基于轨迹数据的信号交叉口排队长度估计

排队长度是信号交叉口最重要的性能指标之一，也是信号交叉口配时优化的关键参数。对于一些偏僻的路口或者固定检测器损坏的交叉口，由于无法获取准确的排队长度信息而无法了解交叉口的实时状态。针对以上情况，提出通过浮动车轨迹数据来估计信号交叉口的排队长度，通过对轨迹数据的分析，建立了基于浮动车集群队列的排队长度估计模型，利用排队长度估计模型可以实现交叉口排队长度的估计。通过SUMO仿真获取早高峰、晚高峰、平峰流量下输入下的浮动车轨迹数据，分别采用早高峰、晚高峰、平峰流量下时间间隔为2、4、6、8、10 s和渗透率为5%、10%、15%的数据集对模型进行验证。验证结果表明，所建立的模型可以较为准确地估计交叉口的排队长度。与相关方法的对比结果表明，在早高峰流量下和晚高峰流量下，所建立的模型误差更低，不同的时间间隔和渗透率下比不考虑集群车队的精准度平均提高约12%,即使在平峰流量、时间间隔大、低渗透率场景下，所建立的模型估计的排队长度误差仍在可接受范围内。相关的研究成果将为交叉口的交通状态评估以及信号配时优化提供支撑。

Study on queue length of signal intersection based on trajectory data

Queue length is one of the most important performance indicators of signalized intersections, and it is also a key parameter for timing optimization of signalized intersections. For some remote intersections or intersections with damaged fixed detectors, it is impossible to know the real-time status of the intersection due to the inability to obtain accurate queue length information. In view of the above situation, this paper proposes to estimate the queuing length of signalized intersections through floating car trajectory data. Through the analysis of trajectory data, this paper establishes a queue length estimation model based on floating car cluster queues. The queuing length estimation model can be used to realize the intersection Estimated queue length. Obtain the floating car trajectory data input under the morning peak, evening peak, and flat peak flow through SUMO simulation. The morning peak, evening peak, and flat peak flow time intervals are 2s/4s/6s/8s/10s and the penetration rate is 5%. /10%/15% of the data set to validate the model. The verification results show that the model established in this paper can estimate the queuing length at intersections more accurately. The comparison results with related methods show that the model established in this paper has lower errors under the morning peak traffic and the evening peak traffic, and the accuracy of the model established in this paper is about 12% higher than that without considering the cluster fleet under different time intervals and penetration rates. Even in the scenario of flat peak flow, large time interval, and low permeability, the error of queue length estimated by the model established in this paper is still within an acceptable range. Related research results will provide support for the assessment of traffic conditions at intersections and the optimization of signal timing.