Unscented Kalman Filter-Based Real-Time Traffic State Estimation

Real-time traffic states and origin-destination (OD) flows are essential inputs for on-line traffic control and management systems. With the final goal to unify the problem of real-time estimation of traffic states and origin-destination (OD) flows, this paper follows the same idea as in the previous work to develop a new traffic state estimator based on new filtering technique, Unscented Kalman Filter (UKF). The model formulates a discrete macroscopic model into a state-space form and applies the UKF as a correction algorithm to make the estimate be consistent with the observed traffic data. The method is tested using real traffic data with traffic condition changes abruptly from light to heavy traffic and finally returns from heavy to light. Its performance is also compared with the method which uses only macroscopic model and the method that uses macroscopic model with Extended Kalman Filter. Numerical results show that the new method outperforms other methods and be capable of capturing abrupt changes of traffic condition despite a long spacing between detectors. This demonstrates a high potential of the UKF for future application in traffic state estimation.