DATA FUSION CONCEPT TO ESTIMATE VEHICLE TRAJECTORIES ON URBAN ARTERIALS

摘要

Fusion of probe and fixed sensor data has been widely used to estimate travel time on urbanarterials. Considering different accuracies and limitations of traffic data from various sources,data fusion is applied to extend the spatial and temporal coverage of data. Majority of existingfusion techniques to estimate travel time are merely based on statistical methods withoutconsiderations for traffic engineering concepts. In addition, probe trajectories include muchricher information of traffic conditions rather than only travel times, which could be usedmore effectively. In this research a data fusion technique is proposed to estimate vehicletrajectories using multi-sensor traffic data and concepts of traffic engineering which can fullyutilize probe trajectory information. Proposed method is based on modeling vehicletrajectories by applying the simplified theory of 3D kinematic waves. Assuming a triangularshaped flow-density curve, variational formulation of kinematic waves is considered toestimate cumulative number of vehicles in time and space. Given cumulative number ofvehicles as heights, a three dimensional surface is created over time-space plane and vehicletrajectories are estimated as contours of the surface. Once vehicle trajectories are estimated,they can be used for several purposes including travel time estimation and signal timingoptimization. Performance of the method regarding travel time estimation is evaluated and itis shown that it is capable of estimating reliable travel times on urban arterials. At the endapplication of the method for signal timing optimization is demonstrated.