汽车工程 ›› 2021, Vol. 43 ›› Issue (11): 1611-1619.doi: 10.19562/j.chinasae.qcgc.2021.11.006

• • 上一篇    下一篇

转向避撞工况下装载激光雷达车辆的障碍物跟踪

赵治国(),王鹏,陈晓蓉,梁凯冲   

  1. 同济大学汽车学院,上海 201804
  • 收稿日期:2021-05-07 修回日期:2021-08-13 出版日期:2021-11-25 发布日期:2021-11-22
  • 通讯作者: 赵治国 E-mail:zhiguozhao@tongji.edu.cn

Obstacle Tracking of a Lidar-equipped Vehicle in Turning for Collision Avoidance

Zhiguo Zhao(),Peng Wang,Xiaorong Chen,Kaichong Liang   

  1. School of Automotive Studies,Tongji University,Shanghai 201804
  • Received:2021-05-07 Revised:2021-08-13 Online:2021-11-25 Published:2021-11-22
  • Contact: Zhiguo Zhao E-mail:zhiguozhao@tongji.edu.cn

摘要:

目前车载激光雷达对周围障碍物的感知与定位多基于车辆坐标系,当车辆转向避撞时,由于航向角的变化和车辆坐标系的旋转,会增加障碍物数据关联、周边车辆运动状态分析和避撞路径规划的难度。为消除车辆转向带来的不利影响,本文中基于车载激光雷达坐标系的变换,提出了一种目标障碍物的跟踪方法。首先,采用改进的K-means算法对提取的道路边界点进行聚类,拟合出道路边界线;其次,根据道路边界利用麻雀搜索算法求解车辆航向角,进而对雷达坐标系进行变换;最后,采用关联算法和粒子滤波器,实现目标障碍物的跟踪。实车试验结果表明:在车辆转向避撞时,提出的算法能够准确地提取道路边界和跟踪障碍物。

关键词: 激光雷达, 道路边界提取, K-means聚类, 麻雀搜索算法, 障碍物跟踪

Abstract:

At present, the perception and positioning of surrounding obstacles by onboard lidarare mostly based on vehicle coordinate system. When the vehicle turns to avoid collision, due to the change of heading angle and the rotation of vehicle coordinate system, the difficulties of the obstacle data correlation, the analysis of surrounding vehicle motion state and the path planning for collision avoidance will increase. In order to eliminate the adverse effects of vehicle steering, atarget obstacle tracking methodis proposed based on the transformation of onboardlidar coordinate system. Firstly, the improved K-means clustering algorithm is used to cluster the road boundary points extracted to fit the road boundary line.Then, sparrow search algorithmisadopted to solve out the heading angle of vehicle accordingtoroad boundary, with the lidar coordinate system transformed. Finally, the association algorithm and particle filter are used to track the target obstacles. The results of real vehicle test show that the algorithm proposed can accurately extract the road boundary and track obstacles in vehicle turning for collision avoidance.

Key words: lidar, road boundary extraction, K-means clustering, sparrow search algorithm, target tracking