Platooning Connected, Autonomous, and Human-Driven Vehicles: A Deep Reinforcement Learning-based Approach

This study introduces a hybrid platooning approach that allows non-connected vehicles to join platoons, addressing limitations in existing models that focus solely on connected vehicles. Utilizing deep reinforcement learning (DRL), the proposed control strategy incorporates vehicle dynamics and traffic flow states to optimize platoon structures, achieving a balance between traffic throughput and stability. Numerical simulations indicate that this method effectively reduces disturbance propagation, enhancing overall traffic stability, safety, and environmental efficiency in mixed traffic conditions.