The advanced traveler information provides realtime information about estimated time of arrival for the bus to the commuters. The bus transmits its location information to the base station frequently and then base station uses this available location information and advanced prediction algorithms to estimate the time of arrival for rest of the stops on the itinerary. This frequent communication between the bus and base station adds a recurring cost to the implementation and management of such a system. In this paper we wish to study performance of the system wherein the bus transmits its information only when it is pulled by base station. Algorithms that determine when to request information are developed for the two scenarios: one where motion on each link on the road network is modeled with a single common statistical model, and another where each link on the road has an individual model for vehicle motion. The second scenario can generate more accurate predictions of vehicle motion, but the algorithms for determining when to request information from the vehicles are more complex. The resulting algorithms are evaluated using Monte Carlo simulations to illustrate the performance of the approaches.