汽车工程 ›› 2021, Vol. 43 ›› Issue (3): 323-329.doi: 10.19562/j.chinasae.qcgc.2021.03.003

• • 上一篇    下一篇

基于自适应神经模糊推理系统的平行泊车路径规划

张家旭1,2,王晨1,郭崇1(),滕飞1,李东燃1   

  1. 1.吉林大学,汽车仿真与控制国家重点实验室,长春 130011
    2.中国第一汽车集团有限公司智能网联研发院,长春 130011
  • 收稿日期:2020-08-21 修回日期:2020-11-10 出版日期:2021-03-25 发布日期:2021-03-26
  • 通讯作者: 郭崇 E-mail:guochong@jlu.edu.cn
  • 基金资助:
    国家自然科学基金面上项目(51775235)

column:Vehicle Parallel Parking Path Planning Based on Adaptive Neuro⁃fuzzy Inference System

Jiaxu Zhang1,2,Chen Wang1,Chong Guo1(),Fei Teng1,Dongran Li1   

  1. 1.Jilin University,State Key Laboratory of Automotive Simulation and Control,Changchun 130011
    2.Intelligent Network R&D Institute,China FAW Group Co. ,Ltd. ,Changchun 130011
  • Received:2020-08-21 Revised:2020-11-10 Online:2021-03-25 Published:2021-03-26
  • Contact: Chong Guo E-mail:guochong@jlu.edu.cn

摘要:

针对常见的平行泊车场景,提出一种基于自适应神经模糊推理系统的平行泊车路径规划方法。以基于优化算法的泊车路径规划方法得出的泊车路径作为训练样本,利用Python脚本语言建立以自适应遗传算法和拟牛顿法为内核的自动化训练框架,使自动训练后的自适应神经模糊推理系统既可继承基于优化算法的泊车路径规划方法适用范围更广的优势,又有效解决该方法求解过程计算量大的问题。通过仿真分析验证所提出方法的可行性和有效性,结果表明:自动训练后的自适应神经模糊推理系统可依据汽车初始泊车位姿和泊车位信息快速规划出可行的平行泊车路径。

关键词: 平行泊车路径规划, 自适应神经模糊推理系统, 自适应遗传算法, 拟牛顿法, 自动化训练框架

Abstract:

Aiming at the common parallel parking scene, a novel parallel parking path planning method for vehicle is proposed based on adaptive neuro?fuzzy inference system. With the parking path obtained from parallel parking path planning based on optimization algorithm as training sample, Python script is used to build the automated training framework with adaptive genetic algorithm and quasi?Newton algorithm as its core for enabling the adaptive neuro?fuzzy inference system automatically trained can not only inherit the merits of wider application scope of parallel parking path planning method based on optimization algorithm, but also effectively get rid of enormous computation efforts. The feasibility and effectiveness of the proposed method are verified by simulation, and the results show that the adaptive neuro?fuzzy inference system automatically trained can quickly plan the feasible parallel parking path based on the initial parking posture of vehicle and parking space information.

Key words: parallel parking path planning, adaptive neuro?fuzzy inference system, adaptive genetic algorithm, quasi?Newton algorithm, automated training framework