Dynamic data driven event reconstruction for traffic simulation using sequential Monte Carlo methods

Simulation models are commonly used to study traffic systems. Accurate traffic predictions need proper characterization of the traffic flow and knowledge of related parameters representing the state of the traffic flow in the models. To correctly estimate the traffic flow in real time, we need to reconstruct the event by answering such critical questions as the source of the congestions. The availability of sensor data from the real traffic provides information that can be assimilated into a traffic simulation model for improving predicted results. In this paper, we use the sequential Monte Carlo methods to assimilate real time sensor data into the simulation model MovSim, an open-source vehicular-traffic simulator, to reconstruct events such as the slow vehicles that cause the traffic jam. Related experimental results are presented and analyzed.