• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2022, Vol. 58 ›› Issue (6): 130-142.doi: 10.3901/JME.2022.06.130

• 运载工程 • 上一篇    下一篇

扫码分享

考虑横摆稳定性的无人车轨迹跟踪控制优化研究

吴西涛,魏超,翟建坤,苑士华   

  1. 北京理工大学坦克传动国防科技重点实验室 北京 100081
  • 收稿日期:2021-06-09 修回日期:2021-11-01 出版日期:2022-03-20 发布日期:2022-05-19
  • 通讯作者: 魏超,男,1980年出生,博士,教授,博士研究生导师。主要研究方向为无人驾驶车辆技术。E-mail:weichaobit@163.com
  • 作者简介:吴西涛,男,1991年出生,博士研究生。主要研究方向为无人驾驶车辆技术。E-mail:wu.xitao@qq.com
  • 基金资助:
    国家自然科学基金资助项目(51875039)。

Study on the Optimization of Autonomous Vehicle on Path-following Considering Yaw Stability

WU Xitao, WEI Chao, ZHAI Jiankun, YUAN Shihua   

  1. Science and Technology on Vehicle Transmission Laboratory, Beijing Institute of Technology, Beijing 100081
  • Received:2021-06-09 Revised:2021-11-01 Online:2022-03-20 Published:2022-05-19

摘要: 横摆稳定性和轨迹跟踪性能对无人车至关重要。为此,提出一种基于模型预测控制的轨迹跟踪控制器,将考虑瞬时极限性能的稳定性判据添加到控制器约束中,并且利用性能驱动的方式对控制器的参数进行优化。首先根据车辆3自由度动力学模型建立横摆角速度-质心侧偏角相平面,分析前轮转角对相平面平衡点的影响,通过建立相平面的等倾几何曲线,分析车辆的稳定性特征,设计出基于包络线的横摆稳定性判据。然后将模型预测控制器的代价函数参数化,根据性能目标设计特定场景的全局代价作为评价函数,利用贝叶斯优化进行预测时域和代价函数权重两类参数的优化,实现目标任务全局性能最优。仿真和实车试验表明,所提算法在保证车辆稳定的前提下,发挥了车辆的动力学极限,采用的贝叶斯优化方法对轨迹跟踪模型预测控制器的参数进行了优化,实现了轨迹跟踪性能的提高。

关键词: 无人驾驶车辆, 轨迹跟踪, 模型预测控制, 横摆稳定性判据, 贝叶斯优化

Abstract: Vehicle yaw stability and the performance of path-following are crucial to autonomous vehicles. A path-following controller based on model predictive control is designed. The stability criterion considering the instantaneous limit performance is added to the controller constraints, and the controller parameters are optimized in a performance-driven manner. Firstly, we established the phase plane of the yaw rate and the sideslip angle according to the vehicle's three-degree-of-freedom dynamic model, and the influence of the steering angle on the phase plane balance point is analyzed. By establishing the isotropic geometry of the phase plane, the stability characteristics of the vehicle are analyzed and the yaw stability criterion based on the envelope is formed. Secondly, we parameterize the cost function of the model predictive controller and design the global cost function of a specific scenario as the performance evaluation function, we use Bayesian optimization to optimize the prediction horizon and cost function weights to achieve the optimal global performance under target task. Finally, the simulation and real vehicle test show the algorithm exerts the vehicle to the dynamic limits and improves the tracking performance.

Key words: autonomous vehicles, path-following, model predictive control, yaw stability criterion, Bayesian optimization

中图分类号: