Travel time prediction by weighted fusion of probing vehicles and vehicle detectors data sources

Travel time information plays an important role in ITS, especially in advanced traveler information system (ATIS). Traditionally, travel time is predicted by a single data source, such as vehicle detectors (VD) or probing vehicles (PV). In this paper, we try to predict travel time by integrating these two data sources by a dynamic weighted fusion scheme. The weights of the data sources are dynamically determined by the distance weight scheme to enhance the prediction precision. The proposed TTP model is applied to a small traffic network located in the east and north district of Tainan City, Taiwan. VD data is provided by traffic bureau of Tainan city government and probing vehicles raw data is collected from a Taxi dispatching system. The experiment results show that dynamic weighted combination of these two data sources can enhance the precision of the TTP, and the prediction stability of the proposed model is better than both the single source TTP models (VD or PV).