On The Application Of Filter Methods To Traffic State Estimation

Traffic state estimation is an important part of intelligent traffic management, the estimation quality will directly affect the results the traffic control and management. Traffic state estimation problem to be solved is how to analyze the data from the traffic flow information collection equipment in the traffic state changes with randomness and uncertainty, finding out the regularity and establishing corresponding forecasting method and model, to estimate traffic state change trend.The traditional traffic estimation algorithms are limited duo to the strong nonlinear characteristics and increasing complexity of the traffic system. In order to deal with this problem, this study attempts to apply particle swarm optimized particle filter and Shadowing Filter to solve traffic state estimation problems.When the particle filter (PF) is applied to solve traffic state estimation problems, importance functions adopted are suboptimal during importance sampling processes. In order to reduce the impact of this suboptimality, this study attempted to introduce particle swarm optimizations (PSOs) into PFs. The PF combining with the PSO is named as PSOPF. To test the effectiveness of the PSOPF, twin experiments using PF and PSOPF respectively are conducted and the variance estimation was used to measure the robustness of PF and PSOPF. The results show that the RFC based on PSOPF in quality and stability is apparently more advantageous than that based on PF.Shadowing Filter (SF) method based on dynamic model can obtain the optimum estimation by adjusting the deviation between estimated value in the filter window and the corresponding estimated value obtained by model integration to minimum the whole deviation.This study attempts to utilize the SF method to get the optimal estimation of model trajectory which is closed to observations. SF and PSOPF methods are compared by experiments. Numerical experiment results show that SF is superior to PSOPF when the observation noise is greater than the system noise.