Processamento Eficiente de Junções Espaçotemporais

It's a well-known fact that the new GIS applications need to keep track of temporal information. Among other operations, a spatiotemporal DBMS should efficiently answer the spatiotemporal join. The best-known spatial index structure, the R-Tree (and its variants), does not preserve the MBRs’ evolution. New indexing structures were proposed in the literature that allows the retrieving of present and past states of data, and most of them are R-Tree based. This paper presents a first study of spatiotemporal join processing using these new structures, particularly a partially persistent R-Tree called Temporal R-Tree and the 2+3D R-Tree. Starting from spatial join algorithms, we present algorithms for processing spatiotemporal joins over time instants and intervals on both spatiotemporal structures. Then, we propose some improvements that lead to a better performance and try to show the correctness of our algorithms. Finally, we implement and test these new algorithms with some spatiotemporal data sets. Our experiments shows that our algorithms performance is good even on extreme cases, like datasets with many changes, showing its good scalability – especially for the TR-Tree. In addition, with minor adaptations, the main ideas of our algorithm can be used for evaluating joins using other partially persistent structures, like the MVB-Tree.