Travel Speed Forecasting Algorithm Based on Data Fusion

Due to the variety of equipment and collection conditions, different traffic data collection methods have their own application fields and different precision. Therefore, in order to acquire more reliable analysis results, the data from different detectors needs to be fused. And then travel speed estimation should be carried out based on the data fusion results. The paper presents a short-term forecasting algorithm for travel speed based on Kalman filter data fusion. After the pretreatment and traffic states division, the multi-source data including microwave detector data, loop detector data and Floating Car Data (FCD) are fused to estimate travel speed according to the different traffic states. This research determines parameters for Kalman filter, and computes the states Transfer Matrix by using the artificial neural network (ANN). The analysis result of model validation shows that the average accuracy of travel speed forecasting algorithm is over 90 percent.