湖北汽车工业学院学报2025，Vol.39Issue(2)：12-18,7.DOI:10.3969/j.issn.1008-5483.2025.02.003
基于DDPG算法的智能车辆轨迹跟踪控制
Trajectory Tracking Control of Intelligent Vehicles Based on DDPG Algorithm
摘要
Abstract
To realize the trajectory tracking control of intelligent vehicles,a reward function combining stepped and proportional functions based on the deep deterministic policy gradient(DDPG)algorithm was proposed.A two-degree-of-freedom vehicle model was established,with lateral position error and yaw angle error taken as optimization objectives.A Critic-Actor evaluation network was constructed;a hybrid reward function combining proportional and stepped functions was designed;the state quantity and action space of the agent were defined;control accuracy,lateral stability,and adaptability were con-sidered as optimization indicators.Based on the MATLAB/Simulink platform,simulation experiments under both straight-line and double-lane-change scenarios were conducted.These experiments veri-fied the superiority of the proposed reward function in terms of tracking accuracy and vehicle stability.The simulation results under lane-change scenarios demonstrate the high tracking accuracy and good adaptability of the controller.关键词
深度强化学习/轨迹跟踪/自动驾驶/奖励函数Key words
deep reinforcement learning/trajectory tracking/autonomous driving/reward function分类
计算机与自动化引用本文复制引用
杨烜,冯樱,高峰,何建彪..基于DDPG算法的智能车辆轨迹跟踪控制[J].湖北汽车工业学院学报,2025,39(2):12-18,7.基金项目
中央引导地方科技发展专项(2019ZYYD019) （2019ZYYD019）
