抵近信控交叉口分心驾驶识别模型  

Recognition Model of Distracted Drivers When Approaching Signalized Intersections

作　　者：张玉婷 陈波佑 张双焱 闫学东[2] 李晓梦 ZHANG Yu-ting;CHEN Bo-you;ZHANG Shuang-yan;YAN Xue-dong;LI Xiao-meng(College of Transportation Engineering,Chang'an Universtiy,Xi'an 710021,China;MOT Key Laboratory of Transportation Industry of Big Data Application Technologies for Comprehensive Transport,Beijing Jiaotong University,Beijing 100044,China;Centre for Accident Research and Road Safety-Queensland(CARRS-Q),Queensland University of Technology,QLD 4059,Australia)

机构地区：[1]长安大学,运输工程学院,西安710021 [2]北京交通大学,综合交通运输大数据应用技术交通运输行业重点实验室,北京100044 [3]昆士兰科技大学,交通事故与道路安全研究中心,昆士兰4059,澳大利亚

出　　处：《交通运输系统工程与信息》2022年第1期217-224,共8页Journal of Transportation Systems Engineering and Information Technology

基　　金：国家自然科学基金(71901036);国家重点研发计划(2019YFE0108000);国家级大学生创新创业训练计划(S202010710019)。

摘　　要：为降低驾驶人抵近交叉口过程因分心导致的交通事故,本文基于双向长短时记忆网络(BILSTM)建立分心驾驶识别模型。依托驾驶模拟实验,采集了45位驾驶人抵近信控交叉口过程中的横纵行为数据,通过方差分析研究分心任务对驾驶行为的影响。结果表明,分心驾驶人需要更长的制动反应时间,制动操作时间缩短,踩压制动踏板的力度下降,同时操纵方向盘的稳定性变差。然后,筛选有显著性影响的6个特征行为指标作为模型的输入,结果表明:BILSTM模型分心状态识别的精确率最高,达到92.6%,F_(1)值为88.7%;准确率、精确率、召回率、F_(1)、AUC和ROC曲线等模型性能均优于单向长短期记忆网络、支持向量机和决策树5.0分心识别模型。研究结果说明BILSTM模型能有效判别抵近信控交叉口驾驶人分心状态,可为交叉口驾驶人分心预警系统的优化设计提供依据和指导。To reduce the traffic accidents caused by distraction when approaching signalized intersections,this study developed a driver distraction detection model based on bi-directional long short-term memory(BILSTM)method.A series of experiment scenarios approaching intersections were performed using a high-driving simulator.The longitudinal and horizontal behaviors of 45 drivers were analyzed to determine the impacts of driver distractions on driving behaviors using the analysis of variance(ANOVA).The results indicate that the distracted driver needs longer braking response time,operates the braking in shorter time and exerts less pressure on the pedal.Meanwhile,the districted driver shows deteriorated stability when operating the steering wheel.Six indicators that have significant impacts on driving behavior were selected as the input to the distraction recognition model.The results show that the recognition accuracy of the model is 92.6% and the F_(1)-scores is 88.7%.Compared with the long short-term memory(LSTM),support vector machines(SVM)and decision tree 5.0 distraction recognition models,the proposed model shows the best performance in terms of accuracy,recall,the F_(1)-scores,the area under the curve(AUC)and the receiver operating characteristic(ROC)curve.The BILSTM model can effectively distinguish the distraction driving status when the driver approaching signalized intersections,which provides basis and guidance for the optimal design of the driver distraction warning assistance system at the intersection.

关 键 词：交通工程 信控交叉口 分心状态识别 BILSTM模型 驾驶模拟 

分 类 号：U491[交通运输工程—交通运输规划与管理]