Real-time data-driven trajectory reconstruction based on rough set theory

Real-time detailed vehicle trajectory data can be utilized to estimate traffic parameters for improving the efficiency of traffic control systems. Due to the low frequency of GPS data, the original vehicle trajectory obtained from the GPS device may be incomplete. This results in information loss or parameter estimation errors. In order to obtain more accurate traffic information from vehicle trajectory data, it is necessary to reconstruct or complete the original vehicle trajectory. However, for the real-time trajectory reconstruction there are two issues that need to be addressed. One is the real time original trajectory construction. The other is the original trajectory reconstruction. In this paper, a distributed traffic management framework and a data-driven trajectory reconstruction algorithm are proposed to address these two issues. GPS data taken from vehicles travelling on the segment are first delivered into the local traffic management center to achieve the map-matching for constructing the original vehicle trajectories. These original trajectories are reconstructed in real time on the local center using the proposed data-driven trajectory reconstruction method which is based on rough set theory. To obtain a more detailed reconstructed vehicle trajectory, reasonable condition and decision attributes are selected to establish a trajectory decision system based on the historical trajectory dataset. Discernibility based attribute weight is calculated using rough set theory to select the trajectory decision rules for vehicle trajectory reconstruction. Simulations illustrate that the proposed real-time data-driven trajectory reconstruction method can achieve vehicle trajectory reconstruction using the historical traffic trajectory dataset.