Abstract: Aiming at the existed most traffic simulation systems which have low precision to estimate the traffic state trend, this paper introduces the particle filter method to pre-estimate the motion state of vehicles, then designs and implements a traffic simulation system based on particle filter. First of all, the design of initialization, state transition, weight calculation and normalization and resample operation are introduced; Then the basic process how the measured data which reflects the real characteristics of the system is dynamically incorporated to adjust the distribution of particles for getting closer to the real state development is discussed; Finally, the system based on microscopic traffic simulation software MovSim is realized. Experiments indicate the proposed system is of high accuracy and stabilization of prediction.

Key words: traffic simulation, state estimation, data assimilation