Traffic prediction and navigation using historical and current information

Abstract

We developed a traffic prediction and navigation system that deals with uncertainty of road traffic conditions by stochastic modeling of road networks. Our system consists of a data collecting system, a data management system, and a path planning system. First, the data collecting system gathers real-time travel time data using a mobile sensor network system, CarTel. GPS sensor units having wireless connectivity were deployed on taxis running around the Boston area, and report their position and time information to the networked database system. Second, the raw GPS data collected from this CarTel system is processed to generate a database storing the statistical information of road travel time. We organize a large amount of data in a form in which they can be accessed efficiently and can capture important aspects of road traffic conditions. Third, we developed efficient stochastic shortest path algorithms that find best paths depending on drivers' goals. We evaluate our algorithms using both simulations and real-world drives. Finally, we implemented a path planning system using historical and current information organized by our data management system. Our system provides a Web-based interface that is publicly usable. The interface provides traffic information, including optimal paths and visualized traffic conditions. Our system also offers analysis tools of users' own driving routes with user track-log uploading interface. We evaluate the system using taxi trajectories and human driving experiments.