Prediction of dynamic freeway travel times based on vehicle trajectory construction

The paper develops a novel approach to construct vehicle trajectories using real-time and historical traffic data to predict dynamic travel times. The approach combines real-time and historical data within a particle filter framework to dynamically predict future traffic state maps. The predicted travel trajectory is then constructed using the velocity spatiotemporal map. Based on the nature of particle filters, the variance of each speed grid traversed during the trip can be calculated and then used to compute the travel time variance. The proposed approach is tested using simulated data along a section of I-66. The results demonstrate the effectiveness of the proposed algorithm in predicting travel times.