Predicted Range Aggregate Processing in Spatio-temporal Databases

Predicted range aggregate (PRA) query is an important researching issue in spatio-temporal databases. Recent studies have developed two major classes of PRA query methods: (1) accurate approaches, which search the common moving objects indexes to obtain an accurate result; and (2) estimate methods, which utilize approximate techniques to estimate the result with an acceptable error. In this paper, we present a novel accurate prediction index technique, named PRA-tree, for range aggregation of moving objects. PRA-tree takes into account both the velocity and space distribution of moving objects. First, the velocity domain is partitioned into different velocity buckets, and moving objects are classified into different velocity buckets by their velocities, thus objects in one bucket have similar velocities. Then we use aTPR-tree, which is based on the basic TPR-tree structure and added with aggregate information in intermediate nodes, to index objects in each bucket. PRA-tree is supplemented by a hash index on IDs of moving objects, and exploits bottom-up deletion algorithm, thus having a good dynamic performance and concurrency. Also new PRA query methods with a more precise branch-andbound searching strategy are developed for PRAtree. Extensive experiments confirm that the proposed methods are efficient and practical.