Traffic Estimation Based on Particle Filtering with Stochastic State Reconstruction Using Mobile Network Data

The popularity of mobile communication provides transportation engineers a low-cost approach to collect position related data and time related data from mobile network for traffic estimation. This paper first introduces the mechanism of traffic estimation by the measurement of cell handoff data of moving vehicles and then presents two traffic estimation models based on Bayesian framework. The first-order model is a basic model which uses traffic speed as the only state variable. On the basis of the first-order model, the second-order model, incorporating the flow of traffic as the second state variable, has a two-level architecture, where macroscopic states and microscopic states are connected by the process of state reconstruction. This idea makes it possible to realize high-order sparse-sampling traffic estimation. Due to the good performance on solving highly nonlinear estimation problems, particle filters are introduced to provide the approximation solution to the traffic state estimation problem with system noise and measurement error. The performance evaluation and practical test of the particle filters are performed by numerical experiments. Estimation error can be controlled at an acceptable level through parameter initialization and adjustment. Finally, further research topics on model improvement and handoff-based transportation applications are discussed.