Using vTree indices for queries over objects with complex motions

We introduce the vTree, an index structure for efficient processing of spatiotemporal queries over sets of objects moving along complex trajectories. The vTree is a tiered structure, and partitions space at different granularities at different tiers. It uses two novel strategies to enhance the performance of spatiotemporal queries. First, it groups objects by velocity, and indexes objects from each group at an appropriate tier in the vTree, to localize the loss of precision induced by fast objects. Second, it accommodates complex trajectories by controlled replication of object descriptors at each tier. These features permit vTree indices to remain useful for longer time durations, and to support very efficient query processing. Our algorithms for vTree joins are designed to limit the portions of index and data space explored, as well as to maximize locality within the portion of space explored.