Variational formulation incorporating spatial capacity changes to reconstruct trajectory for heterogeneous traffic conditions

Abstract

This study deals with the reconstruction of vehicle trajectories incorporating a data fusion framework that combines video and probe sensor data in heterogeneous traffic conditions. The framework is based on the application of variational formulation (VF) of kinematic waves for multiple lane conditions. The VF requires cumulative count and reference trajectory as boundary conditions. The VF also requires generation of lopsided network using fundamental diagram (FD) parameters. In this regard, cumulative count and FD parameters are obtained from the video sensor, while reference vehicle trajectory is obtained from the probe sensor. The analysis shows that the framework can provide an accuracy of 83% in trajectory estimation from the nearest reference trajectory. However, the accuracy decreases as the reference trajectory gets farther away from the estimated one. Additionally, an extension of the VF to accommodate roadway side friction is presented. The FD as well as lopsided network reform when the roadway capacity varies due to side friction. Consequently, the vehicle trajectory bends to accommodate the capacity fluctuation.