Smart Navigation via Strategic Communications in a Mixed Autonomous Paradigm

Abstract

This paper investigates the optimal information design for a system to minimize congestion cost in the presence of both autonomous vehicles (AVs) and human-driven vehicles (HVs). We incorporate asymmetric information between AVs and HVs in a routing game where there are two routes available and one of them has a state-dependent congestion cost. AVs are informed of the state and make choices as a fleet while HVs rely on information provided by the system and make self-interested choices. The system designs information in a Bayesian persuasion manner aiming to mitigate HVs' selfish routing such that traffic congestion cost is minimized. We show that the penetration of AVs can mitigate HVs' overcrowding problem and the first-best can be achieved when the fleet size of AVs reaches a high level. We find that it is optimal for the system to randomize in providing traffic information rather than to provide perfect information to HVs. When the information distortion is mild, HVs overcrowd the more desirable route as in the complete information benchmark. When the information distortion is strong enough, their behaviors are flipped and overcrowd the less desirable route. Interestingly, our research sheds light on the interaction of AV platooning and information provision. Finally, credibility constraint limits the social planner's persuasion power in navigating HVs away from overcrowding the more desirable route. When the fraction of AVs is high enough, the first-best can be achieved through the synergy of AV platooning and information provision.