Unscented Kalman filter for urban link travel time estimation with mid-link sinks and sources

To estimate link travel time, the classical analytical procedure uses vehicles counts at upstream and downstream locations. This procedure is vulnerable in urban networks mainly due to significant flow to and from mid-link sinks and sources. One of the important developments recently done on this topic has yielded to the CUPRITE methodology. This method is derived from the classical analytical procedure. It integrates probe vehicle data to correct deterministically the upstream cumulative plot to match the information of probe vehicles travel times, whilst the downstream cumulative plot is kept unchanged. The algorithm proposed and validated in this research estimates urban links travel times based on an unscented Kalman filter (UKF). This algorithm integrates stochastically the vehicle count data from underground loop detectors at the end of every link and the travel time from probe vehicles. The proposed methodology, which can be used for travel time estimation in real-time, is compared to the classical analytical procedure and to the CUPRITE method in case of mid-link perturbation. Along to its lower sensitivity than CUPRITE, the UKF algorithm makes it possible detection and exclusion of outliers from both data sources.