Distributed Particle Filter for Urban Traffic Networks Using a Platoon-Based Model

Abstract

Raw measurement data are too noisy to directly obtain queue and traffic flow estimates usable for feedback control of urban traffic. In this paper, we propose a recursive filter to estimate traffic state by combining the real-time measurements with a reduced model of expected traffic behavior. The latter is based on platoons rather than individual vehicles in order to achieve faster implementations. This new model is used as a predictor for real-time traffic estimation using the particle filtering framework. As it becomes infeasible to let a truly large traffic network be managed by one central computer, with which all the local units would have to communicate, we also propose a distributed version of the particle filter (PF) where the local estimators exchange information on flows at their common boundaries. We assess the quality of our platoon-based PFs, both centralized and distributed, by comparing their queue-size estimates with the true queue sizes in simulated data.