Performance of Synthetic Road Networks with Dedicated Streets for Altruists

Abstract

We present an empirical study of network ensembles under static traffic assignment with two vehicle classes: an altruistic vehicle class (AVs) – potentially automated vehicles that are optimally routed – and a selfishly routing class, corresponding to human-driven vehicles (HVs). We assess a management measure that consists on reserving some network links for exclusive use by AVs, to mitigate the detrimental effects of HVs on the per-vehicle cost of AVs. We formulate a capacitated, multi-class, static traffic assignment problem (STAP). We solve the STAP as a single-objective, convex optimisation problem with additional flow constraints on AV-exclusive links. We find that randomly selecting the AV-exclusive links can be detrimental at low penetration rates. However, having several sets of links to choose from and selecting the best performing link-set for the traffic conditions steers the system closer to optimal even for modest penetration rates (e.g., 30%).