Deep reinforcement learning-based drift parking control of automated vehicles  

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

机构地区：[1]School of Automotive Studies,Tongji University

出　　处：《Science China(Technological Sciences)》2023年第4期1152-1165,共14页中国科学（技术科学英文版）

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)。

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 conditions.The 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 paper.In 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 control.In the drifting control segment,the deep reinforcement learning algorithm twin-delayed deep deterministic policy gradient is creatively introduced to the controller design.It solves the precise vehicle motion control problem under the condition of the rear wheels having locked brakes.Through various simulations,the superiority and robustness to different initial conditions and abrupt changes in the parking space are verified.The effectiveness of the proposed control system is verified by a ground test.

关 键 词：automated vehicles deep reinforcement learning drifting control vehicle dynamics control 

分 类 号：U463.6[机械工程—车辆工程] TP18[交通运输工程—载运工具运用工程]