Online Dynamic Travel Time Prediction using Speed and Flow Measurements
Abstract
Traffic forecasting is considered nowadays as one of the most important traffic management techniques on road networks. To provide suitable control strategies and advanced traveler information, that improve traffic performance, a continuous short-term prediction is a significant requirement. This paper concentrates on online and continuous travel time prediction between to points of interest. To this aim two different type of sections need to be identified, links (divided in several cells) and nodes. Inside the links, the travel time forecasting is achieved by predicting the densities inside the cells using a CTM-predictor, while the boundary flows are predicted using an adaptive Kalman filtering approach. On the other hand, as we will take into account the latency of the nodes, we assume that sensors located at the node levels provide also speed measurements. The time series characterizing the node latency, computed from the mean speed inside the node, is thus predicted using the same adaptive Kalman filtering approach (AKF). The complete travel time between two points of interest is therefore predicted by the sum of the travel times of the cells in the link and all the nodes. The performance of the proposed method is evaluated by using data of the Grenoble south ring, a case study of the NoE Hycon2.