汽车工程 ›› 2022, Vol. 44 ›› Issue (8): 1162-1172.doi: 10.19562/j.chinasae.qcgc.2022.08.006

所属专题: 智能网联汽车技术专题-规划&控制2022年

• • 上一篇    下一篇

基于改进BP神经网络的智能车纵向控制方法

梁旺,秦兆博(),陈亮,边有钢,胡满江   

  1. 湖南大学,汽车车身先进设计制造国家重点实验室,长沙  410082
  • 收稿日期:2021-12-07 修回日期:2022-03-20 出版日期:2022-08-25 发布日期:2022-08-25
  • 通讯作者: 秦兆博 E-mail:qzb@hnu.edu.cn

Longitudinal Control Method of Intelligent Vehicles Based on the Improved BP Neural Network

Wang Liang,Zhaobo Qin(),Liang Chen,Yougang Bian,Manjiang Hu   

  1. Hunan University,State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body,Changsha  410082
  • Received:2021-12-07 Revised:2022-03-20 Online:2022-08-25 Published:2022-08-25
  • Contact: Zhaobo Qin E-mail:qzb@hnu.edu.cn

摘要:

针对传统PI控制在车辆速度跟踪过程中参数固定且不易整定的问题,提出了一种基于改进BP神经网络的智能汽车纵向控制方法。分别构建驱/制动模式下的BP神经网络,针对BP神经网络初始参数选取困难及反向自学习存在梯度消失等问题,利用粒子群算法和批处理归一化方法对BP神经网络进行改进,最终实现PI控制参数的动态自整定。通过Carsim/Simulink联合仿真与实车测试对该方法进行了验证,结果表明:相比于传统PI控制,所提出的纵向控制方法在实现基于误差快速调整参数的同时提高了车辆纵向控制精度。

关键词: 智能汽车, 速度跟踪控制, BP神经网络, 自整定, 粒子群算法

Abstract:

For the problem that the parameters of traditional PI control are fixed and difficult to be adjusted in the process of vehicle speed tracking, a longitudinal control method of intelligent vehicle based on the improved BP neural network is proposed. The BP neural network in drive mode and brake mode is established respectively. In view of the difficulty in selecting initial parameters of BP neural network and the problem of gradient disappearing in reverse self-learning, particle swarm optimization and batch normalization are used to improve the BP neural network. Finally, the dynamic self-tuning of PI parameters is realized. By Carsim/Simulink co-simulation and real vehicle test, the proposed method is verified. The results show that the proposed longitudinal control method can realize rapid adjustment of parameters based on error and improve the longitudinal control accuracy of the vehicle compared with the traditional PI control.

Key words: intelligent vehicle, speed tracking control, BP neural network, self-tuning, particle swarm optimization