Multi-Agent Learning Empowered Collaborative Decision for Autonomous Driving Vehicles

Abstract

Autonomous vehicles play an important role in intelligent transportation systems. In these vehicles, driving control decision is obtained based on the collection of massive traffic states and intensive information processing. However, the spatial-temporal characteristics of the traffic states and the constrained environmental perception range of an individual vehicle seriously undermine the effectiveness of the state collection. Multi-agent empowered collaborative decision provides a potential approach to address the problem. This paper proposes a multi-dimensional information fusion mechanism, which improves the utilities of vehicular information processing and autonomous driving. Moreover, we design an intelligent distributed decision algorithm for autonomous driving applications, which optimizes road traffic flow under vehicular resource constraints. Numerical results demonstrated that our proposed scheme significantly increases the system revenue.