Efficient Algorithm to Monitor Continuous kNN Queries

Continuous monitoring of k nearest neighbor (kNN) queries has attracted significant research attention in the past few years. A safe region is an area such that as long as a kNN query remains in it, the set of its k nearest neighbors does not change. Hence, the server does not need to update the query results unless the query moves out of its safe region. Previous work uses time-parameterized kNN (TPkNN) queries to construct the safe region. In this paper, we present an efficient technique to construct safe regions by using much cheaper RangeNN queries. Moreover, unlike TPkNN queries, the cost of a RangeNN query is not significantly affected as the value of k increases. Hence, our proposed algorithm scales better with the increase in the value of k. We also present a technique to efficiently update the safe regions when the underlying dataset is dynamic (i.e., objects appear or disappear from the dataset). Extensive experimental results show that the proposed algorithm provides an order of magnitude improvement over existing approach on both the static and dynamic datasets.