汽车工程 ›› 2024, Vol. 46 ›› Issue (10): 1804-1815.doi: 10.19562/j.chinasae.qcgc.2024.10.008

• • 上一篇    下一篇

基于学习型模型预测控制的智能车辆路径跟踪控制

秦洪懋1,2,江曙1,张田田1,谢和平1,3,边有钢1,2(),李洋1   

  1. 1.湖南大学机械与运载工程学院,整车先进设计制造技术全国重点实验室,长沙 410082
    2.湖南大学无锡智能控制研究院,无锡 214115
    3.徐州徐工矿业机械有限公司,徐州 210009
  • 收稿日期:2024-05-11 修回日期:2024-06-12 出版日期:2024-10-25 发布日期:2024-10-21
  • 通讯作者: 边有钢 E-mail:byg10@foxmail.com
  • 基金资助:
    国家重点研发计划项目(2023YFB2504500);国家自然科学基金(52372411);湖南省自然科学基金(2023JJ10008)

Path Tracking Control of Intelligent Vehicle Based on Learning Model Predictive Control

Hongmao Qin1,2,Shu Jiang1,Tiantian Zhang1,Heping Xie1,3,Yougang Bian1,2(),Yang Li1   

  1. 1.College of Mechanical and Vehicle Engineering,Hunan University,State Key Laboratory of Advanced Design and Manufacturing Technology for Vehicle,Changsha 410082
    2.Wuxi Intelligent Control Research Institute of Hunan University,Wuxi 214115
    3.Xuzhou XCMG Mining Machinery Co. ,Ltd. ,Xuzhou 210009
  • Received:2024-05-11 Revised:2024-06-12 Online:2024-10-25 Published:2024-10-21
  • Contact: Yougang Bian E-mail:byg10@foxmail.com

摘要:

路径跟踪控制是智能车辆的一项关键技术。然而,现有车辆跟踪控制方法多依赖于较为精确的车辆控制模型,而实际的车辆控制系统多存在建模误差、参数摄动以及外部扰动等,显著影响路径跟踪控制精度。本文针对性地提出一种考虑车辆未建模动态的智能车辆学习型路径跟踪控制方法。首先建立车辆标称模型,并采用线性预言模型对车辆未建模动态进行近似补偿,以提高车辆模型的精度;然后基于扩展卡尔曼滤波原理实现对未建模动态参数的学习更新;之后构建考虑系统未建模动态的学习型模型预测控制器(LMPC);最后通过CarSim和Matlab/Simulink设计多工况多组别联合仿真试验,验证所提方法在提高路径跟踪精度方面的有效性。

关键词: 智能车辆, 路径跟踪控制, 未建模动态, 参数学习, 学习型模型预测控制

Abstract:

Path tracking control is a key technology for intelligent vehicles. However, the existing vehicle tracking control methods mostly rely on more accurate vehicle control models, while actual vehicle control systems mostly have modeling errors, parameter perturbations and external disturbances, which significantly affect path tracking control accuracy. In this paper, a learning path tracking control method for intelligent vehicles considering unmodeled dynamics of vehicles is proposed. Firstly, a nominal model of the vehicle is established and a linear prediction model is used to approximate the compensation for the unmodeled dynamics of the vehicle to improve the accuracy of the vehicle model. Then, learning and updating of the parameters of the unmodeled dynamics are realized based on the principle of Extended Kalman Filtering. Next, learning Model Predictive Controller (LMPC) considering the unmodeled dynamics of the system is established. Finally, the effectiveness of the proposed method in improving the path tracking accuracy is verified by designing a joint simulation test with Carsim and Matlab/Simulink for multiple operating conditions and multiple groups.

Key words: intelligent vehicle, path tracking control, unmodeled dynamics, parameter learning, learning model predictive control