汽车工程 ›› 2023, Vol. 45 ›› Issue (8): 1417-1427.doi: 10.19562/j.chinasae.qcgc.2023.08.012

所属专题: 智能网联汽车技术专题-感知&HMI&测评2023年

• • 上一篇    下一篇

基于CIDAS事故数据的路口乘用车-两轮车测试场景研究

胡林1,李根1,王方1(),林淼2,巫宁3   

  1. 1.长沙理工大学汽车与机械工程学院,长沙 410114
    2.中国汽车技术研究中心有限公司,天津 300300
    3.波鸿鲁尔大学 ;德国
  • 收稿日期:2023-02-01 修回日期:2023-03-05 出版日期:2023-08-25 发布日期:2023-08-17
  • 通讯作者: 王方 E-mail:wangfang83715@163.com
  • 基金资助:
    国家自然科学基金(52172399);湖南省研究生科研创新项目(QL20220197);湖南省教育厅科研基金(21A0193);长沙市自然科学基金(KQ2208235)

Research on Test Scenarios of Passenger Cars and Two-Wheelers at Intersections Based on CIDAS Accident Data

Lin Hu1,Gen Li1,Fang Wang1(),Miao Lin2,Ning Wu3   

  1. 1.School of Automotive and Mechanical Engineering,Changsha University of Science and Technology,Changsha 410114
    2.China Automobile Technology Research Center Co. ,Ltd. ,Tianjin 300300
    3.Ruhr-University, Germany
  • Received:2023-02-01 Revised:2023-03-05 Online:2023-08-25 Published:2023-08-17
  • Contact: Fang Wang E-mail:wangfang83715@163.com

摘要:

现有两轮车事故场景研究由于不区分事发地点,故对路口两轮车事故场景的提取还远远不足。据此,本文对来自CIDAS数据库的1 239起路口乘用车-两轮车事故案例进行了聚类分析,提取10个典型路口两轮车事故场景。针对强相关变量组,对比了单层聚类和双层聚类两种方法,发现双层聚类在样本划分和结果可解释性方面性能更优。从场景频率和致伤风险两个角度分析了聚类结果,进一步挖掘了速度与场景致伤风险指数之间的相关性。通过事故致因分析,发现了一些特色场景,例如未让行、两轮车逆行、交通灯冲突(包括闯红灯和交通灯切换间隙)以及视觉障碍。提取的路口两轮车事故场景能为智能车及主动安全系统测试提供场景参考依据。

关键词: 事故场景, 乘用车-两轮车事故, 聚类分析, 智能车测试, 致伤风险

Abstract:

The existing research on two-wheeler accident scenarios does not distinguish the accident location, so the extraction of two-wheeler accident scenarios at the intersection is far from enough. Therefore, this paper conducts a cluster analysis on 1 239 cases of intersection accidents involving passenger vehicles and two-wheelers from the CIDAS database, and extracts 10 typical two-wheeler accident scenarios at intersections. For the group of strongly correlated variables, the single-layer clustering and double-layer clustering methods are compared, and it is found that the double-layer clustering method has better performance in sample division and interpretation of results. The clustering results are analyzed from the perspective of scenario frequency and injury risk, and the correlation between velocity and scenario injury risk index is further explored. Through the analysis of accident causes, some characteristics scenarios are identified, such as failure to give way, two-wheeler running in the opposite direction, traffic light conflict (including jumping the red light and traffic light switching gap) and visual obstruction. The two-wheeler accident scenarios extracted in this paper can provide a reference basis for the test of intelligent vehicles and active safety systems.

Key words: accident scenarios, passenger car-two-wheeler accident, cluster analysis, intelligent vehicle testing, injury risk