Dynamic Data Acquisition and Parameters Estimation for Traffic Prediction.

This study proposes a real-time traffic data acquisition system and prediction algorithm. The framework of the system suggests taxi fleets as probevehicles, combining roadside detectors to collect data from urban networks extensively. Then, mathematical models of “link travel time prediction” and “route flow estimation” are built based on generalized least squares and extended Kalman filter. To verify the prediction capability of the models, this study analyzed the results from grid network simulation. The models are proven well functioning with data processing and calibration. The mean errors of flow estimation on the generated network traffic flows are within 15%.