摘要
为更加精细化地分析驾驶人驾驶行为的风险特性,识别驾驶行为风险,提出基于驾驶行为模式转移的个体驾驶行为风险评估方法。通过驾驶模拟试验获取个体驾驶人细粒度驾驶行为数据,依据提取的驾驶行为参数值对驾驶行为模式进行分解,以驾驶模式间的相互转移概率为特征指标,描述驾驶行为模式转移特性;利用最大信息系数及随机森林算法优选特征指标,最终形成个体驾驶人驾驶行为风险评估模型,对驾驶人的驾驶行为风险等级进行辨识。结果表明,该方法可以很好地识别个体驾驶人在驾驶过程中的决策偏好,且驾驶行为风险辨识精度达到85.71%。
In order to refine the analysis of driving behavior risk characteristics of drivers,identify the risks of driving behavior,the paper proposes a risk assessment method of individual driving behavior based on driving behavior pattern shift.Fine-grained driving behavior data of individual drivers is obtained through driving simulation experiment,and the driving behavior pattern is decomposed according to the extracted driving behavior parameter value.The probability of mutual shift between driving modes is taken as the characteristic index to describe driving behavior pattern shift characteristics.Maximum information coefficient and random forest algorithm are used to optimize the characteristic index,and the risk assessment model of individual driver’s driving behavior is ultimately formed,which identifies the risk level of driver’s driving behavior.The results show that this method can identify the decision preferences of individual drivers during driving,and the identification accuracy of driving behavior risk reaches 85.71%.
作者
孙宫昊
荣建
常鑫
刘思杨
高亚聪
Sun Gonghao;Rong Jian;Chang Xin;Liu Siyang;Gao Yacong(Beijing University of Technology,Beijing 100124;Civil Aviation University of China,Tianjin 300300)
出处
《汽车技术》
CSCD
北大核心
2021年第11期22-29,共8页
Automobile Technology
基金
国家自然科学基金项目(5160080899)。
关键词
驾驶行为
驾驶风险
模式转移概率
随机森林
Driving behavior
Driving risk
Pattern transition probability
Random forest
