Query Processing on Historical Uncertain Spatiotemporal Data
(TAO Yufei)

A predictive spatiotemporal database maintains the current motion characteristics of the monitored entities (e.g., vehicles, aircrafts, weather patterns, etc.), and efficiently processes queries that return objects whose future locations satisfy certain spatial and temporal predicates (e.g., "find the aircrafts expected to be in the airspace of Hong Kong 1 hour later"). The existing methods, however, assume that objects are moving linearly with fixed velocities and hence, are not applicable to (real-world) entities with general trajectories. In this project, we aim at eliminating this linearity assumption by developing techniques that support effective retrieval of objects with diverse motion patterns that are unknown in advance. Our first objective is to study theoretical tools for discovering the motion functions of nonlinearly moving objects from their (recent) past locations. The second goal is to design efficient access methods and algorithms that minimize query costs.