Dynamic Travel Time Prediction using Pattern Recognition

Travel-time information is an essential part of Advanced Traveler Information Systems (ATISs) and Advanced Traffic Management Systems (ATMSs). A key component of these systems is the prediction of travel times. From the perspective of travelers such information may assist in making better route choice and departure time decisions. For transportation agencies these data provide criteria with which to better manage and control traffic to reduce congestion. This study proposes a dynamic travel time prediction algorithm that matches current traffic patterns to historical data. Unlike previous approaches that use travel time as the control variable, the approach uses the temporal-spatial traffic state evolution to match traffic states and predict travel times. The approach first identifies candidate historical time intervals by matching real-time traffic state data against historical data for use in prediction purposes. Subsequently, the selected candidates are used to predict the temporal-spatial evolution of traffic. Lastly, dynamic travel times are constructed using the identified candidate historical data. The proposed algorithm is tested on a 37-mile freeway segment from Newport News to Virginia Beach along the I-64 and I-264 freeways using historical INRIX data. The prediction results indicate that the proposed method produces predictions that are more accurate than the state-of-the-art K-Nearest Neighbor methods reducing the prediction error by 15 percent to less than 3 minutes on a 50-minute trip.