关键词: 智能车辆, 轨迹跟踪, 横向控制, 模型预测控制, 强化学习

Abstract: To address the problem that intelligent vehicles struggle to balance trajectory tracking accuracy and driving stability during lateral trajectory tracking,a lateral control strategy for intelligent vehicle trajectory tracking based on Reinforcement Learning combined with Model Predictive Control (RL-MPC) is proposed. First,a vehicle system dynamics model is established. Then,relying on the receding horizon optimization framework of Model Predictive Control (MPC),the trajectory tracking accuracy is quantified in real time. Meanwhile,a reward function is constructed to realize the autonomous optimization of the control strategy,and a feedback correction module is used to correct the prediction deviation. Finally,a simulation model is built using Matlab/Simulink software for simulation tests,and the RL-MPC controller is compared with the traditional MPC controller. The experimental results show that under different working conditions,compared with the traditional MPC controller,the RL-MPC controller achieves better trajectory tracking performance,significantly reduces the lateral tracking error,and improves the handling stability under driving conditions.

Key words: intelligent vehicles, trajectory tracking, lateral control, Model Predictive Control (MPC), Reinforcement Learning (RL)