欢迎访问《兵工学报》官方网站,今天是 分享到:

兵工学报 ›› 2019, Vol. 40 ›› Issue (11): 2336-2351.doi: 10.3969/j.issn.1000-1093.2019.11.019

• 论文 • 上一篇    下一篇

基于Frenet坐标系和控制延时补偿的智能车辆路径跟踪

王威, 陈慧岩, 马建昊, 刘凯, 龚建伟   

  1. (北京理工大学 机械与车辆学院, 北京 100081)
  • 收稿日期:2019-01-15 修回日期:2019-01-15 上线日期:2019-12-31
  • 通讯作者: 陈慧岩(1961—),男,教授,博士生导师 E-mail:chen_h_y@263.net
  • 作者简介:王威(1993—),男,博士研究生。E-mail: wei.wang.bit@outlook.com
  • 基金资助:
    国家部委预先研究项目(LZY20193005)

Path Tracking for Intelligent Vehicles Based on Frenet Coordinates and Delayed Control

WANG Wei, CHEN Huiyan, MA Jianhao, LIU Kai, GONG Jianwei   

  1. (School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China)
  • Received:2019-01-15 Revised:2019-01-15 Online:2019-12-31

摘要: 对智能车辆路径跟踪问题中存在的控制延时问题进行研究。前轮转角表述为纯滞后和1阶惯性延时的串联结构模型,通过使用Matlab/Simulink建立转向控制延时模型,并对实车采集的转向控制数据进行分析,完成延时模型的参数辨识;基于V-REP和ROS搭建仿真测试平台,根据延时模型的辨识结果模拟转向响应特性,实现与实车转向特性等效的控制延时效果;基于Frenet坐标系和运动学、动力学模型构建不考虑控制延时和考虑控制延时的模型预测控制(MPC)路径跟踪控制器,使得控制器可以直接扩展到多车编队行驶场景;在V-REP仿真环境中设置以5 m/s、10 m/s、20 m/s车 速采集的变曲率参考路径,先针对无延时系统考察不考虑控制延时的MPC路径跟踪控制器,获得了平均跟踪误差低于0.22 m的控制效果,验证了不考虑控制延时的MPC控制器在处理无延时车辆系统路径跟踪问题的跟踪性能,再针对大延时车辆系统对比测试两种MPC控制器。试验结果表明:考虑控制延时的MPC控制器相比不考虑控制延时的MPC控制器取得了较大的效果提升,特别是在最大跟踪误差和航向误差指标上表现优异,平均跟踪误差降低了83.7%,最大跟踪误差降低了74.4%;对于高延时系统,低速工况下考虑延时的运动学MPC表现更好,而高速工况动力学MPC表现出了更加稳定的跟踪性能,20 m/s延时试验中仅考虑控制延时的动力学MPC控制器安全地跑完了全程。

关键词: 智能车辆, 路径跟踪, 延时建模, 模型预测控制

Abstract: The path tracking problem for intelligent vehicle with delayed control inputs is studied. The cramping angle is expressed as a series structure model with pure lag and first-order inertial delay, and a steering control delay model is established using Matlab/Simulink. The collected steering control data of an actual vehicle is analyzed for parameter identification of the proposed delay model.The equivalent delay performance in simulation environment based on V-REP and ROS is implemented. The model predictive control (MPC)-based path tracking controllers without or with considering delay control are designed based on Frenet coordinates, and the kinematic and dynamics models, which can also be used for marching vehicle formation. A curvature-variant reference paths collected at 5, 10 and 20 m/s are set in V-REP simulation environment. Three curvature-variant reference paths are presented. For the MPC path tracking controller without delay modeling, the average tracking error is less than 0.22 m for a vehicle platform without control delay. The MPC controllers with and without delay modeling are tested to compare their tracking performances for the vehicle system with long control delay. Simulated results indicate that the average and maximum tracking errors of MPC controller with delay modeling are 83.7% and 74.4% less than those of MPC controller without delay modeling when they are used on a vehicle with delayed control inputs. The kinematics-based MPC controller performs better at low speed, whereas the dynamics-based MPC controller performs better at high speed. Only dynamics-based MPC controller with delay modeling completed the whole test safely at 20 m/s on the vehicle with delayed control. Key

Key words: intelligentvehicle, pathtracking, delaymodeling, modelpredictivecontrol

中图分类号: