Hierarchical Vehicle Management Concept for Automated. Guideway Transportation Systems. REGGIE J. CAUDILL, ALAIN L. KORNHAUSER, AND JOHN R.
TRANSACTIONS IEEE
ON VEHICULAR TECHNOLOGY,
VOL. VT-28, NO.
1, FEBRUARY 1979
11
Hierarchical Vehicle Management Concept for Automated Guideway Transportation Systems REGGIE J. CAUDILL, ALAIN L. KORNHAUSER, AND JOHN R.WROBLE
Abstmcr-A detailed description of ahierarchicalsystemmanagement concept that can operate a large automated guideway transportation system while maintaining achievable communication, vehicle control,andnetworkmanagementrequirements is presented. The require minimal conceptemploys“smart” vehiclecontrollerswhich instructions from the network controller andcan operateautonomously in case of a failure in either the network management system or the communication system. Wayside-to-vehicle communication requirements for the smart vehicle are determined and compared to those of a passive vehiclesystem. Theeconomicconsequencesimposed by the smart vehicle’s additionalonboa(dequipmentontheowners of the privatelyowned automated highway (AHS) vehiclesarediscussed. System-owned communication and control packages are recommended to make the economics of the concept favorable, even for the occasional A H S user.
I. INTRODUCTION
T”
E RESPONSIBILITY for the safe management of vehicles of an automated guideway transportation system (either automated highway or automated guideway transit) can be distributed over a hierarchy of responsibility. A first stratification of this hierarchy can bea macrolevel anda microlevel. The microlevel is concerned with the operation of individual vehicles-control of the longitudinal and lateraldynamics of each vehicle (e.g., headway control, acceleration and deceleration maneuvers,emergency collision avoidance,lanechange, lateral tolerance, and coupling). Themacrolevel, termed network control, has the responsibility of controlling the vehicle fleet, including vehicle dispatching from station, empty-vehicle assignment, vehicle routing, vehicle diverging and merging at intersections, and failure management. Both of these levels are in mostpart autonomous; however,coupling betweenthem exists throughcommunicationrequirements.Thenetwork controller requires some rudimentary knowledge of the state of each vehicle, and the vehicle controller bases its maneuvers on instructions from the network controller. Thispaper discusses thecommunication requirements of various vehicle controlconcepts andpresentsadetailed deManuscript received May l , 1978;revised June 30,1978. This work of the was supported in part by the Federal Highway Administration U S . Department of Transportation. The opinions expressed are those of the authors and not necessarily those of the Department of Transportation. R. J. Caudilland A . L. KornhauserarewiththeTransportation E-420, EngineeringQuadProgram,PrincetonUniversity,Building rangle, Princeton, NJ 08540. Telephone R. J. Caudill (609) 4524596, A. L.Kornhauser (609) 4524657. J. R. Wroble was withtheTransportation Program,Princeton NJ. He is nowwithAlan M. Voorheesand University,Princeton, Associates,Inc., 7798 Old SpringhouseRd., McLean,VA.Telephone (703) 893-4310, ext 224.
scription of ahierarchicalsystemmanagement conceptthat can operate a large automated guideway transportation system having achievable communication, vehicle control, and network management requirements. The concept employs “smart” vehicle controllers, whichrequireminimal instructionsfrom the network controller and can operate autonomously in case of a failure in either the network management system or the communication system. The “smart” vehicle concept is based on the premise of placing substantial amounts ofintelligence onboard the vehicle. With this allocation of intelligence, thecommunicationbetween the wayside and the vehicle and the wayside computer storageand computation requirements are minimized.The significance of these advantages is evident by comparing the communication requirements for the smart vehicle system to those for a “passive” (or “dumb”) vehicle system. Many of theoperational and structural characteristics of automated guideway transit(ACT)and automated highway systems (AHS) are similar or even the same;however,there are some distinguishing characteristics that impose different constraints on the design of the vehcle managementsystem
PI
*
1) AHS mustbe capable of managing a large number of vehicles, perhaps two or three orders of magnitude larger than some ACT systems. Major impact: vehicle autonomyfrom wayside. 2) AHS is basically a corridor-type system; consequently, thenetwork topology is h e a r , as opposed to grid or loop. Major impacts: failure management, routing, and intersection control. 3) The vehicles which operate on AHS are privately owned, as opposed tothe publiclyowned ACT vehicles. Major impacts: on-board equipment costs, empty-vehicle shuttling, pretrip vehicle inspection. 4) AHS is primarily an intercity mode, thus the physical dimensions of AHS are large, compared to ACT. Major impact: roadway modification cost. Although the management hierarchy presented in this paper was devised for an AHS system, the controller structures, functional responsibilities, and basic conclusions are applicable to manyACTconcepts, especially the small vehicle systems which operate at short headwaysand moderate speeds-personal rapid transit (FRT) systems. The remainder of this paper is divided into four sections. The first is a discussion ofthe vehicle management control hierarchy-individual vehicle and network control. An examination of wayside/vehicle communication requirements for the smart and passive vehicle systems is presented in the next sec-
0018-9545/79/0200-0011$00.75 0 1979 IEEE
12
TRANSACTIONS IEEE
ON VEHICULAR TECHNOLOGY, VOL.
tion. The paper concludes with a preliminary evaluation of the economics of system-owned communication and control packages for AHS followed by a brief summary highlighting the basic conclusions of this research.
11. HIERARCHICAL VEHICLE MANAGEMENT Vehicle Longitudinal Control Longitudinal controllers on boardeach vehicle are necessary to regulate the vehicle’s position and velocity, according to specified command profiles. In normal situations, thelongitudinal control system maintains the proper position and velocity during a) mainline operation at constant speed, b) transitions between regions of high- and low-speed operation, c) merges and diverges at stations and intersections, d) maneuvers at intersections to avoid potential conflicts, e) starts and stops within exit area queues. Normal control must be accomplished without subjecting passengers to acceleration or jerklevels, which may cause discomfort. In emergency situations, the longitudinal controller must stop the vehicle, in order to avoid collisions. To minimize the delay time involved in sensing an emergency and initiating braking, an on-board collision-avoidance radar monitorsthe relative motion of the preceding vehicle. If separation or closure rate levels exceed threshold values, emergency braking is initiated and the wayside is alerted. Extensive literature exists describing past research in the design and analysis of longitudinal control systems for various types of automated vehicles [2] -[ 191 . The experience t o date indicates that the vehicular longitudinal dynamics can be adequately controlledelectronically,and that headways shorter than those on conventional highways can be achieved without compromising safety. Traditionally, fured-block control hasbeen used to maintain desired headways in rail transit systems. Minimum headways achievable under conventional block control are on the order of 1 minute. To achieve the flow capacity (vehicles per hour) necessary for AHS and many AGT systems requires headways on the order of one second; consequently, conventional fEedblock control is unacceptable. Short-headway operation can be realized under twopossible longitudinal control philosophies [20] . A point-follower system [3] -[9] regulates each vehicle’s position relative to a moving reference point, while a vehicle-follower system [ 101 [16] regulates the state (position, velocity,acceleration,and jerk) of the vehicle relative to the state ofneighboring vehicles. Significant trade-offs betweenthetwo philosophiesexist in controller design, sensing requirements, and wayside-to-vehicle communication requirements,as described in the following. 1) Controller Design: On-board controller design for pointfollower appears to be somewhat simpler than for vehicle-follower systems in that, with the possible exception of emergency braking, the response of only a single vehicle rather than a string of vehicles must be considered. Since various headway policies [21] (constant separation,constant time headway, and constant safety factor) are implemented by varying slot
NO. VT-28,
1 , FEBRUARY 1979
size, it appears possible to design the on-board controller independently of spacing policy. However, vehicle-follower control systems can be successfully designed, which implement either a constantsafetyfactororconstanttime headway spacing policy and assure stable string behavior [ 131 -[ 161 , [22] . 2) Sensing Requirements: The type of data necessary for point-follower and vehicle-follower systems is basically the same; however, the required accuracy of the various data are different. Therequiredaccuracy of any piece of data is dependent upon how this data is used by the system. The most prominent difference in data usage between point- andvehiclefollower systems is with respect to position data and preceding vehicle separationand closure rate data.In point-follower systems, the position data is used t o regulate the vehicle’s state; consequently, it must be highly accurate. However, the separation and closure rate data are used only to determine if emergency threshold values have been exceeded. Thus the accuracy of this data is less critical. For vehicle-follower systems, just the opposite is true. The preceding vehicle separation and closure rate must be highly accurate, as they are used to reguate the vehicle’s state. The position data is used only to determine where a command manuever is to be initiated. Thus, precise continuous knowledge of the vehicle’s position is unnecessary. The accuracy of data is dependent upon theaccuracy of the measuring equipment (sensors) and the sampling rate at which the data is measured and transmitted. A number of position reference systems, using a variety of technologies, have been proposed by Fenton [18] and others. Positionmeasurement can be accomplished by sensing the passing of equally spaced position indicatorsembedded in theroadway.Interpolation between these incremental position markers would be used to providea piecewise continuous estimate of vehicle position. The accuracy of this estimate is dependent upon the spacing between these position indicators (i.e., the rate at which the estimate is updated). The incremental position measurements would be updated with mileposits every 1000 to 10 000 increments. Thedistance between mileposts is a function of the reliability of the vehicle to sense the incremental markers and the durability of the passive increment display system that is to be embeddedin the highway. It is anticipated thatunder point-follower control,the vehicle would monitor the passing of incremental position indicatorsonthe order of once every 0.1 seconds. Thusat typical AGT/AHS speeds, they would need to be located every 10 feet along the roadway.Under vehicle-follower control, spacing of the position markers may be one or two orders of magnitude greater than for point-follower control. 3) Wayside-to-Vehicle Communication Requirements: Designing the communication system requires that the minimum rate at which data is transmitted be determined. The bit-rate BR can be expressed as BR = BN/T bit/sec, where
B N
longest message to be sent to each vehicle (bit), numberof vehicles to which the wayside transmits
13
CAUDILL et al. : VEHICLE MANAGEMENT FOR AGT
A graphic comparison of the communication requirements betweenpoint-follower and vehicle-follower control is illustrated in Fig. 1 in terms of the minimum bit-rate as a function of lane density p . The lane density is the ratio of the number of vehicles in the region to the maximum number the region can accommodate.Forpoint-follower,theminimunbit-rate increases linearly with lane density. For vehicle-follower, the minimum bit-rate increases linearly until p = 0.5 then decreases. Therefore, lane saturation ( p = 1.0) is the most demanding condition on point-follower communications, while p = 0.5 is 0 .25 .50 .7J 1.0 mostdemandingfor the vehicle-follower system whenthe h. Demit,. 0 capture region is 2 H . In general for vehicle-follower systems, Fig. 1. Minimum communication requirements as a function of lane if the capture region is kH, the minimum communication redensity for vehicle and point-follower control. quirement is BN,/kT bit/sec. Philosophical Viewpoint: On aphilosophical basis, veduring the communication cycle time, hicle-follower longitudinal control is desirable because it proT communication cycle time (seconds). vides a certain autonomy to each of the vehicles and decouples their individual actionsfrom a continuous dependence on Assuming thatthelength of command messages and the eithera local or central controller, while oversight managecommunication cycle time are the same for point- and vehiclement from a centralized source would maintain order and disfollower systems,thenthe differencein communication recipline over the vehicle fleet. This concept is comparable to quirements will depend only upon a difference in the number present automobile behavior as it couldbeenhanced by a of vehicles to which the wayside must communicate during better and continuously updated sign and information system. time T. Communication requirements are based uponcondiThe continuous and precise intervehicular regulation and contions during peak communication demand. trol is to be kept in the domain of neighboring vehicles. Toexecute a commanded maneuver under point-follower Nominal as well as off-nomialbehavior of the preceding vecontrol requires communication with each vehicle that is exhicle can be sensed without the need of having that informapected to participate in the maneuver, since the vehicles are tion transferred through an intermediate central vehicle dynamically decoupled. In a vehicle-follower system, only the managementsystem. The fundamental technologicalrequirelead vehicle of eachstring receives the maneuver command, ment of thevehicle-follower system is that of an intervehicular but the entire string executes the maneuver, since the vehicles communication system that can transmit information on the are coupled through their longitudinalcontrollers. instantaneous position and velocity of each vehcle to an upConsider the following scenario:asection of guideway stream neighboring vehicle. This requirement is compounded under the jurisdiction of a single wayside controller is capable by the need for this information transfer to be accomplished of accommodating N u vehicles; and conditions warrant that all during the merge process where information must be transfervehicles decrease their speed. If the section is saturated with Nu vehicles then,forpoint-followeroperation, Nu vehicles red from vehicles on different physical sections of the guideways. If t h s technological problem is solved, then from a conmust receive proper commandswithin thecommunication ceptual standpoint, thevehicle-follower control system is to be cycle time T. Consequently, if the command is B bits in length, a bit-rate of BNJT bit/sec is required. For a vehicle-follower preferred because of the following reasons. system under these same conditions, communication with only It will reduce the number of electronic elements needed one vehicle (the lead vehicle) wouldberequired. The other in the guideway and therefore greatly reduce the guidevehicles wouldfollow the lead vehicle. Theresulting bit-rate way costs. (This is highly desirable because of the large would be BIT bit/sec. physical size of AHS). This saturatedcondition is themost demanding on the It is more adaptable to local situations. communication system for point-follower but obviously not It is less vulnerable to fduresin that communication befor the vehicle-follower system. The maximum communication tween vehicles is direct and therefore not subject to the requirements for the vehicle-follower system is when the reliability problems oftheintermediate transmission number of vehicles operating in the velocity-command mode is systems. maximum. This condition corresponds to having all vehicles in It has significantly lower wayside-to-vehicle communicathissectionseparated bythe length of thevehcle-capture tion requirements. This is a major benefit because of the region, Assuming thatthe mode transitionfrom velocity large number of vehicles that AHS must accommodate. command to spacing regulation is made when the following vehicle is twice the nominal minimum nose-to-nose separation Network Control distance H , then the vehicle capture region would be W .This A major design issue for vehicle management is the selection section of guideway can accommodate only YJV, vehicles each of an operational management policy. This issue centers about separatedby 2H. Consequently, the bit-rate required for the the degree to which the trajectory (position as a function of vehicle-follower system is BNu/2T, which is one-half that re- time) of each vehcle should be specified. In order that conquired for the point-follower system. ficts at merge junctions be resolved, a vehicle's trajectory, at ~
~~
14
IEEE TRANSACTIONS VEHICULAR ON TECHNOLOGY,
the very least, must be specified prior to traversing those sections of the networklocated a “short” distance upstream from the merge junctions. The upper extremeis that the trajectories be specified for the entire trip. Thus, two1 classes of operational management vehicles have been identified: 1) quasisynchronous and 2) synchronous. For quasi-synchronous management, vehicle trajectories are specified only in the vicinity of merge junctions. In this case, the specification of the trajectory is defined in real time as a function of the existing traffic flow conditions in the vicinity ofthat junction [23] . The second type of network managementhasbeen termed synchronous, since the entire trajectory, fromorigin to destination, has been specified prior to the vehicle’s departure [28], [29] . While synchronous has the inherent ability to estimate accurately arrival time at the destination and to minimize the aggregate delay of all vehicles processed by the system, it is necessary that all vehicles on the system precisely adhere to the programmed trajectory. Any variation in any trajectory can require that the trajectory of all other vehicles be recomputed. This extreme sensitivity to even the slightest malfunctions makes synchronous control very undesirable, especially when the controllers must manage a relatively large number of vehicles. Other considerations make the quasi-synchronous philosophy desirable. These include a) its compatibility with vehicle-follower longitudinal control option as well as the point-follower option, b) thedistribution of trajectorycomputationstomany local controllers that require only local information for resolving merge conflicts, and c) its flexibility under variable operating conditions. In considering the trade-off between the two vehicle management philosophies, the quasi-synchronous philosophy is highly preferred over the synchronouspolicy.
Hierarchy of Management Elements and their Functional Responsibilities The control functions for vehicle managementencompass vehicle routing, stationdispatching, intersection control,failure management,andindividual vehicle control. These control functions are hierarchicalin nature. An infinite numberof hierarchical schemes can be devised to perform the management functions. In this paper management and control of the flow of vehicles is to be distributed over a hierarchy of five controllers ranging from the vehicle level which maintains the proper positionandvelocity over time,tothe global level which maintainssummaryperformancestatistics. Thethree intermediate level controllers-central, regional, andsectorperform system management functions that manage the flow 1 The literature often discusses three classes of vehicle management, the two given above plus “asynchronous”, in whichthe vehicle’s trajectory is nowhere synchronized. This class of vehicle management is clearly not applicable to AHS orany situation that has any merge junctions, since merge conflict will need to be resolved through some sort of synchronization of vehicles in some upstream neighborhood of each merge junction. Some authors use the term “asynchronous” yet provide some synchronization at merge junctions. We consider this to be quasi-synchronous.
VOL. VT-28, NO. 1 , FEBRUARY 1979
(Vahicleroutins)
Fig. 2.
Communicationflow chartandprimary functional responsibilities of management element hierarchy.
of vehicles, resolve merge conflicts,and initiate emergency maneuvers. Fig. 2 depicts the organizational hierarchy of the classes of controllers. The GLOBAL controller monitors the CENTRAL controller,which in turn monitors andmanages the REGIONAL controllers. The REGIONAL controller monitors and manages vehicles traversing the main roadway of theregion and coordinates the operation of the sector controllers (if any are necessary). The sector controllers manage vehicles as they enter the system (DIAGNOSTIC), leave the departure queue and merge onto the main line (STATION/DISPATCH), proceed throughintersections (INTERSECTION), and exitthe system (ARRIVAL). Systemperformanceobjectives, such as minimum headway,maximum speed, and vehicle reaction time, were considered as constraintsonthe controller design. When tradeoffs existed, minimization ofintercontrollercommunication requirements was favored over intracontrollercomputation and storage. The minimization of communication was applied most strenuously between the vehicle and the wayside, where “hard wire” or microwave communication is not feasible. This objective requires that thevehicles be “smart”, i.e., they receive rudimentary commands but generate longitudinal and lateral control command profiles on-board. In addition, the “smart” vehicle monitors its vital systems and senses its actual state.By monitoring its vital systems, the vehicle can alert the wayside of impending malfunctions and enhance the capability of anticipating emergency situations. It is anticipated that each sector controller would monitor the performance of between 100 and 300 vehicles simultaneously. The REGIONAL controller would manage a maximum of 600 vehicles (corresponding to 15 km of roadway; I/ = 25 m/s, h = 1 s) and coordinate the operation of the sector controllers in the region (if any are necessary). In turn, each
CAUDILL MANAGEMENT et al. : VEHICLE
FOR AGT
CENTRAL controller would monitor between 10 and 100 regions. If there would exist an average of one REGIONAL controller per15 km (-10 miles) of roadway, theneach CENTRAL controller could be responsible for approximately 1000 miles. Once the system grew beyond this size, an additional level, the GLOBAL controller, would be required to monitor the CENTRAL controllers. The GLOBAL controller should have the capacity of monitoring up to100 CENTRAL controllers. Functional Responsibility of the Controllers: Thefunctional responsibility of theGLOBAL, CENTRAL, REGIONAL, and thefour generic SECTOR controllers-DIAGNOSTIC, STATION/DISPATCH, INTERSECTION, and ARRIVAL, are discussed below. I ) GLOBAL: The primary responsibility of theGLOBAL controller is to monitor the performance of the CENTRAL controllers andto relay system-wide information between CENTRAL controllers. Hard-wire or microwave (possibly via satellite) communication would exist between all CENTRAL controllers and the GLOBAL controller. 2) CENTRAL: The primary responsibility of the CENTRAL controller is the routing functions for its jurisdiction. Itmonitors region status, maintainsforecasts of linkand station arrival rates, and computes routing commands for all REGIONALcontrollers. These commands are subsequently transmittedtothe INTERSECTIONsector controller.It is responsible for generatingreal-timererouting commands in the case of lane blockage and for coordinating restart maneuvers for regions. It supplies the GLOBAL controllerwith summary data and accepts from the GLOBAL controller the status of gateways that interface between CENTRAL control areas. Communication between the 10 to 100 REGIONAL controllers is by hard wire or microwave to accommodate the high data-transmission rates.
Routing Major technical design issues in theroutingfunctionof networkcontrol concerndetailedconsiderations ofrouting algorithms and the degree to which the routing strategies will be “demand responsive.” For these issues the trade-offs rest on considerations of data availability, service capability, and service availability. Numerousroutingconcepts are feasible; however, the preferred implementation scenario consists of a centralized routing algorithm that wouldbe the recipient of real-time trafficinformation. Based on these data, routing instructions would be transmitted to each interesction in the network. One concept would have these data in the form of a list of destinations for which a diverge maneuver should be executed for those vehicles that are enroute to one of those destinations(referred to as amaneuver instruction list). The list could be given priorities so that under conditions of congestion, diverge priority could be given to those vehicles for which an abort (failure to diverge) would substantially increase the travel time for the vehicle to reach its destination. Each vehicle would be encoded with its destination.As it approaches an intersection, a comparison would be made between its destination and the diverge destination list. If a match occurred at a high enough priority (defined by the ability of the downstream merge or station area to accommodate another vehicle),
15
the vehicle’s lateral control system would proceed to steer the vehicle into the intersectioninterchange. The data requirements and rate at which maneuver instruction lists would be updated are the detailed decisions that need to be addressed for specification of a centralized routing algorithm. In terms of increasing levels of demand responsiveness, the following choices exist. 1) The generation of maneuver commands, based on operational status of each link, could minimize the distance and/or time to the destination from each intersection. This algorithm consists of asystem optimal tree-buildingminimum path finder. Maneuver instructions would require updateonly in the case of a change in the operational status of at least one h k in the network, e.g., the closing of a link. 2 ) Algorithm 1 above could be expanded to include historical data on link densities to create a routing algorithm yielding a weighted minimum distance, and a minimum expected delay for individual vehicles. This algorithm could yield unique sets of maneuver instructions for specific time periods, (say, peak, off peak, weekends, special events) and could be updated as the historical traffic data files were updated. 3) Real-time traffic density data could be used in conjunction with historical traffic data to forecast traffic in real-time. These data could serveas inputs to real-timeapplications of the routing algorithm, which would update the maneuver instructions in real time. If necessary, these updates could be generated at a rate of say, one every five minutes at time of rapidly changing demand. While the third option above would yield minimum delay for the entire network, it seems as if the data requirements of trafficforecasting are large expenses to pay for whatwould probably be marginal improvements in performance. If detailed investigations show t h s conjecture to be valid, it seems that option 2 ) will be preferred. T h s strategy requires no continuous real-timetrafficdensity data, uses datathat should be avadable from operating statistics, and contains the capability of issuing new instructionsif amajor change in status occurs in one or more links. A large part of the failure management system includes the management of vehicles around a failed vehicle. If a vehicle causes a link in the network to be closed, routing instruction must be changed to reflect the fact that a h k of the network has been closed. So that the blockage does not propagate to other links, therouting algorithm and theupdate of the maneuver instruction list must be accomplished in a very short period of time. 3) REGIONAL: The primary responsibility of the REGIONAL controller is to receive routinginformation from CENTRAL and manage the vehicles in its region accordingly. To manage these vehicles, REGIONAL utilizes sector controllers-DIAGNOSTIC, STATION/DISPATCH, INTERSECTION, and ARRIVAL. (The responsibilities of these controllers are discussed in the next section.) REGIONAL receives a preview of vehcles coming into its region from the upstream REGIONAL controller(s). Based on its real-time macroscopic vehicle flow simulation? REGIONALgenerates and transmits nominal and emergency maneuver commands to vehicles under its control.
16
TRANSACTIONS IEEE
ON VEHICULAR TECHNOLOGY,
If necessary, REGIONAL creates gaps between vehicles on the mainline and transmits this “vacant slot” information to the STATION/DISPATCH sectorcontroller so that vehicles waiting in the departure queue can merge onto the mainhe. REGIONAL relinquishes vehicle management responsibility to its SECTOR controllers when a vehicle passes into the SECT O R S control region. When a vehicle reaches its destination, the ARRIVAL sector controller monitors and manages the vehicle after it diverges from the mainline onto the exit ramp. The INTERSECTION sector controller is responsible for vehcles as the approach and proceed through intersections. REGIONAL accepts vehicle management responsibility from upstream REGIONAL controllers and its STATION/DISPATCH sectorcontroller and resumes vehicle management when vehicles leave theintersection. Region status is transmittedto CENTRAL to aid in vehicle routing and failure management. 4 ) DIAGNOSTIC: A DIAGNOSTIC sector controller is the first controller encountered by the vehicle as it enters the AHS system. The controllermonitorsthe vehicle’s entrance and performance in the obstacle course. It either directs the vehcle to the departure queue or rejects it from the system. It transmits vehicle status and destination requests to the REGIONAL controller as well as its associated STATION/DISPATCH controller. 5 ) STATIONIDISPATCH: Its responsibility is to assign each vehicle a departure time, departure position, and maneuver command code that will ensure a safe merge onto the main highway. The controllerreceives vehicle departure status/vacant slot information from the REGIONAL controller. The STATION/DISPATCH sectorcontrollertransmits thedeparture command to each vehicle in the departure queue andtransmitsa preview of vehicle arrivals andsummary data to REGIONAL. The controller monitorsthe positionandvelocity of the vehicle as it proceeds along the entrance ramp to assure that a safe merge is negotiated. 6) INTERSECTION: The INTERSECTION sector controller is responsible for safely merging vehicle streams at the intersection merge points. To accomplish the merge, INTERSECTION maintainsareal-timemacroscopicsimulation of vehicle flowin its sector from which itdetects merge conflicts and generates theproper manuever commandsto resolve these conflicts. This simulation is initialized from the vehicle preview (which contains vehicle route data) fromREGIONAL and updated by direct vehicle-to-wayside communications. Detailed specification of a merge-control policy are coupled withthe selection of the longitudinal controlsystem; gross characteristics are commonto each type.In general, each merge junction must contain a region in which trajectory informationon each vehicle ineach ofthe merging lanes is shared either between vehicles or by the local wayside computer, which then transmits maneuver instructionstothe merging vehicles, This common data region must be of sufficient length t o allow vehicles to manuever and merge without conflict. Under the vehicle-follower longitudinal control system,one or both of the merging streams of traffic can be altered so as to resolve merge conflicts. From previous investigations it
VOL. VT-28, NO. 1 , FEBRUARY 1979
seems that Brown’s [27] adaptive merging process is the most desirable. In this process a parallel data region is assigned in which the two streams of vehicles are merged and assigned a sequence in a virtual sense. Based on this new sequence, thevehicles maneuver so as to achieve a safe separation between the vehicles in the virtualsequence. Safe separation is acheived prior to reaching the end of the parallel data region. 7) ARRIVAL: The ARRIVAL sector controller, controlling each exit point, receives a preview of vehicles from REGIONAL, actuates the station diverge signal, and monitors the vehicle’s entrance into its sector. Itsreal-time simulation of vehicles in its sector is used to assign vehicles to automated-tomanual conversion bays. Maneuvers are transmittedto vehicles and summary statistics are sent to REGIONAL. 111. INTERCONTROLLER COMMUNICATIONS Large amounts of information must be passed between the various levels in the hierarchical vehicle management system. Although decisions about vehicle routing are made bythe CENTRAL controller,they are eventually executedbythe on-board vehicle control system as commands from the REGIONAL or INTERSECTION controller. The communication system links together these levels of control. Vehicles communicate only with the REGIONAL (or SECTOR) controller, which has jurisdiction over the segment of thenetworkon which it is traversing. REGIONAL controllers communicate with one another, with their SECTOR controllers, and withthe CENTRAL controller. CENTRAL interactswithother CENTRALS, its REGIONALS, and with GLOBAL. With theexception of vehicle/wayside communication,information being transferred betweenthe various controllers can be on a“hard-wired” communicationlink. Consequently,theamount of data,its accuracy,and transmission rate can be very high. The weakest link in the communication system is the vehicle/wayside link, since “hard wiring” is not feasible at typical vehicle speeds (15-30 m/s). The recently completed AHS practicality study [ l ] recommends using inductively coupled systems for the vehiclelwayside link dueto reliability, stateofdevelopment,and cost considerations. The channel modem(bit-rate capabilities) is limited toapproximately 7000 bits per second. Unlike the other links where data-vehicle routing, previews, andsummary statistics-can be delayed and transmitted inblocks, vehicle/wayside transmissionmust be in real time. Consequently? due tocommunication hardware limitations and the large number of vehicles in each region, a strict minimization of vehicle/wayside communication is required. A major design issue is the allocation of intelligence between the vehicle and the wayside. T h s allocation mustbe consistentwithand promotetheother system design objectives; e.g., maximize vehicle autonomy from wayside and minimize wayside/vehicle communication. An entire spectrum of intelligence allocationexists; however, to illustrate how the selection of allocation affects communication requirements, twobasic concepts will be examined. These are defined as follows. 1 ) Smart Vehicle: Substantial amounts of equipment and logic are onboard each vehicle. The wayside transmitscommand codes and minimal associated data. The vehicle receives
CAUDILL et al. : VEHICLE MANAGEMENT FOR AGT
17
TABLE I WAYSIDE GENERATED NOMINAL MANEUVERS-SMART VEHICLE SYSTEM Data
Maneuver
Transmitted
V e h i c l e ID number; command code; Xo;
1S)l oStl i p
Value
3)
Speed Change
Command
Vehicle ID number; command code; X,;
4)
D e c e l e r a t ei n t oS t a t i o n
Vehicle ID number; command code; VF
5)
A c c e l e r a t eo u ot fS t a t i o n
V e h i c l e ID number; command code; VF
6)
Diverge
V e h i c l e ID number; command code
XF,
VF
D-number of s l o t s s l i p p e d be
13
m i l e p o s t + number of p o s i t i o n increments (ax = l o f t . )
maneuver
x F - p o s i t i o n a t completion of maneuver 1,2,3, created)
D
Bits Required*
Parameter X - p o s i t i otinon i t i a t e
or
XF;
V e h i c l e ID number; command code; Vo
2) Constant Speed
or
Strip
X.
+
numberp o sf i ti nocnr e m e n t s
10
(or g a p st o 2
4
V ( V ) - i n i t i a l( f i n a l )v e l o c i t y
0,15,25,3On/s
2
Cornand Code
1-10
4
V e h i c l e ID Number
1 of 600*
O F
10
The longest d a t as t r i pc o r r e s p o n d st oas l o t - s l i pm a n e u v e r ,
* Assuming a region of 15,000 m,V
29 b i t s p l u s
communication number ( 1 0 b i t s ) .
= 25 m/s, h = 1 sec.
TABLE I1 WAYSIDE GENERATED EMERGENCY MANEUVERS-SMART VEHICLE SYSTEM Transmitted
Strip
Data
Maneuver
7)
Emergency Stop
8)
Emergency Stop
- maximum deceleration - nominal deceleration
Vehicle ID number; cornand code
Lateral s t e e r t o shoulder Assuming the emergency c-and
Vehicle ID number; comand code Vehicle ID number; comand code
9) Emergency start-up 10)
Vehicle ID number; cotanand code
is initiated upon receiving the comand, the largest
data strip is 14 bits.
thesecodes,generatescommand profiles onboard, senses its actual state, and adjusts the vehicle’s propulsive force to follow the command profiles. In addition, thesmart vehcle monitors its vital subsystems and alerts the wayside of impending malfunctions in order to improve the system’s anticipation of emergency situations. 2) Passive VehicZe: The wayside transmits command profile data, i.e., commanded position, velocity, and acceleration informationupdated several timesper timeheadway.The vehicle then executes these commands through its on-board controller. The main distinction between these two concepts with respect to communication requirements is that the smart vehicle generates its commanded state on-board,whereas the passive vehicle relies on the wayside to generate and transmit the entirecommandprofiles.Thisdifferenceaffects thebit-rate requirementsdue to differencesin the lengths of messages transmitted to each vehicle and the communication cycle time (essentially the update time of commands toeach vehicle).
Smart Vehicle For the smart vehicle, theinformation being transmitted contains the individual vehicle identifying number and a bit
stream composed of the maneuver command code and associated data (see Tables I and 11). For nominal commands, the communication cycling time may be an order of magnitude greater thanthe nominal time headwaybetween vehicles, since typicalmaneuversmayrequire several seconds to be accomplished. However, emergency commands will need to be transmittedona cycling timethat is less thanthe time headway. Table I11 gives the baseline parameter values of vehicle operatingand communication characteristics on which the following communication structure and bit-rate requirements are based. For a region of 15 000 m (-10 miles) in length, the wayside controller must be capable of managing 600 vehicles simultaneously ( V = 25 m/s and h = 1 s). Both nominal and emergency manuever commands will be transmitted over the same channel; consequently, the bit-rate required will be the hgher of the bit-rates for each type of command(nominal or emergency). Forthe givenmessage content (Tables I and 11) and communication cycle times (10 s for nominal and 0.25 s for emergency commands), emergency transmissions are moredemanding onthecommunication system.Infact,thecommunication channels must befre-
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. VT-28, NO. 1 , FEBRUARY
18
TABLE IV WAYSIDE-TO-VEHICLETRANSMISSION FOR PASSIVE VEHICLE SYSTEM
TABLE 111 NOMINAL SYSTEM CHARACTERISTICS AND PARAMETER VALUES 15,000 m
Length of region
__
Required‘ Information; Bits Co-mand (Value)
2 5m/ s
Vehicle velocity, V
Position, X
1 sec 2
Time headway, h
1
of :1
+
2.5m/s2
Maximum comanded jerk, J
2.5mls3
cmax
23
R
Smart Vehicle Nominal Commands 39 bits
Longest data transmission from wayside, Bnon Communication cycle time,T nom
1); (6,000,001) LXC
-6.On/s
Emergency deceleration Maximum acceleration
1979
Y). cvc ’ (6@11
Velocity, i’
1 of (1
+
Acceleratlon, AC
1
of ( I
+
Vehicle Identifying Sumber
1 of S; (600)
10 sec.
““1. cA*
I
10
(18)
10 :*ssage Length
48
bits**
*For system characteristics given in Table 111. **NO additional bits are included in this total for error-correction
Emergency commands
parity bits.
14 bits
Longest data transmission from wayside,BEm Comunication cycle tine, TEm
0.25 sec.
mandedposition ( X , ) , commanded velocity ( V , ) , and cornmandedacceleration ( A , ) . X,, V,, and A , are descretized equivalences of their respective continuous profiles. Fig. 3 indicates the commanded acceleration values the vehicle would A , and an updated receive for a quantization level on A , of A time interval (corresponding to the communicationcycle time) of T . Assuming that all nominal and emergency command profiles are based on constant jerkmaneuvers, a relationship exists between the acceleration quantization level and the communication cycle time based upon the maximum commanded jerk
Passive Vehicle Range on command position, R
15,000 m
Range on command velocity, Rv
30m/s
Range on command acceleration,R
8.5m/s2
Acceleration quantization level,
n.5m/s2
A d C
Velocity quantization level, AVc = Positionquantization level,
&xc
=
0.05n/s
MCT
T‘
0.002Sm
2 c Coumunication cycle time, T = M c / W cmax
0.10 sec.
Jcmax
T=-. A A c Commanded 5 Acceleration Ac 4
-
uhere
This relationship is based on the sampling principle [31] and is illustrated in Fig. 3. In addition? AA, and T determine the quantization levels for commanded velocity and position ( A V ,and A X , , respectively) as follows [ 181 :
max
5
hAc
0
Y
T
1
2J,max
Continuous Command Profile
Commands
,
,
,
2
3
4
AV, = A C T and
Time(s1
Fig. 3. Quantized acceleration commands actually transmitted derived from continuous commandprofile.
as
quencymultiplexed, since a single frequency using state-ofthe-art inductivelycoupled communication hardware cannot handle the high bit-rate requirements. Tables I and I1 indicate that each emergency message is 14 bitsin length. Since 600 vehicles must receive transmissions within 0.25 seconds, the bitraterequired is BEmNv/TEm= 14(600)/0.25 = 33 600 bits per second. Five frequency bands, eachaccommodating 120 vehicles and transmitting at the channel modem (7000 bit/s), would be the minimal communication requirements.
Passive Vehicle Forthe passive vehicle, each wayside-to-vehicle transmission contains the vehicle identifying number (veh ID), com-
AX, = b A A , T 2 . Table IV gives the length of each message and bits required based upon the systemcharacteristics described in Table 111. For A , = 0.5 m/s2, the communication cycle time is 0.10 s. Therefore, 48 bits of information must be transmitted to each of the 600 vehicles in the region every 0.10 s. This corresponds to a bit-rate requirement of 228 000 bits per second. For this passive-vehicle system, aset of 43 separate frequency bands are necessary, each accommodating 14 vehicles and operating at 7000 bits/s. This is compared to the smart-vehiclesystem whichrequiresonly 33 600 bit/sor 5 frequencybands. As shown in Fig. 4, the communication requirements for the passive vehicle can be decreased by increasing the acceleration quantization level; however, increasing the acceleration step size (AA,) increases the initial jerk levels with concomitant
CAUDILL MANAGEMENT e t al. : VEHICLE
19
FOR AGT
TABLE V CONTENTS AND COST ALLOCATION OF THE COMMUNICATION AND CONTROL PACKAGE
700 54, T = 0.05 600
Estimated Capital Cost($)
On-Board Equipment 500
. :: 0
500
, = 48,
175
50
100
50
50
Mount B Receptacles
T = 0.10 825-1625
2
1
225
Microprocessor
e
2
LateralControl
Radar
0
'c"_c_
200
8
100
s
42. T = 0.20
207-407
-,-
year'
Costs
per
Bit-Rate Required for Smart Vehicle
($1 C and C Package
A%
LongitudinalControl Communication
400
Allocated Cost Owner
150
150
200-1000
200-1000
100
100
50
50
Total
275
550-1350
69
Cost
138-338
Cost permile of AiiS C&C Package*'
0.291
-
0.681
'Assuming 5 - y e a r l i f e a t 8%. *'Assuming eachpackageused 50,000 a i l e s / y e a r .
= 36, T = 0.4
B = 28, T =
1.0
0
0
1
2
3
4
5
AcceleratlonQuantization Level
\
My, MIS2
Fig. 4 .
Bit-rate as a function of acceleration quantization level for passive-vehicle concept.
30
,
\ equipment high-qualicy radar \fUser-ovned
I decreases in passenger ride comfort. In addition, large values of Uc may lead to instabilities inthe vehicle's control systems. Consequently,thecommunication requirements forthe passive vehicle system are significantlygreater,perhaps even an order of magnitude greater, than for thesmart-vehicle system. IV. COMMUNICATION AND CONTROL PACKAGES FOR AHS
1°C
II
\
. \
use of
User-owned e4uI~ment low-quality radar
.1
C and C package v i t h
I
1000
I
2000
I
3000
I
4000
1
5000
Annual-nile6 of WS Uae
The major disadvantage of the smart-vehicle concept is that the additional on-board equipmentmay be costly to the owner of the vehicle. This is especially true if he is only an occasional AHS user. The economics of this equipment, if totally bome by theowner, may dictatethat an individualmust use the AHS at least a thousand miles per year. T h s could drastically reduce the patronagefor AHS. T h s is a majorconcernfor AHS, since it consists of privately owned vehicles. The cost borne by the AHS user of the additional on-board equipment can be reduced substantially by the use of systemowned communication and control (C&C) packages. This package would be leased by the AHS user, attached to the vehicle upon entering, and removed when the vehicle leaves the system. Since each package will be used repeatedly, the user cost will be minimal. This leasing arrangement makes the "smart" vehicle concept economical, even for the occasional user. The equipment contained in the C&C package would be the type of hardware that isused exclusively onthe AHS and which could be readily attached to the vehicle. Table V lists 6 components which would be contained (at least in part) in thecommunication and control package. The cost of each component is estimated and an allocation between user-owned and C&C package is made. To illustrate the economics of the C&C package, consider the cost associated with this on-board equipment for an indivi-
Fig. 5 . Comparison of totally user ownedequipmentwith C and C package on a user cost per mile basisas a function of annual AHS mileage.
dual who uses the AHS 1000 miles each year. Assuming that if the equipment is totally user owned, the mountand receptacles are unnecessary, then the annualuser cost of this equipment is $194-$395 (depending on the quality of the radar purchased). This results in equipment cost of 19.44 and 39.54 per mile of AHS use. If the individual purchases the mounting and receptacles and leases the C&C package, his cost reduces to 7.24 to 7.64 per mile. It should be noted that the cost of the C&C package equipment is only 0.3 to 0.7d of the 7.2 to 7.6k per mile; consequently, very reliable and highqualitycomponents can be used in the C&C package with only marginal increases in user cost. A comparison of the economies of totally owned equipment with the C&C package on a user cost-per-mile basis as a function of annual AHS usage is presented in Fig. 5 . Given a threshold value of user cost per mile, these curves indicate theannual AHS usage necessary to achieve this threshold. As an example, let this threshold value be 104 per mile. That is, an individual would not invest in theequipment necessary for AHS if it would cost more than 10d per mile to use theequipment. At the 104 per mile threshold value, the
20
TRANSACTIONS IEEE
ON VEHICULAR TECHNOLOGY, VOL.
occasional user would be satisfied with only 750 miles annual use, if C&C packages were available. If theequipment is totally user owned, thisthreshold is not reached until 2000 and 4000 miles per year, forlow-quality and high-quality equipment, respectively. For AHS to be a practical intercity mode, the cost borne by the individual user mustbe minimized while the performance, safety, and reliability of AHS are maximized. Leasing system-owned communication and control packages by AHS users makes the “smart” vehicle concept economical, even for the occasional user, thus achieving both of these goals.
V. CONCLUSIONS Vehicle management for an automated vehicle system is based on a hierarchy of controller functions.Careful consideration must be given to the formulation of the management scheme inorder to assure thatthecommunication, vehicle control, and network management requirements are achievable. The inherent characteristics of the automated vehicle system significantly impact the design requirements for longitudinalcontrol and vehicle management.The sheer size of small vehicle AGT and AHS systems, in terms of number of vehicles and physical dimension, dictate that wayside/vehicle communications, wayside computer storage, and computational requirements be minimized. To achieve these goals requires that each vehicle be as autonomous from thewayside as possible; consequently, substantial amounts of intelligence must be placed on-board each vehicle. With the smart vehicle concept the communication requirements are significantly less than for the passive vehicle system. In fact, several-fold reduction may be possible. However, the additional on-board equipment may be costly to the owner. To overcome these additional costs, system-owned communication and control packages are recommended for AHS systems. With respect to longitudinal control,the vehicle-follower control concept is preferred over point-follower if the problem of information transfer during the merging process can be resolved. The linear property of the AHS network makes vehicle routing somewhat simplistic while imposing stringent requirementson failure management. To enhance failure detection and prevention, each vehicle monitors its ownvital subsystems and alerts the wayside if a malfunction occurs. This monitoring system is a component of the “smart” vehicle concept.
23 1
REFERENCES [ 11 Calspan Corp., Barton-Aschman, Assoc. and Princeton University, “PracticalityofAutomated HighwaySystems,”Final Report, DOT-FH-11-8403, NoV. 1977. [2] Kirk,P. M., “Automated personaltransit control systems,” ASME Paper ##73-ICT-41,1973. [3] Lang, R. P.,“MorgantownPersonalRapidTransitLongitudinal Control System Design Summary,” DOT Report No. UMTA-MA06-0048-75-4, December, 1975. M., “NormalandEmergency [4] Whitney,D.E.,andTomizuka, Control of a String ofVehiclesby Fixed Reference SampledDataControl,” ZEEE Trans. on Vehicular Technology, VT-21, pp. 128-138, November, 1972.
24 1
VT-28, NO. 1 , FEBRUARY 1979
Garrard, W. L., and Kornhauser, A. L., “Design of Optimal FeedbackSystemsforLongitudinalControlofAutomatedTransit Vehicles,” Transportation Research, 7, June, 1973. Garrard, W. L., and Kornhauser, A. L., “Use of State Observers in Optimal Feedback Control of Automated Transit Vehicles,” ASME Journal of Dynamic Systems, Measurement, and Control, June, 1973. Brown, S. J., “Design Considerations for Vehicle State Control by thePoint-FollowerMethod,” PersonalRapidTransit IZ, J. E. Anderson, Ed., Univ. of Minnesota, 1973. Fenton, R. E., Bender, J. G., Mayhan, R. J., Brinner, T. R., “On Synchronous Longitudinal Control for Automated Ground Transport-TheoryandExperiment,”7thAnnualPrinceton Conference on information Sciences and Systems, March, 1973. Larson, V., “AnOptimalStochasticControllerforAccurate Position Control (Personal TransportationStudy)” Aerospace C o p . Report #ATR-72 (8124)-1, El Segundo,CA. 1971. Brown, S. J., “Characteristics of a Linear Regulator Control Law AIAA forVehicles in anAutomaticTransitSystem,”1971 GuidanceandControlConference,StonyBrook, New York, August, 1971. TypeControlSystems Brown, S. J., “DesignofCar-Follower with Finite Bandwidth Plants,” 7thAnnual Princeton Conference oninformation Science and Systems,Princeton,NewJersey, March, 1973. Garrard, W. L., Hand, R. G., Raemer, R., “Suboptimal Feedback Control of a String of Vehicles Moving in a Single Guideway,” Transportation Research, 6, June, 1972. Garrard, W. L., Caudill, R. J., and Reed, W. B., “Longitudinal Controland Crashworthiness for Small AutomatedTransit Vehicles,”NTISPB-243-353,Springfield,Va., January,1976. Caudill, R. J., and Garrard, W. L., “Vehicle-follower,LongitudinalControlforAutomated GuidewayTransitVehicles,” to Appear in High Speed Ground Transportation Journal, 1977. Garrard, W. L. and Caudill, R. J., “Dynamic Behavior of Strings of Automated Transit Vehicles?’ SAE paper No. 77028, 1977. Caudill, R. J., “Vehicle-Follower Longitudinal Control for Small AutomatedTransit,” PH.D.Thesis,Univ.ofMinnesota, 1976. (AlsopublishedbyUrban Mass TransitAdministration, U.S. DOT as Report No. UMTA-MN-11-0002-77-1.) Hadju, L. P., et al.,“Design and Control Consideration for Automated Ground Transportation Systems,” Proceeding ZEEE, 56, pp. 493-513, 1968. Fenton, R. E., et al., “Fundamental Studies in the Longitudinal Control of AutomatedGround Vehicles,”FederalHighway AdministrationReport No.FHWA-RD-77-28,December 1976. Garrard, W. L., “Longitudinal Control for Automated Guideway Transit Vehicles,” U.S. DOT Urban Mass Transit Administration Report No. MN-ll-002-77-2-UMTA, July 1977 MacKinnon,Duncan,“TechnologyDevelopmentforAdvanced Personal Rapid Transit,”Personal Rapid Transit IZ, J. E. Anderson, Ed. University of Minnesota, Dept. of Audio-visual Extension, 1974, pp. 57-64. Morag, D., “Operating Policies for Personal Rapid Transit,” U.S. DOT Urban Mass Transit Administration, Report No. RDD-8-742, May, 1974. Caudill, R. J. and Garrard,W. L., “Vehicle-Follower Longitudinal Control for Automated Transit Vehicles,” to Appear in ASME Journal of Dynamic Systems, Measurement and Control, 1977. Munson, A.V., etal., “Quasi-SynchronousControl of HighCapacity PRT Networks”, Personal Rapid Transit,J. E. Anderson, J. L. Dais, W. L. Garrard, and A.L. Kornhauser, Eds. Department of Audio-visual Extension, University of Minnesota, pp. 325-350,1972. Brown, S. J., “Merge Control in AutomatedTransitSystem Networks,” ASME Publication, 73-ICT-100, 1973. Cauda, R. J. and Youngblood, J. N., “A Study of VariousMerge Control AlgorithmsApplicable toAutomatedTransitSystems Under Quasi-Synchronous Control,” Personal Rapid Transit ZZZ, University of Minnesota, 1976. Irving J. M., et al., “VehicleManagement on LargePRTNetworks,” Personal Rapid Transit IIZ, Editors D.A. Gary, W. L. Garrard, A.L. Kornhauser, Department of Audio-visual Extension, University of Minnesota, Minneapolis, 1976. Brown, S. J., “Adaptive Merging Under Car-Follower Control,” Johns Hopkins Applied Physics Lab. Report, August, 1975.
CAUDILL MANAGEMENT er al. : VEHICLE
FOR AGT
[28] Wikie, D. F., “A Moving Cell ControlSchemeforAutomated TransitSystems,” TransportationScience, pp. 347-364, 1970. [29] TRW SystemsGroup,“StudyofSynchronousLongitudinal Guidance as Applied toIntercity Highway Networks,” NTIS PB180582, Sept., 1969. [30] Bender,J. G., andFenton, R. E., “On Vehicle Longitudinal Dynamics,” Traffic Flow andTransportation, Gordon Newell, Ed., Elsevier Press, 1973. [31] Stein Jones, and Modern Communication Principles, McGrawHill, 1967.
21
Main L. Kornhauser was borninFrance,on June 12, 1944. He received the B.S. and M.S. degrees in aerospace engineering fromPennsylvania State University in 1965 and 1967, the and respectively, Ph.D. aerospace degree in andmechanical sciences fromPrinceton University in 1971. He joined the faculty of the University of Minnesota in 1971 as an Assistant Professor of Aerospace Engineering and Mechanics. In 1972 he became an Assistant Professor in Civil Engineering and Associate Director of the Transportation Program at PrincetonUniversity.Thereafter,in 1976 hebecame Associate Professor and Director of the Transportation Program. Most recently he has been on sabbatical from Princeton as a Visiting Professor at MassachusettsInstitue of Technologyand on special assignmentinthe Resgie J. b u d i l l was born in Branson, MO, Department of DevelopmentandPlanning atPortAuthorityof New on October 20,1949. He received the B.S. York and New Jersey.Recentprimary research interests have included degree inmechanical engineering andtheautomated vehicle systemsandrailroadfreighttransportationmodeling. MS. degree in engineering mechanics from Dr. Kornhauser is theauthor of several papers in bothof these the University of Alabama in 1971 and 1973, fields. Heis active in an array of professional societies and has taught a respectively,andthePh.D. degree in mech- variety of courses in transportationandquantitativemethods. anical engineering from the University of Minnesota in 1976. After receiving his undergraduatedegree, he was employed as an engineer in theDynamics Analysis Branch o f Teledyne-Brown Engineering Company(Teledyne-Brown is a NASA contractor in John R. Wroble was born in St. Louis, MO, Huntsville, AL). Upon completing his Ph.D., he joined the faculty at on January 2, 1955. He received the B.S. in the University of MissouriColumbia, where hetaughtcourses degree,withhonors,in civil engineering from transportationsystemsandplanning.Sincethesummer of 1977, he the University of Missouri, Columbia, in 1976, has been with the Transportation Program at Princeton University as and the M.S. degree from Princeton University an Assistant Professor ofCivil Engineering. His primary teaching and in 1978. research activitiesconcerntheapplication of dynamicsandcontrol He is currently an Engineer for Alan M. techniques to transportation problems. Voorhees and Associates, Inc. in McLean, of several papers oncommunication, Dr. Caudill is theauthor VA, where he is working in the area of transcommand, and control for automated transportation systems. portation operations.
