Cycle by Cycle Queue e Length Estimation for Signalized Intersections Using Sampled Trajectory Data

摘要

Queue length is an important measure of intersection performance. The problem of queuelength estimation has been investigated for over fifty years. Different approaches basedon different data sources have been presented. With the latest development in vehicledetection technologies, especially probe vehicle technologies, the utilization of vehicletrajectory data has become possible. Many studies focus on using trajectory data to detecttraffic shockwaves. One major limitation is that these studies assume perfect information,requiring the trajectories of entire vehicle population.In this paper, an improved queue length estimation method for signalizedintersections is proposed. This method assumes sampled vehicle trajectories as the onlydata source and is able to provide cycle-by-cycle queue length estimation. The keystoneof the entire approach is the concept of Critical Points (CPs), which represent thechanging vehicle dynamics. A CP extraction algorithm is introduced to identify CPs fromraw trajectories. Using the CPs related to queue formation and dissipation, an improvedshockwave based queue length estimation method is proposed. The performance of thisapproach is evaluated using several data sets under different flow and signal timingscenarios, including a recently collected GPS logger data set. The results indicate that thistrajectory based approach is promising.