Interpolating and Using Most Likely Trajectories in Moving-Objects Databases

In recent years, many emerging database applications deal with large sets of continuously moving data objects. Since no computer system can commit continuously occurring infinitesimal changes to the database, related data management techniques view a moving object’s trajectory as a sequence of discretely reported spatiotemporal points. For each pair of consecutive committed trajectory points, a spatiotemporal uncertainty region representing all possible in-between trajectory points is defined. To support trajectory queries with a non-uniform probability distribution model, the query system needs to compute (interpolate) the “most likely” trajectories in the uncertainty regions to determine the peak points of the probability distributions. This paper proposes a generalized trajectory interpolation model using parametric trajectory representations. In addition, the paper expands and investigates three practical specializations of our proposed model using a moving object with momentum, i.e., a vehicle, as the exemplar.