INTRODUCTION An Automated Highway System is a vehicle and road based system that can drive a vehicle automatically. This is done using sensors that serve as the vehicle’s eyes, determining lane position and the speed and location of other vehicles. Actuators on the throttle, brake, and steering wheel give the vehicle the commands that a driver normally would. AHS vehicles often also have equipment to communicate with other AHS vehicles. The concept of an Automated Highway has been around for a long time. eneral !otors displayed a working model of an automated highway at the "#$# %orld’s &air in 'ew (ork )ity. Automated trains have been in use since the "#*+’s. ut it has not been until recently that the technology has become available to build automated highways and vehicles.
1 AUTOMATED HIGHWAY HIGHWAY SYSTEMS
WHY AUTOMATED HIGHWAYS? Safety 'inety percent of all vehicle accidents result from driver-related factors /nattentiveness, inability to respond quickly enough, or bad driving decisions. 0ver 1+,+++ people die annually in motor motor vehicle vehicle accide accidents nts and proper property ty damage damage is estim estimate ated d at over 2"3+ billio billion. n. 4lectr 4lectroni onicc systems that never take over some or all of the 5river’s responsibilities are the most promising method of reducing these accidents. Analyses of causes show that highway accidents could be reduced between $" and 63 percent with the use of electronic collision avoidance technologies and fatal crashes could be reduced by at least 3+ percent.
Efficiency The 'ational highway system carries eighty-nine percent of passenger ground miles traveled and thirty-two percent of the ton-miles of commercial freight travel. The number of vehicles on the roads has doubled in the last ten years while highway capacity has for the most part, remained the same. )onsequently, the average speed on urban highways and crowded corridors during rush hour is $* miles per hour. This loss of time and productivity costs the 7S 23+ billion annually. 0ne solution that engineers are e8ploring now is an Automated Highway System 9AHS:. y reducing or eliminating driver error and placing cars more closely together safely on the highways, highway throughput could be doubled or tripled.
Air Quality AHS has the potential to reduce traffic delays and traffic ;ams. As a result, fewer vehicles will be stopped stopped in traffic where they continue continue to pollute pollute while while they aren’t going anywhere. Smoothing Smoothing out the flow of traffic would also reduce fuel consumption. AHS could eliminate the
HISTORY 2 AUTOMATED HIGHWAY HIGHWAY SYSTEMS
An early representation of the driverless car was 'orman el eddes>s &uturama e8hibit sponsored by eneral !otors at the "#$$ %orld>s &air, which depicted electric cars powered by circuits embedded in the roadway and controlled by radio. The history of autonomous vehicles starts in "#?? with the Tsukuba !echanical 4ngineering @ab in apan. 0n a dedicated, clearly marked course it achieved speeds of up to $+ kmBh 9C+ miles per hour:D by tracking white street markers 9special hardware was necessary, since commercial computers were much slower than they are today:. /n the "#6+s a vision-guided !ercedes-enE robot van, designed by 4rnst 5ickmanns and his team at the undeswehr 7niversity of !unich in !unich, ermany, achieved "++ kmBh on streets without traffic. Subsequently, the 4uropean )ommission began funding the 6++ million 4uro 47F4GA rometheus ro;ect on autonomous vehicles 9"#6?"##3:. Also in the "#6+s the 5AFA-funded Autonomous @and Iehicle 9A@I: in the 7nited States achieved the first road-following demonstration that used laser radar 94nvironmental Fesearch /nstitute of !ichigan:, computer vision 9)arnegie !ellon 7niversity and SF/:, and autonomous robotic control 9)arnegie !ellon and !artin !arietta: to control a driverless vehicle up to $+ kmBh. /n "#6?, HF@ @aboratories 9formerly Hughes Fesearch @abs: demonstrated the first off-road map and sensor-based autonomous navigation on the A@I. The vehicle travelled over *++m at $ kmBh on comple8 terrain with steep slopes, ravines, large rocks, and vegetation. /n "##1, the twin robot vehicles Ia! and Iita-C of 5aimler-enE and 4rnst 5ickmanns of 7niw! drove more than one thousand kilometers on a aris three-lane highway in standard heavy traffic at speeds up to "$+ kmBh, albeit semi-autonomously with human interventions. They demonstrated autonomous driving in free lanes, convoy driving, and lane changes left and right with autonomous passing of other cars. /n "##3, 5ickmannsJ re-engineered autonomous S-)lass !ercedes-enE took a "*++ km trip from !unich in avaria to )openhagen in 5enmark and back, using saccadic computer vision and transputers to react in real time. The robot achieved speeds e8ceeding "?3 kmBh on the erman Autobahn, with a mean time between human interventions of # km, or #3K autonomous driving. Again it drove in traffic, e8ecuting maneuvers to pass other cars. 5espite being a research system without emphasis on long distance reliability, it drove up to "36 km without human intervention. /n "##3, the )arnegie !ellon 7niversity 'avlab pro;ect achieved #6.CK autonomous driving on a 3+++ km 9$+++-mile: L'o hands across AmericaL trip. This car, however, was semiautonomous by natureM it used neural networks to control the steering wheel, but throttle and brakes were human-controlled. &rom "##*C++", Alberto roggi of the 7niversity of arma launched the AF0 ro;ect, which worked on enabling a modified @ancia Thema to follow the normal 9painted: lane marks in an unmodified highway. The culmination of the pro;ect was a ;ourney of C,+++ km over si8 days on the motorways of northern /taly dubbed !ille!iglia in Automatico, with an average speed of #+ kmBh. #1K of the time the car was in fully automatic mode, with the longest automatic stretch 3 AUTOMATED HIGHWAY SYSTEMS
being 31 km. The vehicle had only two black-and-white low-cost video cameras on board, and used stereoscopic vision algorithms to understand its environment, as opposed to the Llaser, radar - whatever you needL approach taken by other efforts in the field. Three 7S overnment funded military efforts known as 5emo / 97S Army:, 5emo // 95AFA:, and 5emo /// 97S Army:, are currently underway. 5emo /// 9C++": demonstrated the ability of unmanned ground vehicles to navigate miles of difficult off-road terrain, avoiding obstacles such as rocks and trees. ames Albus at '/ST provided the Feal-Time )ontrol System which is a hierarchical control system. 'ot only were individual vehicles controlled 9e.g. throttle, steering, and brake:, but groups of vehicles had their movements automatically coordinated in response to high level goals. /n C++C, the 5AFA rand )hallenge competitions were announced. The C++1 and C++3 5AFA competitions allowed international teams to compete in fully autonomous vehicle races over rough unpaved terrain and in a non-populated suburban setting. The C++? 5AFA challenge, the 5AFA urban challenge, involved autonomous cars driving in an urban setting. /n C++6, eneral !otors stated that they will begin testing driverless cars by C+"3, and they could be on the road by C+"6. /n C+"+ Iis@ab ran I/A), the Iis@ab /ntercontinental Autonomous )hallenge, a "$,+++ km test run of autonomous vehicles. The four driverless electric vans successfully ended the drive from /taly to )hina via the arriving at the Shanghai 48po on C6 0ctober.
II. ISSUES IN AHS
4 AUTOMATED HIGHWAY SYSTEMS
resently, the field of AHS faces some important problems and needs to resolve some important issues. The 7S50T, )ongress, the auto industry and consumers are asking the same questions. Three of the most important issues that challenge AHS areM A. !i8ed AutomatedB!anual Traffic or 5edicated AHS @anes . @ow @evel or High @evel of /ntelligent Highway /nfrastructure ). @iability This section will e8plain each issue and its history, give the pros and cons of differing opinions, and discuss what is being done now to solve each problem. The outcome of these conflict points will determine the ultimate success or failure of the AHS program. /t is important that policy makers and engineers work together to find solutions that are technologically feasible, socially acceptable, and that accomplish the goals set forth by )ongress. !uch research has been done and is being done now on these three issues as they relate to human factors, economic impact, land use impact, traffic safety and efficiency, and cost to benefit ratios. An overview of that research will be presented here.
A. MIED AUTOMATED!MANUA" DEDICATED AHS "ANES
TRA##IC
OR
De$icate$ "ane% There is a question whether automated vehicles should drive on normal roadways with manual drivers or on separate lanes built for automated use only. )ost and equity must be weighed with safety and efficiency to find an elegant solution to this question.
Hi%t&ry &or the first few years of AHS under the /ST4A legislation, the predominant thinking among engineers was to plan for dedicated lanes. !any factors influenced that decision. !ost importantly, they did not know how to deal with the random factor of human drivers. 4ngineers knew that if they could control and coordinate every vehicle on a set of dedicated lanes, traffic accidents and slowdowns could be virtually eliminated. The ideal traffic situation would be one where every lane changes and merging maneuver, acceleration and braking is controlled by computers. This would eliminate the human error that causes accidents and slows traffic. The early thinking was that these benefits could not be gained in a mi8ed traffic scenario because human drivers behave randomly, so engineers planned to use AHS only on dedicated lanes. 0ne more reason for dedicated lanes was that proposed systems of that time 9"##"-#3: were very infrastructure dependent. lanners thought that automated lanes would need to be highly specialiEed, therefore should be reserved for automated vehicles, and not crowded by manual vehicle
Safety feature%
5 AUTOMATED HIGHWAY SYSTEMS
After it was decided that dedicated lanes would be necessary, there was a question of how those lanes would be separated from the rest of the highway and how automated vehicles would enter and e8it the lanes. !ost engineers agreed that concrete barriers with chain-link fencing above them would provide the necessary isolation from manual traffic. They chose such drastic separation because of the potential difficulty that automated vehicles could have with road obstacles 9stalled cars, animals, or parts thereof:. Automated vehicles cannot detect and avoid obstacles with nearly as much efficiency as human drivers can. 4ngineers also determined that entrance onto the dedicated AHS lanes should require some kind of vehicle system diagnostic check. efore a vehicle could enter the lanes, the highway infrastructure should communicate with the vehicle to verify that its communications equipment, navigation computer and vehicle control actuators are fully functional. /t would be unsafe for a malfunctioning vehicle to enter automated traffic. 48iting the automated lanes would require a check that the driver is awake and prepared to take control of the vehicle.
C&%t% The largest factor that led engineers to consider a mi8ed traffic scenario was cost. /t was estimated that dedicated lanes could cost up to 2$+ million per mile for design, right of way acquisition, and construction. !any urban areas that could benefit most from AHS simply have no room to add more lanes to e8isting highways. &rom an economic point of view, the costBbenefit ratio of AHS must be compared to conventional highways and mass transportation systems. At 2$+ million per mile and with benefits that will not be realiEed until many AHS equipped vehicles are on the roads, dedicated lanes face large financial and political barriers. /n contrast, the )alifornia ATH 9artners for Advanced Transit and Highways: program on the /"3 corridor has demonstrated a mi8ed traffic AHS infrastructure for around 2"+,+++ per mile.
E'uity Another recently acknowledged problem with dedicated lanes is the potential lack of equity. rototype AHS equipped vehicles have up to 2C++,+++ of automation equipment. This includes numerous sensors to detect lane position, location of other vehicles and road conditions, actuators for the throttle, brake and steering, e8tensive communications gear, and multiple entium processors. %hen the systems become commercially available, carmakers hope to offer AHS equipment as a 2"+++-2C3++ option on new vehicles or perhaps more for upgrade of an older vehicle. /t is unlikely that the average consumer will want to invest that much money in a system that could only be used in the few places that have dedicated AHS lanes. The wealthy and elite may be the only ones to buy the AHS option. /f that becomes the case, dedicated AHS lanes could become a publicly funded highway used only by those wealthy enough to afford AHS equipped vehicles. This is a ma;or political and social barrier to dedicated lanes.
Infra%tructure nee
%$6 AUTOMATED HIGHWAY SYSTEMS
The last and most important problem with the use of dedicated lanes is a question inherent in all infrastructure dependent ;umps in technology. )ities do not want to invest 2$+ million per mile to build automated lanes before automakers have made available and people have purchased vehicles to drive on those lanes. And automakers do not want to build and people do not want to buy AHS vehicles before dedicated lanes are built. &or dedicated lanes to become a reality, large infrastructure investments would have to be made on the hope that the vehicles would quickly occupy the AHS lanes. )ities and private investors are unwilling to take that risk this early in the technology development. 0n the flip side of that coin, consumers will not buy cars with the AHS option in hopes that they may get to use it someday.
Mi(e$ Traffic All of these factors have led engineers and planners to more closely consider a mi8ed traffic scenario. !i8ed traffic would solve the problems of cost, equity, and a large infrastructure investment. However, some of the safety and efficiency benefits would be lost or reduced. 4ngineers had hoped with dedicated lanes to completely eliminate traffic accidents. /n a mi8ed traffic scenario, the most they could hope for is that no automated vehicles would be the cause of an accident. 4fficiency would also be reduced since high speed, close-spaced platooning could cause a safety haEard in mi8ed traffic.
)lat&&nin* latooning is considered the best way to increase highway capacity and decrease delays. Automated vehicles will have the ability to communicate to each other information about speed, headway 9distance between vehicles:, braking ability and other factors that would enable them to follow each other very closely under automated control. raking and acceleration actions could be communicated electronically so that even with very short headways 9N"m:, rear-end collisions could still be avoided. 4ngineers estimated that platooning could allow up to three times as many vehicles per lane compared to manual traffic. There are also aerodynamic advantages of platoons which would add fuel mileage and emissions advantages. 0bviously platooning raises questions about safety in a platoon of mi8ed vehicle types with largely variable braking ability or platooning on wet or icy road conditions. There has been much research into ideal spacing distance and estimation of braking ability. /f platoons were to be used in a mi8ed traffic scenario, some restrictions would be necessary. &irst, the number of vehicles in a platoon might have to be limited. A platoon of si8 or more cars could block any mi8ed vehicles from entering their lane. /f that large platoon was driving in the right lane, it could prohibit other vehicles from merging onto the highway or changing lanes to get off the highway. This is because drivers of automated vehicles are less likely to make room for other vehicles since that would require them to override automatic control to make room for manual vehicles. /n many cases, this would be an annoyance to other drivers. /n some cases, it could cause serious accidents.
C&nclu%i&n %ithout the advantage of increased efficiency through platooning, the only advantage of using automation in mi8ed traffic is improved safety. Although safety is an important improvement, it 7 AUTOMATED HIGHWAY SYSTEMS
may not be enough to ;ustify investment. The costBbenefit ratio may be too low for government and consumers to make an investment, especially since the value of added safety is difficult to measure. However, mi8ed traffic intelligent vehicles may be an important first step in the use of AHS that will lead to the building of more and more dedicated lanes. Iehicles should be able to run in either scenario, but perhaps would have some functions limited while driving in mi8ed traffic. As more AHS equipped vehicles are on the roads, cities and states will begin to build dedicated lanes which will ma8imiEe the efficiency of the highway system. /f the new dedicated lanes only displaced non-AHS lanes and were not entirely new roadways, there would also be significant cost savings over the construction of new lanes. B.
H&+ %,&ul$ Traffic Intelli*ence an$ inf&r-ati&n e $i%triute$ et+een /e,icle% an$ t,e r&a$+ay?
The two necessary ingredients of an automated highway system are information and intelligence. /nformation about other vehicles and the road and intelligence to make decisions based on that information. The information gathering system and the decision making system must be distributed between the road and the vehicles. /deally, there would be some overlap for redundancy. 4ngineers must balance cost, performance, reliability, and safety when deciding what should be the functions of the infrastructure and the vehicles.
Intelli*ent infra%tructure /%. intelli*ent /e,icle% Infra%tructure intelli*ence The e8treme case of an intelligent infrastructure AHS system would be one where the roadway has complete control over every vehicle. /n this type of system, each vehicle would require some communication equipment and actuators on the throttle, brake and steering. Iehicles would become
0e,icle Intelli*ence The other e8treme would be the complete absence of new infrastructure. /n this type of system, sensors onboard the vehicles would need to read lane striping, sense other vehicles and obstacles, and communicate with other AHS equipped vehicles for merging and platooning maneuvers. There are, in fact, prototype vehicles under testing that use video imaging to read lane striping, radar or laser ranging systems to determine pro8imity and relative velocity of other vehicles, and communications systems that enhance safety by coordinating some actions and sharing information about vehicle locations. This type of system would require more comple8 sensors and onboard computing than an infrastructure dependent system. 8 AUTOMATED HIGHWAY SYSTEMS
C&%t% /f most of the intelligence is onboard the vehicles, the cost of the system will be passed on to the consumers who buy those cars. 0n the other hand, a highly intelligent highway infrastructure could cost millions, or even billions of ta8payer dollars. 4ngineers need to balance vehicle cost, infrastructure cost and total system cost in the design of AHS systems. Since the distribution of costs and money are largely tied up in these issues, all of the stakeholders have strong opinions.
A$/anta*e% an$ $i%a$/anta*e% &f Intelli*ent Infra%tructure %y%te-% The advantages of an infrastructure dependent system are safety and efficiency. 0ne single controlling computer can make better decisions than hundreds of onboard computers which are all trying to communicate with each other to make their own decisions. /nfrastructure systems also tend to be more reliable since siEe, e8ternal appearance and "C volts of electrical potential generally limit onboard systems. 0nboard systems are also vulnerable to harsh environments including dirt, vibrations and inclement weather 9ground based systems can be shielded from these effects:. 0ne other consideration is the quality of onboard sensors available. &or e8ample, if the range of any available forward-looking sensors is smaller than the stopping distance of the vehicle, those sensors cannot prevent a collision with a fi8ed ob;ect in the roadway. Totally infrastructure dependent systems have a higher cost than vehicle dependent systems. Since the roadways will be paid for by ta8payer dollars, another question of equity is involved. oorer citiEens should not have to pay for highways that can only be taken advantage of by the wealthy who can afford AHS equipped vehicles. &inally, the most important question of infrastructure dependent systems is system integrity. )ould a hacker break into the system and cause traffic delays, accidents, or even fatalitiesO To what e8tent would the flow of traffic be affected if a car were to veer off the road and strike a communications beaconO /f all of the traffic intelligence lies in the infrastructure, AHS vehicles cannot drive without it.
A$/anta*e% &f Intelli*ent 0e,icle %y%te-% The advantages of vehicle based systems are faster deployment and wider use. These advantages are derived from the fact that it could take years to decide which highways and corridors to build an intelligent infrastructure on. /t could take less time for intelligent vehicle systems to become available. 'on-infrastructure dependent systems could then be used on any roads, not ;ust those that had been upgraded. )onsumers could get more use out of an intelligent vehicle system.
C&nclu%i&n The solution to this question lies somewhere between the two e8tremes. A wise distribution of intelligence promises ma8imum benefits at a lower cost. 0verlapping or redundant vehicle and infrastructure intelligence would provide a higher level of safety, reliability, and system integrity. /nfrastructure control is more efficient for lane changing, merging and platoon management. 9 AUTOMATED HIGHWAY SYSTEMS
However, vehicle systems have a better ability to sense other vehicles and obstacles, especially in a mi8ed traffic scenario.
"ane 1ee2in* The ne8t question to be addressed is how intelligent vehicles should keep their lanes. The three main types of systems that are being used today are video imaging systems, radar reflective striping, and magnetic markers. Iideo imaging is an e8citing technology because it can read current lane striping so it would require no road improvements. However, the onboard systems are very comple8 and costly. Fadar requires special reflective lane striping that can be seen by onboard sensors. The cost per mile for the new striping is very low and the onboard systems are relatively cheap. The ma;ority of systems being tested today use onboard magnetometers that read small magnetic
"e/el &f c&--unicati&n The third question to be answered in this topic is the level of vehicle to vehicle and vehicle to road communication. Some stakeholders are concerned that driver’s privacy could be compromised if a central computer can obtain his origin and destination of travel. /t is also undesirable, from the consumer’s point of view, for the infrastructure to have the ability to monitor vehicle speed and possibly even issue traffic tickets. 0bviously, the infrastructure must be given this information, but safeguards must be set up to guard user privacy.
S,&ul$ c&ntr&l e *i/en t& infra%tructure &r /e,icle%? &inally, the last question of a distributed intelligence is the division of vehicle control. Should the vehicles or the infrastructure have ultimate control of the vehicleO This becomes an especially difficult question when the two systems are giving conflicting instructions. /f the infrastructure issued an emergency brake command, but the vehicle’s forward-looking sensor did not detect a problem, what should the vehicle doO 4ither braking or maintaining speed could cause an 10 AUTOMATED HIGHWAY SYSTEMS
accident. /n another situation, the infrastructure could tell a vehicle to change lanes, but that vehicle>s radar might detect another vehicle in that lane. There is no easy solution to this question. Fedundancy is necessary to add safety and reliability, but problems can result when the vehicle and infrastructure issue conflicting instructions. The system then would need to both alert the driver and give him control, or choose the option that will least likely cause an accident.
C. "IA3I"ITY @iability has been an issue in the AHS program from the beginning. /t poses the largest threat to AHS technology because acceptance by automakers, consumers, and the government could be lost if an adequate solution is not found. ecause of the danger of platooning and even normal automated control, system malfunctions can cause serious accidents and even fatalities. %hen those accidents do happen, legal suits could be brought against automakers, drivers, and states if proper safeguards are not installed.
Aut&-a1er% Automakers will be unwilling to build AHS vehicles if they could be sued for failure of their systems. 7nder current law, drivers are responsible for maintaining safe control of their vehicles. /f the brakes or steering or any other system were to fail, the driver would be responsible for any accidents that he might cause. This is because drivers are not only responsible to drive safely, but also to maintain their vehicles properly and replace parts that could wear out. Automakers must conform to quality assurance tests which are most commonly !ean Time between &ailure 9!T&: tests. AHS systems would need to have similar quality control procedures. AHS systems must also be
G&/ern-ent Another concern is that the parties responsible for construction and maintenance of AHS highways could be held liable for a system failure that results in an accident. Puality controls and maintenance records would be necessary to keep the construction companies and states free from liability.
Dri/er% )urrent AHS systems are said to be
driver’s attention will inevitably shift from the road around him to his morning newspaper or conversations with other passengers. Although AHS systems may warn drivers that they must be in control of their vehicles, a simple warning may not be enough to keep drivers’ attention on the road and their vehicles. /n order to protect drivers from legal liability, there should be required maintenance checkups with a qualified service technician that would insure that the vehicle’s AHS systems are functioning properly and are in good condition. /t may also be necessary to certify drivers to drive AHS vehicles.
T&rt ref&r0verall, attempts should be made to limit tort liability. There needs to be a limit on damages that can be collected from any of the parties involved. Although tort reform has been relatively unsuccessful in many areas, there is good reason to believe that it could work for AHS. The airline industry has established a precedent of liability management. There are many similarities between AHS and the airline industry. oth systems involve the necessity of quality control in manufacturing control and maintenance checkups. oth also involve a state or federally funded infrastructure. /n addition, both involve the transport of passengers who rely on technology and infrastructure to get them to their destination safely. 0ne suggestion that has come from this comparison is that AHS vehicles should be equipped with a data recorder that would help determine the cause of any problems or accidents. AHS can learn a lot from the airline industry regarding liability and consumer acceptance.
C&nclu%i&n% The problems and decisions in AHS right now need careful consideration and more research before coming to hasty conclusions. The goals of safety and efficiency must be kept at the forefront of every system without overlooking the important issues of cost, equity and liability. 4ngineers and policy makers will need to decide together what the best solution is for the future of AHS.
A"TERNATI0E CONCE)TS #OR AUTOMATED HIGHWAY SYSTEMS SYSTEM CONCE)T
"OCA" )OSTION 4EE)ING
"ANE CHANGING
O3STRUCTION #"OW ON ROADWAY CONTRO"
Aut&n&-&u%5 &ully automated vehicles employing sensors and
Iehicles automatically senses vehicle
@ooks for and moves into an opening
Iehicle brakes for detected obstacles, changes lanes if possible 12
AUTOMATED HIGHWAY SYSTEMS
computer operate along with manually driven vehicles without requiring infrastructure assistance and communication C&&2erati/e5 Iehicles equipped with onboard sensors and computers would share information with other vehicles to coordinate maneuvers and enable fully automated travel Infra%tructure %u22&rte$5 &ully automated vehicles operate dedicated lanes, using global information and two way communication with smart infrastructure to support vehicle decision making Infra%tructure -ana*e$5 The automated roadside system provides intervehicle coordination during entry, e8it, merging and emergencies Infra%tructure c&ntr&lle$5 Same as above, but infrastructure takes the entire control in all driving situations
ahead roadway problems
and
Iehicles sensors, )ooperative communications negotiation among from other vehicles vehicle for land changes or platoons
Iehicle senses, communicates warning and coordinates maneuvers
Same as cooperative, but within guidelines from the infrastructure
Same cooperative
/nfrastructure or vehicle senses, communicates to vehicleM vehicle coordinates
Iehicle sensors, communications from other vehicles and
Iehicle lane
infrastructure needed
as
/nfrastructure sense vehicle positions and sends commands to control throttle, braking and steering
as
/nfrastructure senses sends commands to vehicles based on with infrastructure or vehicle for detection, or vehicle actions
request changeM
infrastructure responds commands surrounding vehicles
/nfrastructure determines need for lane change from origin-destination data, controls all necessary vehicles
/nfrastructure senses, sends commands to the vehicles based on /nfrastructure or vehicle detection, or vehicle actions
/nfrastructure monitors traffic, formulates responses, send parameters to local group of vehicles
/nfrastructure monitors individual vehicles, commands vehicles as needed including entry and e8it
/nfrastructure monitors individual vehicles, performs optimiEing strategy through control of individual vehicles
MO0EMENTS O# THE 0EHIC"E 13 AUTOMATED HIGHWAY SYSTEMS
"ateral M&ti&n Han$lin* The lateral 9side-to-side: motion of the vehicle has a number of different functions, from vehicle centric maneuvers such as lane keeping to those involving merging in heavy traffic. &irst, if the vehicle is to stay within the lane, it needs to know where the lane boundaries are. "ane $etecti&n is currently achieved through a number of different technologies, including a vision system, magnetic nails buried in the roadway which are then sensed by the vehicle, or a radar-reflective stripe 9unpublished work at 0hio State 7niversity:. %ith the advent of lane detection capability, the system can then detect where it is within the lane, leading to lane $e2arture +arnin*% when a vehicle strays out of the lane unintentionally. This is an attractive function to have available for incremental deployment as $"K of all highway fatalities are a result of single-vehicle, run-offroad accidents.9?: !arketable systems might be in the form of warnings which alert the driver when a lane change is attempted without prior activation of a turn signal, or might involve a driver based model that adapts to the characteristic driving patterns of the driver. 0nce a vehicle knows the lane boundaries and is able to determine its own position within the lane, lane 1ee2in* also becomes possible. @ane keeping is the ability of the vehicle to drive down the center of the lane, taking into account upcoming curves and the required pre-steering used to maneuver into them gracefully. A vision system uses a preview area of the lane image for this purpose. !agnetic markers and radar reflective strips could encode upcoming curve information in the roadway, or rely on navigational position information and an electronic map in order to alter the vehicles> tra;ectories appropriately. Si-2le lane c,an*in* is the ability of the vehicle to move smoothly out of one lane and into another in light traffic conditions. The technical requirements for such a system include side-looking sensors that detect a gap, and the ability to cross between ad;acent lanes and begin lane keeping in the new lane. Such a system could be considered
space, while the latter is concerned with the vehicle departing from its own lane une8pectedly. Staility au*-entati&n is a function wherein the vehicle gently resists changing lanes when a turn signal is not used. The steering wheel resists the motion of the driver, making it feel as if the vehicle is driving with its wheels in ruts, helping keep the vehicle in its lane. This type of function might also prove useful in high crosswind situations, where the vehicle assists the driver in stabiliEing a position within the center of the lane.
"&n*itu$inal M&ti&n Han$lin* The longitudinal 9front-to-back: motion of the vehicle also has a variety of functions which range from simplistic in-vehicle handling to tactical driving within a congested traffic scene. S2ee$ 1ee2in* is the most elementary function within this category, involving the maintenance of a constant travel speed. /t is widely deployed in the form of
intelligence. /t is also clear that all higher-order fully automated systems depend on this capability in order to function. &or this reason, vehicle motion prediction is identified as a critical link in the deployment of AHS and is shown in black in order to highlight its importance. C&r$ial $ri/in* , the ne8t step in maneuvering, is best described as driving which accommodates the desires of other vehicles in a friendly way. &or e8ample, if a vehicle is in a merge lane with its turn signal activated, manual drivers tend to 9but do not always: give way so that vehicle can merge. )ordial driving by an automated or semi-automated vehicle can be enabled once the vehicle can discern the intentions of surrounding vehicles. /n &igure ", the assumption is that communications are not universally and dependably availableD therefore, the vehicle must infer intention by vehicle location, vehicle motion, and vehicle signaling. 9Signaling could be augmented with, for e8ample, a radio transmission but such signaling is not assumed to be universally available.: This type of inference is, therefore, a large part of vehicle motion prediction. A/&i$ance /ia lane c,an*e in traffic is a capability that enables the vehicle to change lanes given a haEardous situation, such as the sudden and une8pected slowdown of a preceding vehicle. This capability is a reasonable and attractive alternative to hard braking, as it may avoid secondary and tertiary accidents that could ensue when braking is used. /n order to perform this function safely in traffic, the vehicle system must be acutely aware of the operating environment, and predict the behavior of surrounding vehicles with enough accuracy that the lane change can be done safely and efficiently. 'ote that lane changing in traffic, a lateral motion handling capability, must be supported in order for this function to be available, full tactical $ri/in* is introduced. Tactical driving is the ultimate form of full automation within an individual vehicle. At this stage in the deployment process, the vehicle not only tracks and reacts to other vehicles, but also proactively plans out series of maneuvers which are e8ecuted to achieve a goal or goals. &or e8ample, if the vehicle 9or the driver: desires to increase speed and the vehicle is
O%tacle Han$lin* 0bstacle avoidance capabilities reduce or eliminate safety haEards caused by obstacles on the automated highway roadway. This includes rocks, vegetation, dropped vehicle parts, disabled vehicles, and animals such as deer. 0ne way to reduce the need for obstacle avoidance is to implement obstacle e8clusion. To a limited degree this is already deployed with fencing and highway department maintenance of the interstate highway system. 0bstacle e8clusion can significantly reduce the frequency of obstacles on the roadway, but it seems implausible that any e8clusion method can be "++K effective. Thus, more sophisticated forms of obstacle handling will be needed in most foreseeable end-state AHS implementations. 'ote that obstacle e8clusion, as an infrastructure-based system, it is depicted on a later diagram that introduces infrastructure functionality. O%tacle $etecti&n an$ t,reat $eter-inati&n is a much more difficult task than vehicle detection due to the technical difficulties of sensing obstacles and identifying whether those obstacles present a threat. A mylar toy balloon, for e8ample, is easy to sense but would preferably not trigger a severe braking maneuver that could risk minor in;ury to vehicle occupants. 0n the other hand, deer present a much greater threat but are more difficult to sense and track. /t seems likely that obstacle detection and threat determination will be available well after vehicle detection and collision avoidance functions are fielded. 'otice that a dotted arrow 16 AUTOMATED HIGHWAY SYSTEMS
links the two functions as there may be value to the research done for the detection of vehicles in the development of obstacle detection capability. iven that obstacles can be detected and properly categoriEed, %i-2le &%tacle +arnin* may be possible, whereby the driver is warned of haEards in and around the roadway. This function can provide obstacle information for the driver to use, or it could be coupled with the a/&i$ance /ia ra1in* capability. /n that avoidance via braking was fielded previously as a longitudinal motion handling function, it is likely that it could be marketed as an obstacle avoidance function as soon as reliable warning systems are available O%tacle -&ti&n $etecti&n 6 2re$icti&n may be a particularly difficult capability to develop. 7nlike vehicles, which are physically constrained in realiEable maneuvers, obstacles may not behave in readily predictable ways. 5eer may run into the road and stop abruptly. @oose tires can bounce randomly, depending on road surface, tire wear, and angle of incidence. !ylar balloons, plastic bags, and tarps are at the whim of the wind an d thermal pressures created by hot roadways. 'one of these are easy to quantify, and as such, create very difficult detection and prediction problems. 'otice that vehicle motion detection and vehicle motion prediction might provide benefit to the development of this capability.
DI##ERENT SENSORS AND THERE USE ". @ong-range sensors that measure the distance to an obstacle or vehicle in the environmentM These include both active sensors such as laser and radar, as well as passive ones such as stereo vision. C. Short-range sensors that detect ob;ects in the immediate vicinity of the ego-vehicleM Again, these include both active sensors such as sonar, and passive ones such as capacifle8or. These sensors are typically used to check blind spots or parking spaces before attempting lateral maneuvers. $. Sensors that measure the relative velocity of a sensed vehicleM Some range sensors 9such as 5oppler radar: do this automatically. &or other sensors, this may involve sophisticated processing 9such as optical flow:, or require differential range measurements. 1. ositioning sensors that enable estimation of current ego-vehicle locationM These include on-board systems such as encoders, in addition to the lobal ositioning System 9S:. 3. Sensors that measure the ego-vehicle>s current velocityM This may be done through odometry, by sensing relative motion with the ground using various techniques, or by using S. *. @ateral lane-tracking systemsM This includes direct measurement of lane deviation through infrastructure-assisted methods such as magnetic nails as well as sophisticated software for vision-based lane-trackers. ?. )locksM Since modern clocks do not lose an appreciable amount of time over the duration of an traffic scenario, modeling a time sensor for AHS simulations is trivial. This is certainly not true for all applications. 6. 4nvironment sensorsM This category includes sensors that detect dangerous road conditions such as ice or wet pavement, or weather conditions such as fog and snow 9which can adversely affect other sensors:. Since these phenomena are typically unmodeled in micro simulations, environment sensor models do little more than reporting the initial conditions as specified by the user.
17 AUTOMATED HIGHWAY SYSTEMS
SYSTEM DESCRI)TION 4ach platoon is composed of a leader that is the first car of the platoon and a set of followers. A platoon that contains one vehicle is called free agent. &igure " shows three platoonsM p" with three vehicles, a leader and two followers, pC is a neighboring platoon, and p$is an e8ample of free agent. The intra-platoon distance 9Q8: ranges usually between one to three meters. The inter platoon distance between two platoons 9Qp: in the same lane varies between thirty and si8ty meters.
#i*ure 75 C&nte(t &f a 2lat&&nin* a22licati&n
18 AUTOMATED HIGHWAY SYSTEMS
The ATH research program has defined hierarchical control architectures for platooning applications. The platoons use lateral and longitudinal positioning controllers 9magnetic equipments: to allow the vehicles to follow each other safely. The vehicles are coordinated by means of communications, based among other things on information from the magnetic equipments. Several maneuvers have been defined to allow the system to be in safe conditions in the absence and in the presence of failures 9fail-safe mode:. The main maneuvers consist in splitting a platoon, merging platoons, or making a vehicle e8it or enter the platoon. /n case of a failure affecting a vehicle in the platoon, the maneuvers allow the vehicle to leave its platoon without any haEard, for the purpose of continuously running the platoon without any problem. efore starting a maneuver, the faulty vehicle communicates with its platoon’s leader 9that initialiEes the coordination of the maneuvers:. According to the failure mode, some maneuvers may require a communication between the leaders of neighboring platoons in addition to communications with ad;acent vehicles. /f the faulty vehicle is the leader, specific maneuvers must be applied to allow the platoon vehicles to select a new leader. %e briefly present background information on the ATH architecture that is needed to understand our safety models. %e mainly focus on the failure modes considered and the recovery maneuvers used to ensure AHS safety, taking into account different strategies for intra-platoon and inter-platoon coordination.
0EHIC"ES COORDINATION latooning applications require coordination between the vehicles in the platoon 9intra-platoon: and with neighboring platoons 9inter-platoon:. A vehicle is involved in the coordination process when i: it creates a platoon, ii: enters an e8isting platoon, or iii: when it leaves a platoon to switch to manual driving. Iarious communication models 9centraliEed and decentraliEed: have been proposed in for the inter- and intraplatoon coordination, based on the ATH architecture. They are briefly summariEed hereafter .
Inter82lat&&n c&&r$inati&n 5 )ommunications between platoons can be achieved only through the leaders, and the coordination can be centraliEed or decentraliEed. /n the centraliEed coordination model the coordination between the leaders of neighboring platoons is performed through a centraliEed Service Access oint 9SA: that is on the road-side. The coordination between different maneuvers is achieved at the level of the SA. &igure $ presents an e8ample considering an AHS composed of two lanes with two platoons, p" followed by pC on laneC, and two free agents, v? and v6, on lane". @et us assume that i: v? and v6, which ;ust entered the highway, decide to ;oin respectively platoons pC and p", and ii: simultaneously and independently, vehicles vC and v3, belonging respectively to platoons p" and pC, are coordinating maneuvers to e8it the AHS after passing through lane". The SA determines the priorities between the maneuvers involving the four concerned vehicles and communicates its decision to the leaders of the platoons including the concerned vehicles. The decision would be to assign the highest priority to the maneuvers requested by v? and v6, because it is important to release lane" as quickly as possible, so that vC and v3 can leave the highway.
19 AUTOMATED HIGHWAY SYSTEMS
#i*ure 95 Centrali:e$ inter82lat&&n c&&r$inati&n
/n the case of decentraliEed inter-platoon coordination, the decision is made by the leaders of the concerned platoons. The information related to the state of all vehicles is stored in an onboard system that contains a knowledge base of the neighborhood traffic. This coordination strategy has an impact on the implementation of some atomic maneuvers. )ompared to the centraliEed strategy, it involves fewer vehicles in the accomplishment of some maneuver. @et us consider as an e8ample the case of a faulty vehicle that needs to perform a Take /mmediate 48it-4scorted 9T/4-4: maneuver with the support of a neighboring platoon. /f the inter-platoon coordination is centraliEed, the implementation of this maneuver involvesM ": all the vehicles in front of the faulty vehicle 9including the leader: and the vehicle ;ust behind it, and C: the leader of the neighboring platoon. However, in the decentraliEed inter-platoon coordination strategy, only the leaders of the two platoons and the vehicles ;ust in front and behind the faulty vehicle contribute to the maneuver.
Intra82lat&&n c&&r$inati&n5 /n the centraliEed intra-platoon coordination model the coordination of operation and maneuvers involving the vehicles of a platoon is centered on one vehicleM the leader. &or e8ample, during a split maneuver that is initiated to allow the safe e8it of a faulty vehicle, three vehicles are involvedM the leader, the splitter, and the vehicle following the splitter 9if it e8ists:. The faulty vehicle should announce the need to initiate this maneuver to its platoon’s leader. The leader then calculates the distance and the speed to be respected by the vehicles that are involved in the maneuver, and orders the involved vehicles to change them accordingly. /n the case where the intra-platoon coordination is decentraliEed, each platoon member has knowledge of the platoon formation and can react independently, by communicating directly with other vehicles. The leader is informed of changes as it is the representative of the platoon for inter-platoon coordination.
ACTIONS TA4EN IN CASE O# #AI"URES IN THE SYSTEM
20 AUTOMATED HIGHWAY SYSTEMS
Several failure modes, with various severity levels, can affect the vehicles involved in platoons and their safety. 5epending on the failure severity, various maneuvers can be considered to ensure the safety. Some maneuvers may need to stop the faulty vehicle or help it to e8it safely from the highway as soon as possible with the assistance of ad;acent vehiclesC. /n the case where the failures have a minor effect on safety, the faulty vehicle could e8it from the highway without the assistance of other vehicles. /n the following, we first present the failure modes that might affect a single vehicle, their severity and the associated maneuvers. Then, we discuss the case of failures affecting multiple vehicles. &inally, we present the catastrophic situations that could lead the automated highway system to an unsafe state.
SING"E 0EHIC"E #AI"URES5 Si8 potential failure modes have been identified, presented in Table ". This table shows for each failure mode, an e8ample of cause leading to the failure mode, the severity class, and the maneuver that ensures the safe continuity of service despite the presence of failures.
Tale 75 #ailure -&$e% an$ a%%&ciate$ -aneu/er%
The severity classes associated with the failure modes are ranked by decreasing order. )lass A is the highest, gathering the most critical failures that need to stop the vehicle on the highway. Three maneuvers are defined for this purposeM entle Stop 9S, where the fault vehicle uses its brakes smoothly to stop:, )rash Stop 9)S, where the faulty vehicles uses ma8imum emergency braking:, and Aided Stop 9AS, where the faulty vehicle is stopped by the vehicle immediately ahead:. Specific control laws are then used to ease congestion, divert traffic away from the incident, assist emergency vehicles, and get the queued vehicles out. The severity classes 9 and ): include the failure modes that can be recovered by allowing the faulty vehicle to get out of the highway without stopping the traffic. The corresponding maneuvers can be achieved either without assistance or with the cooperation of some ad;acent vehicles. Three maneuvers are defined too, namelyM Take /mmediate 48it-4scorted 9T/4-4:, Take /mmediate 48it 9T/4:, Take /mmediate 48it- 'ormal 9T/4-':. /t is noteworthy that the severity class also determines the priority of the corresponding maneuver. This is important when multiple failure modes occur. The priorities within each class are as followsM %ithin )lass A, A$ has the highest priority and AC has higher priority than A". /n )lass , " and C have equal priority. /n case of occurrence of multiple failure modes in the same vehicle, the maneuver with the highest priority is applied. 21 AUTOMATED HIGHWAY SYSTEMS
5etails about the atomic maneuvers composing each of the si8 maneuvers presented in Table " and the inter-vehicle coordination required to implement them, are presented in the successive failure of maneuvers may eventually lead to a state where no maneuvers are available to recover the faulty situation. This is illustrated by the state machine in &igure C, where vRG0 identifies such a state. The transitions correspond to the occurrence of failure modes, or to the results of maneuver e8ecutions that might succeed 9transitions to the safe state, vR0G: or fail 9G0 transitions:. %hether the state vRG0 corresponds to an unsafe state for the AHS or not, depends on the state of the ad;acent vehicles.
CONC"USION ". y using this system we can increase the capacity of the road from C+++ vehicleBlaneBmile to 6+++ vehicleBlaneBmile. C. 5riving safety will significantly be increased. Since human errors will be removed. Some estimates states that about 3+K of an improvement can be achieved by using AHS. $. %eather and environmental conditions have little impact in high performance driving. 1. &uel consumption can be reduced. 3. @and can be used more effectively. *. Travel time can be saved. Traffic will not be a ma;or issue in deciding the travel time.
22 AUTOMATED HIGHWAY SYSTEMS
RE#ERENCES 1. @. AlvareE. Automated Highway Systems: safe platooning and traffic flow control . h5 thesis, 7niversity of )alifornia at erkeley, "##* 2. @. AlvareE and F. HorowitE. Safe platooning in automated highway systems. part iM Safety regions design. Vehicle System Dynamics Special Issue: IVHS , $C9":MC$3*, uly "###. 3. . @. AlvareE and F. HorowitE. Safe platooning in automated highway systems. part iiM Ielocity tracking controller. Vehicle System Dynamics Special Issue: IVHS , $C9":M3?61, uly "###. . @. AlvareE, F. HorowitE, and . @i. Traffic flow control in automated highways systems. I!A" #. $ournal on %ngineering &ractice, ?M"+?""+?6, "###. '. !. arth and .!. 'orbeck. Transportation modeling for the environmentM &inal report. Technical report, ATH Fesearch Feport 7)-/TS-FF-#*-*, /nstitute of Transporation Studies, 7niversity of )alifornia, erkeley, )A #1?C+, "##*.
23 AUTOMATED HIGHWAY SYSTEMS
