DDPG-Based Resilient Control for CACC Under Communication Delays and DoS Attacks

Cooperative adaptive cruise control (CACC) has steadily gained recognition as a key technology for reducing traffic congestion and enhancing safety, driven by the rapid advancements in autonomous driving and intelligent transportation systems. Traditional control techniques often struggle to address complex traffic scenarios due to their inflexible parameter adjustments and limited adaptability to dynamic environments. To maximize the benefits of cooperative driving in multi-vehicle queues and effectively address network communication delays and denial of service (DoS) attacks, this research explores a reinforcement learning methodology to enhance CACC systems, where vehicle acceleration control are optimized through the deep deterministic policy gradient (DDPG) framework. A multi-vehicle simulation model is developed using the joint simulation platform of CarSim, Simulink and Python for assessing the system's operational effectiveness. The experimental results demonstrate that the DDPG-based CACC system achieves more precise and stable following control in dynamic traffic environments.