A Low-level PRT Microsimulation

by Markus T. Szillat

PhD thesis, University of Bristol, U.K. 218 pp., 2000

Main document and Appendix A is available as a pdf document


This dissertation examines low-level control of an automated public transit system known as Personal Rapid Transit (PRT). This system consists of a large network of small (automobile-sized), computer-controlled vehicles. It provides passengers with on- demand, non-stop travel similar to a taxi. It should provide many benefits to individuals and society - greatly improved safety, reliability, and accessibility compared to the automotive system. Vehicles should also be more efficient in many respects: energy consumption, land usage, labour usage, and resource usage for construction of the system and vehicles. The idea dates back to at least the 1950's. Various concepts and early prototypes are presented.

A fairly common goal for PRT systems is to operate large numbers of vehicles at close spacings, which introduces safety issues. This thesis is primarily concerned with "low- level" control - the safe motion of vehicles. Various safety concepts from other transit systems are examined. These are compared to the needs of PRT. A methodology called "model follower" is selected. In it, vehicle controllers plan trajectories of future motion based on present information. This methodology is presented in some depth. As evaluated in this thesis, vehicles plan trajectories consisting of segments defined by constant jerk (change in acceleration over time), acceleration, or velocity.

Initial simulations showed that finding a single algorithm which could generate safe trajectories in all circumstances is quite difficult. To ensure safety, numerical methods were added. These evaluate the safety of planned trajectories against local vehicles and speed restrictions. This evaluation provides for the possibility that leading vehicles will alter their trajectories and ensures that the vehicle in question will be able to safely react. As vehicles receive new information, new trajectories must be generated.

A microsimulation was coded to examine this control method. The method was then expanded to cover operation at junctions. Algorithms were also added to deal with adjusting trajectories due to insufficient power. Detailed descriptions of the algorithms employed are given. Poor performance of algorithms which only examine safety issues led to the examination of optimisation techniques for junctions. Optimising algorithms attempt to provide the vehicle controllers with additional information before it is required for safety purposes. As a result, vehicles can adjust to merging future vehicles well in advance of junctions - yielding improved throughput.

Initially, the research was intended to examine high-level control issues - roughly divided into routing and scheduling. However, coding low-level control proved to be quite complex. Even though high-level control was not simulated, many basic concepts were examined. In particular, the shortest path problem, which is critical to routing, was examined in some depth. A possibly new, fast, shortest path algorithm is presented in the appendix. Methods for fault recovery, potential station layouts, and more general high-level concepts such as determinism are presented.

Finally, conclusions from the research to date are drawn. These indicate that while significant issues remain, the chosen model follower control method combined with safety verification techniques may provide a reasonable basis for safe, high- performance PRT operation.

Table of Contents (major headings only)

1. Introduction (5 pp)

2. Control Strategies and Safety Implications (8 pp)

3. Automatic Vehicle Operation (21 pp)

4. Early Simulation Models (5 pp)

5. Detail Description of Safety Evaluation Method (15 pp)

6. Detailed Safety Evaluation at Intersections (22 pp)

7. Trajectory Generation Algorithms (31)

8. Conflict Zone Optimisation Controllers (27)

9. Other Research, Future Tasks and Conclusions (33 pp)

Appendix A - Sequential All Pairs Shortest Path Algorithm

Appendix B - Linear Shift Feedback Register Encoding

Appendix C - Infrastructure Differentiation

Appendix D - Leader/Follower Arbitration Examples

Appendix E - List of Symbols Used

Endnotes and References

Includes 60 figures in the nine chapters and 6 in the five appendices