Vehicle route prediction and time of arrival estimation techniques for improved transportation system management

Many intelligent transportation system services are making use of vehicle position and time information to help manage traffic congestion and in the case of public transportation, improve service reliability. Position and time data can be acquired in a variety of ways including probe vehicles, embedded road sensors, tagging electronic toll collection transponders, automatic vehicle location (AVL) systems, etc. As part of advanced public transportation systems, many transit agencies are placing AVL systems on their buses to better manage their fleet and to provide their passengers with better information on vehicle arrivals and departures. Another public transport system that can benefit from this technology is shared-use vehicle systems (i.e., carsharing, station cars). In particular, multiple-station shared-use vehicle systems need to manage their fleet closely to avoid vehicle distribution problems at their stations. In this paper, we describe a route prediction and time-of-arrival estimation methodology that has been implemented with a real-world multiple-station carsharing system. Each vehicle has AVL hardware, producing position and time data that can be used at the system level for estimation. This is slightly different than bus arrival estimation techniques in that the routes are not known a priori. The developed algorithm was tested against actual routes and arrival times, producing encouraging results. It was determined that the estimation procedure could be improved if the frequency of position/time data was increased.