Vehicle Trajectory Reconstruction Incorporating Probe and Fixed Sensor Data

Abstract

Trajectory estimation is essential for obtaining a complete picture of traffic flow with limited and continuously detected traffic data, which are helpful in evaluating transportation performance and developing precise control measures. Most existing models assume the first-in-first-out principle, which generally is violated by the overtaking action in microscopic simulations and observations. This study focused on improving the accuracy of trajectory reconstruction by incorporating probes and fixed sensor data in multilane facilities. Accordingly, we developed a staircase vehicle order–changing model to describe the overtaking behaviors of vehicles. A field-test data set containing Global Positioning System (GPS) trajectories and automatic vehicle identification (AVI) observations was collected from some probe position units and fixed vehicle-identification cameras. Empirical studies demonstrated that the estimated error of the proposed algorithm was approximately 7%, which was approximately 22% and 12.5% less than that of two benchmark models. These results verified the superiority of our proposed algorithm and confirmed the importance of considering the overtaking behavior of vehicles in trajectory reconstruction.