Proximal Policy Optimization-Based Driving Control Strategy of Connected Cruise Vehicle Platoons to Improve Traffic Efficiency and Safety

This paper investigates connected cruise control (CCC) based on the deep reinforcement learning algorithm to mitigate oscillations and improve traffic safety in stop-and-go waves. A proximal policy optimization (PPO)-based control strategy is proposed for connected and autonomous vehicles (CAVs) in a heterogeneous longitudinal car-following platoon. The method receives the velocity and position signals of n human-driven vehicles ahead through wireless vehicle-to-vehicle communication to obtain an appropriate driving behavior and improve traffic efficiency and safety in real-time. The effectiveness and advantage of the proposed strategy are verified by using traffic simulation software (Simulation of Urban Mobility) for two simulation scenarios of trajectory curves with noticeable acceleration changes of the leading vehicle and two sections of actual traffic speed data. The results show that the proposed PPO-CCC method can successfully suppress the speed oscillations and improve the travel efficiency of CAVs in the car platoon.