车路协同环境下驾驶员行为识别方法研究  

A Study on Driver Behavior Identification Method Under Environment of Vehicle-Road Integration

作　　者：马雷[1] 陈珂 王明露 曲瑞 Ma Lei;Chen Ke;Wang Minglu;Qu Rui(College of Vehicle and Energy,Yanshan University,Qinhuangdao 06600)

机构地区：[1]燕山大学车辆与能源学院,秦皇岛066004

出　　处：《汽车工程》2018年第11期1330-1338,共9页Automotive Engineering

基　　金：国家自然科学基金(51275442)资助

摘　　要：本文中提出了一种介于车辆操纵稳定性和智能交通系统的驾驶员行为识别方法。首先通过微观仿真软件实现不同行驶状态下局部路网仿真,获取大量基本仿真数据,根据汽车动力学理论,实现基本行驶参数到行驶状态参数的转化;然后应用邻域粗糙集来进行特征约简,再使用总体平均经验模态分解(EEMD)、相关系数法和样本熵相结合的方法进行样本数据挖掘,将得到的样本熵数值作为聚类的特征向量;最后将特征向量输入GG模糊聚类进行聚类,利用微观交通软件和UC-Win/Road驾驶模拟器仿真得到的样本,采用最小平均贴近度择近原则实现驾驶行为识别验证,并根据最大贴近度和次最大贴近度计算待测样本属于某类驾驶行为的隶属度。实验表明,该方法取得了良好的识别效果。A driver behavior identification method in comprehensive consideration of vehicle handling stability and intelligent transportation system is proposed in this paper. Firstly, the local road network simulation under different driving conditions is achieved by Microscopic traffic simulation software, and massive basic simulation data is obtained. The transformation from basic driving parameters to running status parameters is realized based on the theory of vehicle dynamics. Secondly, neighborhood rough set is applied for thature reduction. Sample data mining is realized by combined use of ensemble empirical mode deeomposition （EEMD） , correlation coefficient and sample entropy. And the obtained sample entropy value is used as the eigenveetors. Finally, the eigenveetors are put into GG fuzzy clustering for clustering. Then on the basis of the samples, which are obtained by Microscopic traffic software and UC-Win/Road driving simulator, different driver behavior identification is achieved by minimum average closeness degree. Based on the maximum closeness degree and the seeondary maximum closeness degree, the driving behavior membership degree of the test sample is calculated. The experiment demonstrates that the method has achieved good effect in driver behavior identification.

关 键 词：驾驶员行为 属性约简 数据挖掘 模糊聚类 

分 类 号：U467[机械工程—车辆工程]