Development of a Hybrid Model for Dynamic Travel-Time Prediction

Abstract

Travel-time prediction has been an interesting research subject for decades, and various prediction models have been developed. A prediction model was derived by integrating path-based and link-based prediction models. Prediction results generated by the hybrid model and their accuracy are compared with those generated by the path-based and link-based models individually. The models were developed with real-time and historic data collected from the New York State Thruway by the Transportation Operations Coordinating Committee. In these models, the Kalman filtering algorithm is applied for travel-time prediction because of its significance in continuously updating the state variables as new observations. The experimental results reveal that the travel times predicted with the path-based model are better than those predicted with the link-based model during peak periods, and vice versa. The hybrid model derives results from the best model at a given time, thus optimizing the performance. A prototype prediction system was developed on the World Wide Web.