Mixed-Autonomy Traffic and Wireless Charging Problem for Autonomous Electric Vehicles Using Reinforcement Learning

Abstract

As the adoption of electric vehicles (EVs) continues to increase, integrating vehicle-to-grid (V2G) technology while maintaining mobility and optimizing traffic flow becomes crucial. In this paper, we formulate a mixed-autonomy traffic and wireless charging problem where both autonomous and human-driven EVs coexist on the road. We aim to mitigate traffic congestion and improve V2G functionality, and therefore we propose a framework that adopts reinforcement learning, specifically the Soft Actor-Critic algorithm, to jointly enhance traffic throughput, balance the state-of-charge (SOC) among EVs, and maximize wireless charging energy in the mixed-autonomy ring-shaped road with wireless charging facilities. Experimental results demonstrate the effectiveness of our approach in stabilizing traffic, achieving SOC balance, and increasing the energy charged to EVs. This study showcases the potential for solving mixed-autonomy traffic and wireless charging problems in future wireless charging transportation systems.