Accurate Segment Travel Time Estimation Based on Individual Vehicle Data

Travel time is a crucial indicator of motorway network performance for drivers and transport managers alike. In addition, travel time estimation is a key factor for many research subjects; one example is to determine a search window in the vehicle re-identification problem. The typical method of estimating travel time is based on spot speeds, which are particularly inaccurate during peak hours. Even though loop detectors gather high-resolution traffic data on many highway networks, this data has not been used for segment travel time estimation. In this paper, we calibrated and applied the adaptive smoothing filter on individual vehicle data to provide accurate segment travel time estimation. We validated the accuracy of the calibrated adaptive smoothing filter in speed estimation at any given point in the spatio-temporal plane between pairs of loop detectors, and we demonstrated the travel time accuracy against a large sample of visually detected vehicles that travel between pairs of loop detectors. The results represent an accurate travel time estimation when the average errors of the segment travel time in non-congested and congested regimes are 2.12%, and 3.91%, respectively. The proposed method can be reused in vehicle re-identification algorithms.