欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

JASE ›› 2018, Vol. 9 ›› Issue (2): 141-148.DOI: 10.3969/j.issn.1674-8484.2018.02.003

• 汽车安全 • 上一篇    下一篇

自动驾驶汽车控制系统参数辨识与学习（英文）

王乐一1，殷 刚2，赵广亮3，李升波4，徐 彪4，李克强4   

  1. （1. 韦恩州立大学 电气与计算机工程系， 底特律市 MI 48202，美国；2. 韦恩州立大学 数学系，底特律市 MI 48202，美国；3. 通用电气公司 全球科研中心，尼斯卡于纳市 MI 48202，美国；4. 清华大学 汽车工程系，北京 100084，中国）
  • 收稿日期:2017-11-17 出版日期:2018-06-30 发布日期:2018-07-04
  • 作者简介:第一作者 / First author ： 王乐一 / WANG Leyi (1955—)，男/ male ( 汉族)，上海，教授。E-mail: lywang@wayne.edu。
  • 基金资助:

    The National Natural Science Fund Project / 国家自然科学基金资助项目（ 51575293、51622504）；National Key R&D Program of China（2016YFB0100906）；International Sci &Tech Cooperation Program of China ( 2016YFE0102200)。

Identification and Learning in Autonomous Vehicle Control Systems

WANG Leyi 1, George Yin2, ZHAO Guangliang3, LI Shengbo4, Xu Biao4, LI Keqiang4   

  1.  (1. Department of Electrical and Computer Engineering, Wayne State University, Detroit, MI 48202, USA; 2. Department of Mathematics, Wayne State University, Detroit, MI 48202, USA; 3. GE Global Research Niskayuna, NY 12309, USA; 4. Department of Automotive Engineering Tsinghua University, Beijing 100084, China)
  • Received:2017-11-17 Online:2018-06-30 Published:2018-07-04

PDF

293

摘要：

       针对自动驾驶汽车在行驶过程中会遇到随时间和交通环境变化的不确定性，须对自动驾驶系统参数进行辨识和学习。通过获取闭环状态下的系统行为和学习数据驱动下的系统参数，可极大提升系统辨识可靠性和稳定性。本文对自动驾驶车辆跟车的典型场景，综合考虑车辆控制和动力学特性，构建了基于车辆运行数据和置信椭圆的系统参数辨识的学习算法，提升系统辨识的可靠性和鲁棒性。结果表明提出的参数辨识和学习方法可准确估计车辆参数。

关键词: 汽车控制, 自动驾驶车辆, 参数辩识, 学习, 鲁棒性

Abstract:

System parameters of autonomous vehicles need to be identified and learned during operation to solve the problem that autonomous vehicles encounter many uncertainties that change with time, operating conditions, and environments. By capturing system behavior in a closed-loop setting and using data to learn the
related parameters, system reliability and robustness can be quantitatively established. This paper focuses on a basic scenario of an autonomous vehicle following its front vehicle. By integrating control actions with vehicle dynamics, a learning algorithm using operational data and confidence ellipsoids was employed to support robustness and reliability. A simulation case study was used to illustrate the strategies. The results show the proposed method can estimate the vehicle’s parameters accurately.

Key words: vehicle control , autonomous vehicle , identification of parameters, l earning, robustness