基于大规模运动轨迹的交叉口人车交互行为建模及优化调控研究

The interaction between pedestrian and vehicle frequently induces the conflicting of right of way , which is the most primary cause of the traffic accidents involving vehicles and pedestrians at intersections. This project investigates the pedestrian-vehicle interaction behavior. With the understanding of characteristics of pedestrian-vehicle interaction behaviors such as the uncertainty and time-variation, the large-scale trajectory data of pedestrians and vehicles are mined to analyze, model, simulate and regulate the pedestrian-vehicle interaction behaviors. Specifically, the rule of pedestrian-vehicle interaction behavior is investigated through feature expression, pattern recognition, spectrum construction; On the basis of identifying the decision points and dilemma zones, the models for characterizing the decision-making behaviors for pedestrian crossing the street and vehicle yielding to pedestrian are developed to reveal the underlying decision-making mechanisms. Meanwhile, the real-time pedestrian and vehicle trajectory prediction models are integrated into the microscopic pedestrian vehicle simulation model to explore the relations between the pedestrian and vehicle. Moreover, the pedestrian and vehicle path planning model is developed to optimize and regulate the behaviors of pedestrians and vehicles in an interaction. Simulation analysis and empirical study are applied to verify the validity of this method. The research outcomes provide strong theoretical foundation and technical support to address the problem of conflicting of right of way for vehicles and pedestrians in the intersection, which will be very helpful to reduce the traffic accidents and improve the traffic safety at intersections.

我国行人与机动车的交互干扰常形成“车不让人，人不让车”现象，是交叉口人车交通事故的最主要原因。本项目以人车交互行为为研究对象，围绕行为的不确定性与时变性等特点，通过深度挖掘大规模行人与机动车运动轨迹数据，进行行为特征解析、决策建模、微观仿真及优化调控的多层次分析与建模。具体为：按照行为特征表达-模式判别-谱系构建，层层解析人车交互行为特征；根据运动轨迹进行决策点辨识及两难区间判定，建立行人过街行为决策模型和机动车避让行人行为决策模型，揭示人车交互行为决策机理；构建行人和机动车运动轨迹实时预测模型并耦合成交叉口人车微观仿真模型，探寻人车之间的相互作用关系；提出基于人车运动路径协同规划模型的人车交互行为优化调控方法，结合仿真分析和实证研究进行方法有效性验证。研究成果为有效解决 “车不让人，人不让车”问题提供有力的理论依据与技术支撑，对减少交叉口人车交通事故、提升交通安全性具有关键作用。

我国行人与机动车的交互干扰常形成“车不让人，人不让车”现象，是交叉口人车交通事故的最主要原因。本项目以人车交互行为为研究对象，围绕行为的不确定性与时变性等特点，通过深度挖掘大规模行人与机动车运动轨迹数据，进行行为特征解析、决策建模、微观仿真及优化调控的多层次分析与建模。具体为：按照行为特征表达-模式判别-谱系构建，层层解析人车交互行为特征；根据运动轨迹进行决策点辨识及两难区间判定，建立行人过街行为决策模型和机动车避让行人行为决策模型，揭示人车交互行为决策机理；构建行人和机动车运动轨迹实时预测模型并耦合成交叉口人车微观仿真模型，探寻人车之间的相互作用关系；提出基于人车运动路径协同规划模型的人车交互行为优化调控方法，结合仿真分析和实证研究进行方法有效性验证。研究成果为有效解决 “车不让人，人不让车”问题提供有力的理论依据与技术支撑，对减少交叉口人车交通事故、提升交通安全性具有关键作用。.在项目执行期间，项目负责人及其团队共发表标注基金号的学术论文10篇，其中多篇论文发表在交通领域顶级期刊Transportation Research Part B、Transportation Research Part C、Safety Science等，9篇论文被SCI/SSCI检索，申请国家发明专利9项，授权5项，项目负责人获中国智能交通协会科技进步二等奖（排名第2）。参加国内外学术会议9次，培养硕士生2名，项目负责人依托本项目研究成果入选浙东青年学者、全国香江学者，项目负责人及3名团队成员全部由讲师晋升为副教授。