轨迹数据驱动的单点信号交叉口运行效率评价方法

Single-point Signalized Intersection Operation Efficiency Evaluation Method Based on Trajectory Data

为解决信号交叉口运行效率评价方法偏理论化、实用性不强等问题,以出租车、公交车、驾图(车联网)多源GPS轨迹数据为基础,充分利用车辆减速、停车、加速等连续速度变化特征及位置信息,提出交叉口个体车辆排队长度、通行时间、停车次数等交通参数提取技术。基于此,构建以信号交叉口运行指数为一级指标,车辆平均通行时间,第95%分位排队长度,两次停车率为第二级指标的评价指标体系,基于高斯混合聚类模型对综合运行指数进行分级量化,最终确定五级服务水平评价标准。实例表明:采用本文方法能够比较准确地反映不同时段交叉口各转向、进口道及交叉口整体运行水平,最大排队长度与实际相对误差约15%左右,优于仿真结果;除样本不足情形以外,服务水平评价结果比仿真更贴近实际运行情况,验证了本文方法的技术可行性及评价结果的客观真实性。

The existing method for signalized intersection operation efficiency evaluation is mostly theoretical and not practical.This paper proposes a method to evaluate vehicles'performances at signalized intersections including queue length,travel time,number of stops,etc.The study collected multi-source GPS trajectory data of taxi,bus and driving map(car network).and obtained continuous speed change characteristics and position information of vehicle deceleration,stop and acceleration.The performance evaluation takes the operation status of the signalized intersection,as the first-level indicator and the average vehicle travel time,the 95 percentile queue length,and the frequency of cycle failure as the second-level indicators.Based on the Gaussian hybrid clustering model,the study graded and quantified the comprehensive performance measures and determined a five-level of service evaluation criteria.The example analysis shows that the method is able to accurately reflect the performances of the turning movements,approaches,and overall operation level of the intersection during different time periods.The relative error of the maximum queue length calculated from the proposed method and the queue length in the field is about 15%,which is smaller than what obtained from the simulation.Except for some cases of insufficient samples,the level of service evaluation results is closer to the actual operation than the simulation.The example analysis verifies the technical feasibility of the proposed method and the accuracy authenticity of the evaluation results.