中国安全科学学报 ›› 2023, Vol. 33 ›› Issue (7): 24-31.doi: 10.16265/j.cnki.issn1003-3033.2023.07.1120

• 安全社会科学与安全管理 • 上一篇    下一篇

考虑行车安全事件严重程度和个体差异的驾驶行为风险评估

张晖1,2,3(), 刘永杰1,2, 吴超仲1,2, 丁乃侃1,2, 张琦1,2, 肖逸影4   

  1. 1 武汉理工大学 智能交通系统研究中心,湖北 武汉 430063
    2 国家水运安全工程技术研究中心,湖北 武汉 430063
    3 运输车辆运行安全技术交通运输行业重点实验室,北京 100088
    4 武汉市交通发展战略研究院,湖北 武汉 430063
  • 收稿日期:2023-02-12 修回日期:2023-05-06 出版日期:2023-07-28
  • 作者简介:

    张晖 (1983—),男,安徽铜陵人,博士,研究员,主要从事道路交通安全、驾驶行为及车路协同等方面的研究。E-mail:

    吴超仲,教授

    丁乃侃,副研究员

  • 基金资助:
    国家自然科学基金(52072289); 国家重点研发计划(2019YFB1600803); 运输车辆运行安全技术交通运输行业重点实验室开放课题(KFKT2018-04)

A risk assessment method of driving behavior considering severity of safety-critical events and individual heterogeneity

ZHANG Hui1,2,3(), LIU Yongjie1,2, WU Chaozhong1,2, DING Naikan1,2, ZHANG Qi1,2, XIAO Yiying4   

  1. 1 Intelligent Transportation Systems Research Center, Wuhan University of Technology, Wuhan Hubei 430063, China
    2 National Engineering Research Center for Water Transport Safety, Wuhan University of Technology, Wuhan Hubei 430063, China
    3 Key Laboratory of Operation Safety Technology on Transport Vehicles, Ministry of Transport, Beijing, 1000888, China
    4 Wuhan Transportation Development Strategy Institute, Wuhan Hubei 430063, China
  • Received:2023-02-12 Revised:2023-05-06 Published:2023-07-28

摘要:

为解决驾驶行为风险评估方法存在事件危险程度度量缺失及个体行为习惯考虑不足的问题,采集15位被试的自然驾驶试验数据,通过配对T检验和具有噪声的基于密度的聚类方法(DBSCAN),聚类得到行车安全事件中指标偏离正常状态的显著程度和驾驶人风险倾向性等级;选取指标量化单次行车安全事件严重程度,修正驾驶风险权重,构建考虑行车事件严重程度和个体差异的驾驶行为风险评估方法;开展实例分析,使用车头时距(TH)验证模型的有效性。研究结果表明:速度标准差、速度极差、加速度均值和最大值对于驾驶风险评估的重要度较高;使用优化后的评估方法得到的驾驶行为风险评分范围为[21,42.6],均值为32.93,标准差6.62,相较于传统评分方法,该风险评分与实际情况更为接近;采用上述指标评价综合驾驶行为风险,有助于提升驾驶风险辨识准确性。

关键词: 行车安全事件, 严重程度, 个体差异, 驾驶行为, 风险评估, 具有噪声的基于密度的聚类方法(DBSCAN)

Abstract:

Aiming at the lack of risk degree measurement and insufficient consideration of individual differences in the driving behavior risk assessment method, the natural driving experimental data of 15 subjects were collected, and the paired T-test and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering were used to obtain the deviation of the indicator from the normal state in driving safety events and driver risk propensity level. The indicators were selected to quantify the severity of a single driving safety event, and the driving risk weights were corrected to construct a driving behavior risk assessment method that considered the severity of driving events and individual differences. The validity of the model was verified by using time head (TH). The results show that speed standard deviation, speed range and mean and maximum value of acceleration are more important for driving risk assessment. The risk score obtained by the optimized evaluation methods ranges from [21,42.6], with a mean value of 32.93 and a standard deviation of 6.62. The driving behavior risk score in this study is closer to the actual situation than the traditional score. The above indicators can be used to evaluate the comprehensive driving behavior risk and improve the accuracy of driving risk identification.

Key words: safety-critical events, severity, individual heterogeneity, driving behavior, risk identification, density-based spatial clustering of applications with noise (DBSCAN)