Processing of Continuous k Nearest Neighbor Queries in Road Networks

Continuous Nearest Neighbor (NN) monitoring in road networks has recently received many attentions. In many scenarios, there are two kinds of continuous k-NN queries with different semantics. For instance, query "finding the nearest neighbor from me along my moving direction" may return results different from query "finding the nearest neighbor from my current location". However, most existing continuous k-NN monitoring algorithms only support one kind of the above semantic queries. In this paper, we present a novel directional graph model for road networks to simultaneously support these two kinds of continuous k-NN queries by introducing unidirectional network distance and bidirectional network distance metrics. Considering the computational capability of mobile client to locate the edge containing it, we use memory-resident hash table and linear list structures to describe the moving objects and store the directional model. We propose the unidirectional network expansion algorithm and bidirectional network expansion algorithm to reduce the CPU cost of continuous k-NN queries processing. Experimental results show that the above two algorithms outperform existing algorithms including IMA and MKNN algorithms.