交叉口困境区驾驶员行为决策建模与安全防控方法研究

When vehicle approach an intersection at the end of green light, driver may find himself neither too close to the intersection to stop nor yet too far from the intersection to pass through the intersection, or fall in hesitation to decide whether to stop before the stop line or accelerate to pass through the intersection. These situations can lead to read-end crashes, right-angle crashes and so on. Previous works research on this phenomenon mostly based on kinematics theory, ignoring the road-traffic environment, driver behavior characteristics, vehicle condition impact on this phenomenon, and lacking the quantitative description of the influence. This project focuses on driver behavior characteristics in dilemma zones. Firstly, via field data collection, driving simulation experiment and questionnaire investigation, researches on visual and auditory information selective attention calculation model based on cognitive linguistics, as to quantitatively describe road-traffic environment of dilemma zones influence on driver behavior characteristics. Then, based on theory of planned behavior, researches on driver behavior decision-making models, quantitative describing the progressive coupling relationship among road-traffic environment, driver behavior characteristics and driver behavior decision-making. Thirdly, Based on multicomponent and multiphase flow coupling theory, analyses the dilemma zones formation mechanism and dynamic evolutionary process in complex environment, in order to integrate and describe the vehicles movement feature and traffic flow characteristics in dilemma zones. Finally, in the view of the difference between driver behavior characteristics and traffic flow characteristics in dilemma zones, researches on the prevention and control methods in dilemma zones for individual vehicle and traffic flow respectively. The result of this project will providing theoretical basis for traffic signal optimization, dilemma zones safety protection for smart vehicle, microscopic traffic simulation and so forth, and provide references for making the traffic law and regulation, defining the traffic accident liability and so on.

当车辆驶入交叉口遇到绿灯信号转换时，会遇到既不能安全停车又不能安全通行，或犹豫是选择停车还是加速通行的困境状态，容易造成追尾、侧碰等交通事故。国内外学者主要从运动学模式对这一现象进行研究，忽视外部环境、驾驶员自身特性等产生的影响，缺少定量描述影响程度的方法。本项目从交叉口困境区驾驶员行为特性出发，通过现场数据采集、模拟驾驶实验、问卷调查等，研究驾驶员视听觉选择性注意计算模型，定量描述交叉口困境区外部环境对驾驶员行为特性的影响；研究驾驶员行为决策模型，从计划行为理论的角度定量描述外部环境-驾驶员自身特性-行为决策之间的递进作用关系；基于多组分多相流耦合理论，解析交叉口困境区形成机理和动态演化过程，实现微观个体行为和宏观交通流特性的统一表征；针对个体车辆和交通流的差异，分别研究交叉口困境区防控方法。为交通信号优化、智能车辆、微观仿真等提供理论基础，为交通法规制定、交通事故责任界定提供参考依据。

从交通工程上分析：当车辆驶入信号交叉口遇到绿灯信号转换成黄灯信号时，驾驶员会陷入两难或犹豫的抉择困境，从而使交叉口发生车辆追尾、刮蹭、侧碰、斜碰的交通事故率大大增加。因此，只有从驾驶员行为特性出发，结合车辆的运行学特征，解析交叉口困境区的形成机理及其动态演化过程，才能对交叉口安全问题进行有效防控。.本项目主要将认知神经科学与计算机视觉计算相结合，以车辆的挡风板、后视镜、仪表盘等为视觉窗，定量描述交叉口困境区道路环境、交通环境以及车内环境对驾驶员行为特性的影响，解决驾驶员行为特性影响因素及其敏感性难以量化的难题；基于行为计划理论，将交叉口困境区驾驶员驾驶行为表征为外部环境因素、自身特性因素递进作用下的行为决策，并以驾驶员的行为态度、主观规范、知觉行为控制能力三个要素定量描述外部环境影响下的驾驶员自身特性变化及其对驾驶行为决策的影响，解决驾驶员驾驶行为决策外部环境、自身特性对驾驶员行为决策的影响难以量化计算的难题；从交通流理论车辆跟驰、车辆换道的建模思想和建模过程出发，引入交叉口困境区驾驶员行为特性和驾驶员行为决策，实现对驾驶员行为决策-车辆运动特征、车辆微观行为-交通流宏观现象的一体化描述和统一表征，解决基于运动学模式研究交叉口困境区所展示的交通流特性与实际交通流特性不一致，不符合交叉口困境区驾驶员行为特性及车辆运动规律等难题。.项目研究成果可为交通信号控制、交叉口渠化设计、驾驶员行为分析等领域提供理论基础和技术指导，为交通法规的制定、交通事故责任划分等提供参考依据。.本项目取得的研究成果如下：.1）撰写项目研究报告1套；.2）发表学术论文16篇，其中SCI论文6篇，EI论文9篇；.3）出版学术著作1部；.4）申请国家发明专利29项，其中已授权国家发明专利8项，授权实用新型专利5项；.5）获计算机软件著作权3项；.6）培养研究生6名，其中博士研究生3名，硕士研究生3名。