Deep reinforcement learning-based drift parking control of automated vehicles

作     者：LENG Bo YU YiZe LIU Ming CAO Lei YANG Xing XIONG Lu 

作者机构：School of Automotive StudiesTongji University 

出 版 物：《Science China(Technological Sciences)》 (中国科学（技术科学英文版）)

年 卷 期：2023年第66卷第4期

页      面：1152-1165页

核心收录：

学科分类：12[管理学] 08[工学] 0810[工学-信息与通信工程] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 082304[工学-载运工具运用工程] 081104[工学-模式识别与智能系统] 080204[工学-车辆工程] 0805[工学-材料科学与工程（可授工学、理学学位）] 0802[工学-机械工程] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0702[理学-物理学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 0823[工学-交通运输工程] 

基　　金：supported by the National Key R&D Program of China(Grant No.2021YFB2501201) the National Natural Science Foundation of China(Grant No.52002284) the Young Elite Scientists Sponsorship Program by China Association for Science and Technology(Grant No.2021QNRC001) 

主　　题：automated vehicles deep reinforcement learning drifting control vehicle dynamics control 

摘      要：Drift parking usually requires precise control of a vehicle by a professional driver,which can reflect the performance of the vehicle under critical *** obstacles to implementing this action include the high coupling between the longitudinal and lateral states,the high precision required for the vehicle initial state when the drift is triggered,and the difficulty in determining the reference state variables in the drift process.A two-segment drift parking control system is proposed in this *** the approaching control segment,the vehicle achieves the drift-triggered vehicle speed and pose,which relies on a pathtracking algorithm based on linear time-varying model predictive *** the drifting control segment,the deep reinforcement learning algorithm twin-delayed deep deterministic policy gradient is creatively introduced to the controller *** solves the precise vehicle motion control problem under the condition of the rear wheels having locked *** various simulations,the superiority and robustness to different initial conditions and abrupt changes in the parking space are *** effectiveness of the proposed control system is verified by a ground test.