Matching and routing for shared autonomous vehicles in congestible network

Abstract

The dispatch and routing problems of shared autonomous vehicle (DR-SAV) have been widely studied, and one of the key challenges is the modeling complexity and computational burden associated with network congestion. Existing studies on DR-SAV are limited in the sense that traffic congestion and demand–supply interaction, resulting from dispatch and routing decisions, are not fully integrated into the optimization framework. In this work, we explicitly consider the congestion effect of SAV operations in a mixed traffic environment, consisting of SAVs and conventional vehicles (CVs), by proposing a computationally tractable traffic assignment framework for optimal matching and routing of SAV trips, while allowing the sharing of vehicle by up to two trips simultaneously. This problem is formulated as a Stackelberg game where the upper level is inherently a matching and routing problem for SAVs, and the lower level involves a user equilibrium among CVs. Two strategies are proposed to improve the tractability of the proposed problem: (1) a novel convex programming formulation of the joint SAV matching–routing problem based on the system optimal traffic assignment principle, and (2) the invocation of shareability network to facilitate path set generation. Numerical experiments of the proposed method show that the proposed SAV matching and routing scheme could lead to significant reduction in total travel cost.