. Mobility models should reflect the nature of movement in the environment. In the METIS project, mobility is modeled in two ways. In the first approach, users located outdoors, either pedestrian or in-vehicle users, move along the streets making random decisions on the next direction of movement on street crossings. The second approach is different for the pedestrian users. A source-destination solution is used where each user is moving to a certain destination by selecting the shortest path. The realism of mobility is improved by considering traffic lights in the environment. This approach is
Fig. 2. Example of deployment in a simplified MGM scenario (diamonds - macro BS, circles - micro BS/relays, trianges - UEs)
TABLE I. Parameter Station types Number of stations Transmit power [dBm] Cell range expansion [dB] Bandwidth [MHz] Angenna gain [dB] Tilt [degrees] Environment model UEs deployment
IV.
SIMULATION PARAMETERS Value Macro BS Micro BS Relays 3 18 9 46 37 30 0 5 5 20 20 20 17 17 0 12 0 0 simplified METIS TC 2 (see Fig. 2) indoor - 70 %, pedestrians - 15 %, in cars - 15%
SELECTED ASPECTS OF A 5G SYSTEM
One of the most challenging scenarios considered in the METIS project was the Ultra Dense Network (UDN), which is characterized by a very high density of UEs. Such a scenario imposes several challenges for a wireless system design, such as the provisioning of ubiquitous access to all UEs at the possibly low cost or mitigation of interference resulting from the densification of heterogeneous BSs that reuse the same frequency bands. Moreover, the presence of direct D2D communications and two-way relaying introduces new interference patterns that have to be handled. Therefore, in this section, we demonstrate examples of the evaluation of interference mitigation aspects, two-way relaying and direct D2D communications in UDN using the system-level tool described in previous sections. A. Long-term interfence mitigation
The current 3GPP LTE proposal for long-term interference mitigation in the downlink adopts time-domain muting mechanisms known as Almost Blank Sub-Frames (ABSF), where the macro-cell to small cell interference is reduced through muted time frames [6, 7]. However, the enhanced inter-cell interference coordination (eICIC) approach focuses mostly on improving the performance of small cell-edge users, thus impacting the overall performance of the system. As an alternative to eICIC, a game-theoretic approach has been proposed in the framework of the METIS project [8, 9], that relies on a resource sharing scheme, where each BS selects an action that specifies its maximum downlink transmit power in a predefined time-frequency grid. To further improve the performance of the proposed approach, the exchange of information on selected actions and related utilities is performed periodically. To evaluate the performance of the game-theoretic solution in comparison to state-of-the-art systems, a periodic update of each BS transmit power in the time-frequency resource grid is performed every 10 TTIs. The time-frequency resources are divided into three groups in frequency (e.g., for a 20-MHz bandwidth, this corresponds to groups comprising 33, 33 and 34 resource blocks) and two slots in time (each slot corresponding to 5 TTIs). For each of the six groups, a maximum power level is specified out of the set {P high , Plow}. A total of 16 different actions are specified for each BS. The game-theoretic solution has been evaluated in two versions employing two concepts from the game theory:
Correlated equilibrium (CE) – where the optimization process is based on the values of total data rate achieved in each cell [9]. Satisfaction equilibrium (SE) proposed in [10], where the optimization is based on the satisfaction levels (related, e.g., to minimum rate requirement) of UEs attached to BSs. Apart from the results for the game-theoretic solutions, also the performance of the following configurations of the system has been evaluated. These configurations are: A system with no interference mitigation (no ICIC) A system with eICIC proposed for LTE-Advanced in two configurations – with 2 (LTE-A (2 ABSF)) and 4 (LTE-A (4 ABSF)) ABSFs used per 10 sub-frames. Fig. 3 presents the average cell downlink (DL) spectral efficiency (understood as the total throughput per unit bandwidth) achieved with the investigated system configurations. The best performance is observed for the game-theoretic solution based on CE. However, one can notice that with the increase in UE density, the LTE-A eICIC scheme performance approaches the result obtained with CE, which indicates that the use of the time-frequency resource sharing approach gains diminish with the increasing traffic load.
1.5 ] z H / 1.25 s p b [ y c n 1 e i c i f f e l a 0.75 r t c e p s l l e 0.5 c e g a r e v 0.25 A
0
no ICIC LTE-A (2 ABSF) LTE-A (4 ABSF) CE SE 260
520
780
1040
Fig. 3. Average cell spectral efficiency in the DL achieved for scenarios with different numbers of UEs
Apart from the presented results, it is also possible to include other metrics in performance evaluation when using the presented simulator, e.g., the average BS transmit power, average BS energy efficiency or cell spectral efficiency distribution. One should notice that the simulations of an ultra-dense metropolitan network in a challenging environment show that the techniques proposed for 4G systems and their evolution do not necessarily improve the performance of the considered heterogeneous system. On the other hand, the use of alternative techniques can provide significant gains in terms of overall system performance. B. D2D resource allocation
Typically, the D2D communications underlay may use radio resources simultaneously with cellular users (CUEs). However, enabling D2D communications reusing the cellular spectrum poses new challenges, for instance, in terms of new interference patterns. The D2D underlay of the cellular network should not have a negative impact on cellular
transmissions. On the other hand, cellular transmissions should not degrade the performance of D2D communications. Thus, an interference mitigation scheme should be applied when introducing the D2D communications in the system. DT
CC
spectral efficiency on both UL and DL links for most cases. The exception is the case of 260 UEs where the baseline system achieves a better UL spectral efficiency than the system with a D2D underlay. This is due to the fact that the user density in this case is relatively low for the considered BS deployment. Comparing the D2D resource allocation methods, one can observe that LRA has a better performance than RRA for the evaluated cases. The significant difference is visible in the D2D link analysis, which implies that the LRA enables a better protection from interference for DUEs. 2 1.8
DR
BS
Fig. 4. Geometry taken into account in selection of CUE for resource sharing.
One solution for the management of interference inherent to D2D communications is the location-based resource allocation (LRA) method that has been proposed in the METIS project [8]. In this proposal, user location information is used to select CUEs that will share resources with D2D UEs (DUEs). The selection of the CUE for resource reuse is performed by a central entity aiming at distance maximization. Assuming that the large-scale fading including path loss and shadowing has the largest impact on interference, increasing the distance between the nodes that share resources also minimizes the experienced interference. Fig. 4 presents the geometry considered for the sharing CUE selection procedure when cellular UL resources are used for D2D communications. A more detailed description of the proposed LRA can be found in [8]. The advantage of this resource allocation (RA) method is that it requires no channel state information from the DUEs, since location information is used. Moreover, a centralized approach gives more control of D2D links and allows for a better control of interference related to D2D communication. To evaluate the performance of D2D communications, a D2D control node (DCN) is added to the network, and the transmission mode of 10% of CUEs is set to the D2D mode. The DUEs can be deployed in three possible locations, i.e., in buildings, pavements and cars, taking into account the maximum allowed distance between DUEs forming a D2D pair set to 100 m. The macro BS and micro BS, considered in the evaluated scenario, are operating on different carrier frequencies, i.e., 800 MHz and 2600 MHz, respectively. We assume that DCN is associated only with the macro BS, and the D2D users are allowed to share UL resources with CUEs. In each TTI, DCN performs D2D RA by selecting CUEs that will share resources with DUEs. The set of CUEs available for selection is limited to CUEs that are scheduled for transmission in the current TTI. In the evaluation scenario, two D2D resource allocation methods were investigated, namely, the proposed LRA and random resource allocation (RRA) which randomly selects the CUEs for resource sharing. Additionally, the performance achieved by the system when these methods are utilized has been compared with a scenario where D2D communication is not supported. Fig. 5 presents the average cell spectral efficiency achieved by the considered system for three types of links and three user density settings. One can see that the i ntroduction of D2D communications in the network increases the overall
1.6
] l l e c \ 1.4 z H \ s p1.2 b [ y c n 1 e i c i f f E0.8 l a r t c0.6 e p S
No D2D, 260 UEs D2D, RRA, 260 UEs D2D, LRA, 260 UEs No D2D, 520 UEs D2D, RRA, 520 UEs D2D, LRA, 520 UEs No D2D, 1040 UEs D2D, RRA, 1040 UEs D2D, LRA, 1040 UEs
0.4 0.2 0
UL
DL
D2D
Fig. 5. Spectral efficiency achieved for different links in the considered user density set-ups.
C. Two-way relaying
Relaying has been standardized as part of 4G radio access networks, namely LTE-Advanced and IEEE 802.16m. Primarily, they were standardized as a coverage extension technique. The main drawback of the current relaying technology applied in TDD systems is the half-duplex mode of operation. In this section, we consider a cellular system in which the relays are capable of two way transmission realized in two time slots, which is equivalent to the full-duplex mode of operation. This type of relaying technology has been extensively investigated in the METIS project [11]. In the investigation, we consider a broadband system utilizing the OFDMA multiple access technique and TDD duplex. The relays are capable of performing half-duplex transmission but operate in two phases called the multipleaccess (MA) and broadcast (BR). In the MA phase, both relayed UE and the BS transmit simultaneously to that relay. Therefore, the relay must apply detection techniques, such as Successful Interference Cancellation (SIC) or MIMO processing to jointly detect two data streams. In the BR phase, relays transmit jointly the two data blocks received in the MA phase. Then, both the BS and UE can utilize the knowledge of the previously transmitted data block to cancel the inter-block interference from the received signal (i.e., using the network coding principle) [12]. When deployed in a cellular network, such relays aggregate some of the traffic but still operate simultaneously with directly connected users. Therefore, a new kind of interference appears in such a network. The exact interference analysis for the two cells separately for the MA and BR phases is shown in Fig. 6 and Fig. 7, respectively. In the analyzed scenario. we assume that the directly connected users and relay-connected users in the same cell use orthogonal resource blocks. Therefore, interference can occur only between different cells. In the MA phase, relays and UEs directly connected to BS (CUEs) receive the signal if scheduled. The interference from another cell can be caused by another BS or relay-connected UEs (RUEs). In the BR
phase, BSs and UEs connected to relays receive the signal from them. Thus, the interference from another cell can be caused by CUEs or relays. It should be noted that UEs operating in a different mode (directly connected to BS vs. served by relays) experience D2D-like interference. The current implementation of the resource allocation scheme does not consider jointly DL/MA and UL/BR phases. Although two-way relaying in the TDD mode requires a similar or almost equal data flow in uplink and downlink directions, this is not considered. This separation allows us to observe the two phases independently. In Fig. 8, the achieved spectral efficiency is shown for different numbers of full-buffer users. When the number of users increases, the average downlink spectral efficiency increases, whereas for the uplink, the trend is reversed. This result suggests that in the UL/BR phase, the densification of UEs increases interference. Therefore, the mitigation of interference in this phase is more crucial for the overall system performance. BSi
MA PHASE RNl
CUE j
RUEk
RX
TX
RX
V.
CONCLUSIONS
In this paper, we have presented a system-level simulation tool developed at the Poznan University of Technology in the framework of the METIS project for the evaluation of future 5G wireless systems. A description of the environment and channel models used in evaluation has been given. Moreover, the simulation of selected research aspects of future 5G systems using the described tool has been presented. The simulation results indicate that the approach to environment and channel modeling proposed in the framework of the METIS project can highlight additional aspects of the impact of the deployment scenarios on system performance.
Acknowledgment This work has been funded by the Polish Ministry of Science and Higher Education within the status activity task DS-81-152 in 2015. Part of this work has been performed in the framework of the FP7 project ICT-317669 METIS, which is partly funded by the European Union.
TX
References BSm OR
OR
[1]
A. Osseiran et al., "Scenarios for 5G mobile and wireless communications: the vision of the METIS project," IEEE Communications Magazine, vol.52, no.5, pp.26-35, May 2014
[2]
METIS consortium, “D1.2 – Initial channel models based on measurements”, project report, Apr. 2014
[3]
METIS consortium, “D6.1 – Simulation guidelines”, Oct. 2013
OR
RNp RUEo
CUEn
RX
Fig. 6.
TX
RX
TX
[4] MACHINE - a set of software simulation tools for mobile networks rsztaty Telekomunikacyjne PWT’
Multiple access phase interference BSi
BR PHASE RNl CUE j
TX
RUEk
RX
TX
[5]
Evaluation of Radio Interface Technologies for IMT-Advanced and Beyond", Ch. 13 in: "Simulation Technologies in Networking and Communications: Selecting the Best Tool for the Test", CRC Press, 2014
[6]
3GPP TS 36.300, “Evolved Universal Terrestrial Radio Access (EUTRA) and Evolved Universal Terrestrial Radio Access Network (EUTRAN); Overal description”, v. 11.3.0, September 2012
[7]
LG Electronics, “3GPP TSG RAN WG1 Meeting #62 R1-104661
[8]
METIS consortium, “D4.3 - Final report on network-level solutions”, project report, Feb. 2015
[9]
P. Sroka, A. Kliks, “Distributed Interference Mitigation in Two-Tier Wireless Networks Using Correlated Equilibrium and Regret-Matching learning”, Proc. European Conference on Networks and Communicatons 2014, Bologna, Italy, June 23-26, 2014
RX
BSm
RNp CUEn
TX
Fig. 7.
RUEo
RX
TX
RX
Broadcast phase interference
[10] S. M. Perlaza, H. Tembine, S. Lasaulce, M. Debbah,“Quality-of-Service Provisioning in Decentralized Networks: A Satisfaction Equilibrium Approach”, in IEEE Journal on Selected Topics in Signal Processing, vol. 6, pp. 104-116, 2012 [11] METIS consortium, "D3.3 - Final performance results and consolidated view on the most promising multi-node/multi-antenna transmission technologies", project report, Feb. 2015 [12] -way Relaying for 5G Systems - Comparison of Network Coding and MIMO Techniques", IEEE Wireless Communications and Networking Conference, IEEE WCNC 2014, 6-9 April 2014, Istanbul, Turkey
Fig. 8.
Spectral efficiency for cellular network with two-way relays
