Modelling reservation-based shared autonomous vehicle services: A dynamic user equilibrium approach

Abstract

Shared Autonomous Vehicles (SAVs) are expected to be used for regular and pre-planned trips. Such trips are suitable for reservation-based services, wherein the customer needs to book for a trip in advance. Systems enabling reservation of trips can allow for better planning of routes and schedules, and if optimally designed, enable higher efficiency. The primary objective of this research is to model the effects of such a system, by formulating and solving the combined Dynamic User Equilibrium and Shared autonomous vehicle Chain Formation (DUESCF) problem. The problem is formulated as a bilevel model based on game theory, involving road users and SAV service operator. Given a situation where conventional private and shared autonomous vehicles co-exist, road users select paths and departure times to maximize a perceived utility (commonly treated as minimizing a disutility) by forming a DUE (fixed point problem), and the SAV service operator tries to maximize the performance by forming appropriate SAV chains (combinatorial problem). The final objective of this bilevel model is a traffic assignment that includes SAV chain formation, such that both road users and SAV service operator obtain optimal solutions by reaching a Nash equilibrium, where no player is better off by unilaterally changing their decisions. A solution approach, based on Iterative Optimization and Assignment (IOA) method, is proposed with path flow and SAV performance changes as convergence criteria. Furthermore, the solution approach is tested for its robustness, and a scenario analysis is carried out to evaluate the impacts of reservation-based SAV services. The results show that a ridesharing SAV system is better compared to a carsharing and a mixed system consisting of both, in terms of total system travel time, congestion levels, total vehicle kilometres travelled and vehicle requirements.