System optimal routing of autonomous vehicles under equilibrium

Abstract

Different demand management strategies have been proposed in the literature. While tolling is a common approach to achieve system optimum for an entire network, it has been criticized for its potential unfairness to certain demographics. But is it possible to maintain system optimum performance for the whole network without relying on tolling? This is the potential of a routing strategy that could be implemented with the advent of autonomous vehicles. The routing of autonomous vehicles under a potential transportation management centre has been a topic of significant interest in recent years, offering a promising future for transportation management. In this regard, a transportation management centre is envisaged for future autonomous vehicles that distributes traffic over a cycle such that the optimal system flows are maintained every day and travellers from the same origin destination pair experience minimal and similar average travel time at the end of a fixed cycle (number of days), thereby preserving the notion of equilibrium. Travellers are assigned to distinct paths within different path combinations according to time proportions within the fixed cycle. At the end of cycle, all travellers from same origin destination pair will attain equal average travel times, thereby creating a 'win-win scenario' for both the system and individual. This means that the system operates at its most efficient level, while individual travellers experience the least possible average travel time. The thesis formulates the problem as a non-linear complementarity problem and proposes a gap function approach for solving it efficiently. It also investigates the impact of user subscription to the transportation management centre, leading to the imposition of tolls on non-subscribers to encourage subscription. Sensitivity analysis concerning multiple parameters such as market penetration, tolls, value-of-time and average travel time is investigated. Additionally, the research extends the concept to multi-modal transportation management centre, addressing congestion and mode choices, thereby introducing the concept of multi-modal average travel time equilibrium. Travellers from the same origin destination pair are allowed to travel on different modes with different path sets according to the routing instructions by the centre such that they experience the same multi-modal average travel time at the end of a fixed cycle. The thesis also delves into different multi-modal system optimum concepts and explores toll imposition to offset the advantage of non-subscribers in the multi-modal scenario. Sensitivity analysis concerning multiple parameters such as market penetration of different modes, crowdedness cost, value-of-time, tolls and multimodal average travel time is investigated. Lastly, the thesis presents a solution that converts the formulation from a non-linear complementarity problem to a linear problem, making it manageable for extensive networks. This transformation significantly reduces the computational burden and overall run times, demonstrating the practicality and efficiency of the proposed approach. The thesis also extends the concept of conventional average travel time equilibrium to average vehicle travel time and emission cost equilibrium, further enhancing the effectiveness of the proposed solution. The case of toll imposition on non-subscribers for extensive networks is thoroughly discussed, providing a comprehensive understanding of the dynamics between parameters such as tolls, values-of-time, and travel cost differences between subscribers and non-subscribers. Overall, the thesis offers a new and efficient way of utilizing autonomous vehicles as regulators to achieve system optimal performances and equilibrium in transportation networks without resorting to tolling or sacrificing the travel times of certain travellers.

Date of Award	2024
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Hong Kam LO (Supervisor)