论文部分内容阅读
In this article, we describe the development of a transit trip planner (TTP) for mobile devices called Transitr, and evaluate its performance. The system predicts the shortest paths between any two points in the transit network using real-time information provided by a third party bus arrival prediction system, relying on GPS equipped transit vehicles. Users submit their origin and destination through a map-based iPhone application, or through a JavaScript enabled web browser. A server implementing a dynamic K-shortest paths algorithm with predicted link travel times returns personalized route directions for the user, displayed on a map. To assess the optimality and accuracy of the predicted shortest paths, an a posteriori comparison with a schedule-based transit trip planner and the GPS traces of the transit vehicles is performed on six-hundred origin destination pairs in San Francisco. The results show that routing using the predicted bus arrivals marginally increases the accuracy of the total travel time and the optimality of the route. Suggestions to improve the accuracy and optimality using real-time information are proposed.