跟驰自动驾驶车时人驾车行为研究:实证与建模  

Analyzing human driving behavior when following autonomous vehicles:real vehicle testing and modeling

作　　者：刘怿轩 张慧永 王猛 吴欢 宗芳[4] LIU Yi-xuan;ZHANG Hui-yong;WANG Meng;WU Huan;ZONG Fang(CATARC Automotive Test Center(Wuhan)Co.,Ltd.,Wuhan 430000,China;Technical Development Department,FAW-Volkswagen Automotive Co.,Ltd.,Changchun 130011,China;National Science Library,Chinese Academy of Sciences,Beijing 100190,China;Transportation College,Jilin University,Changchun 130022,China)

机构地区：[1]中汽研汽车检验中心(武汉)有限公司,武汉430000 [2]一汽-大众汽车有限公司,技术开发部,长春130011 [3]中国科学院,文献情报中心,北京100190 [4]吉林大学,交通学院,长春130022

出　　处：《交通运输工程与信息学报》2023年第2期14-28,共15页Journal of Transportation Engineering and Information

基　　金：国家自然科学基金项目(61873109)。

摘　　要：随着中国新基建战略的提出及自动驾驶和网联通信技术的不断发展,智能网联车辆(Connected and Automated Vehicle,CAV)、自动驾驶车辆(Autonomous Vehicle,AV)和人工驾驶车辆(Human-driven Vehicle,HDV)混行的状态将在未来一段时间内存在。在混行条件下,车辆间的交互影响模式将发生变化。本文以HDV跟驰AV的驾驶行为为研究对象,通过分析驾驶实验数据将跟驰AV时HDV的驾驶风格量化并分为迟疑型、平稳型和信赖型三类。同时考虑驾驶风格、车辆的转弯能力和转弯半径等参数改进智能驾驶人模型(Intelligent Driver Model,IDM),建立了前车为AV时的HDV跟驰模型。该模型通过对三类不同风格HDV跟驰AV时的驾驶参数的标定,能根据不同跟驰风格采取相应的跟驰策略。经数据拟合检验,该模型在启动加速、匀速行驶和制动减速阶段均能以较高精度拟合实际驾驶数据,其中直行跟驰的平均拟合精度为96.2%,转弯跟驰的平均拟合精度为91.4%。可见,本文提出的模型可以刻画HDV跟驰AV时的行为特征。在目前难以进行大规模混流实车实验的情况下,可用于混流条件下的跟驰行为仿真,也可为未来AV及HDV混行交通流的道路交通管理及基础设施设计等提供理论依据或模型基础。The proposal for new infrastructure construction strategies,combined with the develop-ment of autonomous driving and connected network communication technologies,will contribute to a traffic flow mixed with connected and automated vehicles,autonomous vehicles(AV),and human driven vehicles(HDV).The interaction patterns among vehicles under mixed traffic conditions will differ compared to the driving scenario of a single vehicle type.Therefore,this study investigates the driving behavior of HDVs following AVs.The data collected from actual vehicle tests are analyzed to quantify the car-following behavior of HDVs following AVs,and the driving styles are divided in-to three categories,i.e.,hesitant,steady,and trusting.By considering driving styles,turning ability,and turning radius,we improve the intelligent driver model to construct the driving model of HDVs following AVs.By calibrating the parameters of three types of drivers with different driving styles re-spectively,the model can be adopted to express the driving strategies of different driving styles.The fitting test results show that the model can fit the actual driving data with high precision during start-ing acceleration,uniform driving speed,and braking deceleration,with an average fitting accuracy of 96.2%and 91.4%in the straight-and turning-following,respectively.The proposed model can de-scribe the HDV-AV following characteristics and simulate the car-following behavior when large scale actual vehicle experiments for mixed traffic are difficult.In addition,it can assist road traffic management and infrastructure layout.

关 键 词：交通工程 跟驰模型 混行交通流 驾驶风格 数值仿真 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]