Decentralized Game-Theoretical Approaches for Behaviorally-Stable and Efficient Vehicle Platooning

Abstract

Cooperative vehicle platooning enabled by connected automated vehicle (CAV) technology has shown to bring various benefits including energy savings and a reduction in driving effort. Nevertheless, because these benefits vary over different platoon positions, vehicles from different owners may not be willing to platoon together; even if they form a platoon, they may attempt to change positions. To address such a behavioral-instability issue, it is necessary to redistribute the benefits among platoon members. To this end, this study investigates a decentralized multi-agent system where individually rational agents form platoons through peer-to-peer coordination under designated mechanisms that simultaneously determine the benefit reallocation. Depending on whether the scope of coordination is one-to-one or many-to-many, we introduce two types of mechanisms based on the bilateral trade model and one-sided matching. As the privacy of information sharing in the decentralized system is a common concern in practice, we further discuss two settings under each mechanism, differing by whether complete information is or is not known by the other agents. We indicate both theoretically and numerically that the decentralized platooning system is flexible and scalable, and can be implemented in real-time by leveraging the CAV technology.