Roman, Charlotte Daisy (2021) Routing choices in intelligent transport systems. PhD thesis, University of Warwick.

Preview

PDF WRAP_Theses_Roman_2021.pdf - Submitted Version - Requires a PDF viewer. Download (4Mb) | Preview

Request Changes to record.

Abstract

Road congestion is a phenomenon that can often be avoided; roads become popular, travel times increase, which could be mitigated with better coordination mechanisms. The choice of route, mode of transport, and departure time all play a crucial part in controlling congestion levels. Technology, such as navigation applications, have the ability to influence these decisions and play an essential role in congestion reduction. To predict vehicles' routing behaviours, we model the system as a game with rational players. Players choose a path between origin and destination nodes in a network. Each player seeks to minimise their own journey time, often leading to inefficient equilibria with poor social welfare. Traffic congestion motivates the results in this thesis. However, the results also hold true for many other applications where congestion occurs, e.g. power grid demand. Coordinating route selection to reduce congestion constitutes a social dilemma for vehicles. In sequential social dilemmas, players' strategies need to balance their vulnerability to exploitation from their opponents and to learn to cooperate to achieve maximal payouts. We address this trade-off between mathematical safety and cooperation of strategies in social dilemmas to motivate our proposed algorithm, a safe method of achieving cooperation in social dilemmas, including route choice games. Many vehicles use navigation applications to help plan their journeys, but these provide only partial information about the routes available to them. We find a class of networks for which route information distribution cannot harm the receiver's expected travel times. Additionally, we consider a game where players always follow the route chosen by an application or where vehicle route selection is controlled by a route planner, such as autonomous vehicles. We show that having multiple route planners controlling vehicle routing leads to inefficient equilibria. We calculate the Price of Anarchy (PoA) for polynomial function travel times and show that multiagent reinforcement learning algorithms suffer from the predicted Price of Anarchy when controlling vehicle routing. Finally, we equip congestion games with waiting times at junctions to model the properties of traffic lights at intersections. Here, we show that Braess' paradox can be avoided by implementing traffic light cycles and establish the PoA for realistic waiting times. By employing intelligent traffic lights that use myopic learning, such as multi-agent reinforcement learning, we prove a natural reward function guarantees convergence to equilibrium. Moreover, we highlight the impact of multi-agent reinforcement learning traffic lights on the fairness of journey times to vehicles.

Item Type:	Thesis (PhD)
Subjects:	H Social Sciences > HE Transportation and Communications Q Science > QA Mathematics T Technology > TE Highway engineering. Roads and pavements T Technology > TL Motor vehicles. Aeronautics. Astronautics
Library of Congress Subject Headings (LCSH):	Intelligent transportation systems, Traffic congestion -- Management -- Mathematical models, Traffic flow -- Mathematical models, Automated vehicles
Official Date:	2021
Dates:	Date Event 2021 UNSPECIFIED
Institution:	University of Warwick
Theses Department:	Mathematics for Real-World Systems Centre for Doctoral Training
Thesis Type:	PhD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	Turrini, Paolo
Format of File:	pdf
Extent:	x, 176 leaves : illustrations
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item

Downloads