Multi-Objective Parallel Human-machine Steering Coordination Control Strategy of Intelligent Vehicles Path Tracking Based on Deep Reinforcement Learning

• Proposes a deep reinforcement learning-based multi-objective parallel human-machine steering coordination strategy for path tracking, addressing driver misoperation and external disturbances. • Integrates a driver steering model with Tube MPC for robust path tracking, and employs DQN, DDPG, and TD3 algorithms to optimize coordination. • Demonstrates via simulations and hardware-in-the-loop experiments that the TD3-based strategy outperforms fuzzy and other DRL methods in tracking accuracy, lateral safety, and reduced human-machine conflict. • Provides a comprehensive evaluation framework with multiple indices (tracking accuracy, lateral safety, human-machine conflict, driver steering load) under varied driving conditions.