Accuracy and Resource Concumption in Tracking and Location Prediction

Tracking is an enabling technology for many location based services. Given that the location of a moving object changes continuously but the database cannot be updated continuously, the research issue is how to accurately maintain the current location of a large number of moving objects while minimizing the number of updates. The traditional approach used in existing commercial transportation systems is for the moving object or the cellular network to periodically update the location database; e.g. every 2 miles. We introduce a new location update policy, and show experimentally that it is superior to the simplistic policy currently used for tracking; the superiority is up to 43% depending on the uncertainty threshold. We also introduce a method of generating realistic synthetic spatio-temporal information, namely pseudo trajectories of moving objects. The method selects a random route, and superimposes on it speed patterns that were recorded during actual driving trips.