基于轨迹数据的无信号交叉口行人过街行为研究

Decision-Making Behavior of Pedestrians at Unsignalized Intersections Based on Trajectory Data

随着我国城市道路运行效率的降低和交通安全形势的日益严峻,行人交通安全已成为普遍关注的问题。本文从无信号道路交叉口的实测轨迹数据中提取相关参数,建立基于随机森林算法的行人过街决策模型并进行效果检验和验证,模型正确率可达93.1%。同时挖掘出各个解释变量对过街决策的重要度,证明将随机森林算法用于行人过街决策分析具有较好的拟合度与预测精度。本研究可应用于人车交互仿真工具中模拟行人行为,为进一步研究人车混行的无信号道路交叉口仿真和控制系统奠定了基础。

The low operating efficiency and severe situation of urban traffic lead to pedestrians’ safety problems. Based on observed trajectory data at an unsignalized intersection, this work analyzed pedestrians’ crossing behavior. The decision-making behavior of pedestrians was modeled using random forest algorithm, and the model was then calibrated and validated according to empirical data, which accuracy reached 93.1%. This work has found several key factors on the decision-making behavior of pedestrians, which were never reported before, verified that random forest model had a great goodness of fit and prediction. The research can be applied to simulation tools, which laid the foundation of further study on the simulation of unsignalized intersections.