Contextualized Probability for Approximation of Spatiotemporal Data Distributions

In spatial database applications, moving objects are frequently characterized not only by their position, but also by the uncertainty associated with the measurement of that position. In turn, uncertainty models allow for the specification of interpolators and extrapolators as a means to parametrically estimate historic and future trajectories. In nearly every instance, however, uncertainty models fail to account for how the environment through which an object is moving is likely to affect the set of probable locations associated therewith. This paper introduces a new approach to modeling uncertainty wherein environmental objects are incorporated in a manner that contextualizes the uncertainty utilizing the semantic characteristics of the objects in question. This new approach to modeling uncertainty offers new opportunities for managing moving objects within a database environment and, consequently, dependent applications such as sensor networks, geographic information systems and related technologies.