An analysis model of queue length fluctuation at signals using vehicle trajectories

Abstract

Floating car systems have attracted much attention. In particular, trajectory information of vehicles, which cannot be obtained in legacy traffic monitoring systems, would be significant for more detailed and precise traffic prediction. In this paper, we focus on vehicle queues formed in front of intersections and present a model that analyzes the dynamics of their length using vehicle trajectory information. In general, a queue grows in a red phase of the traffic signal cycle and shrinks in a green phase. We assume that vehicles arrive at a signal with a Poisson process, and estimate the queue length for each signal cycle. Our method was evaluated using a traffic simulator where the real field data were injected. Evaluation results have shown that the queue length could be estimated and the mean error was about 5 vehicles, which is acceptable compared with the whole queue sizes.