Efficient query processing techniques for spatio-temporal databases

Abstract

Spatio-temporal databases deal with data whose geometry changes over time. There are many real-life applications that create such data, including intelligent traffic systems and multimedia applications. This thesis presents an adaptive indexing technique for timestamp queries and interval queries using query workloads, and an efficient and scalable approach to nearest neighbor queries and continuous nearest neighbor queries in the presence of a road network. Many spatio-temporal access methods, such as the HR-tree, the 3DR-tree, and the MV3R-tree, have been proposed for timestamp and interval queries. However, these access methods have the following problems: the poor performance of the 3DR-tree for timestamp queries, the huge size and the poor performance of the HR-tree for interval queries, and the large size and the high update cost of the MV3R-tree. We address these problems by proposing an adaptive partitioning technique called the Adaptive Partitioned R-tree (APR-tree) using workloads which consist of timestamp and interval queries. The APR-tree adaptively partitions the time domain using query workloads. Since the time domain of the APR-tree is automatically fitted to query workloads, the APR-tree is superior to the other access methods for various query workloads. The size of the APR-tree is on average 1.3 times larger than that of the 3DR-tree which has the smallest size. The average update cost of the APR-tree is similar to that of the 3DR-tree, which also has the smallest update cost. A continuous search in a road network retrieves the objects which satisfy a query condition at any point on a path. For example, return the three nearest restaurants from all locations on my route from point s to point e. We deal with NN queries as well as continuous NN queries in spatio-temporal databases. The performance of existing approaches based on the network distance such as the shortest path length depends largely on the density of objects of interest. ...