Continuous K Nearest Neighbor Queries of Moving Objects in Road Networks

Existing algorithms for continuous k-nearest neighbor(kNN) queries of moving objects in road networks adopt the incremental query mechanism,which is verified to be inefficient when data update frequently.Considering the ubiquitous multi-core and multi-threading technologies,a multi-threading based framework for continuous kNN queries of moving objects is proposed.In the framework,all the queries are re-evaluated periodically,and the query process is divided into two sequential phases:the data updating phase and the query execution phase,task parallel and data parallel mechanism are used respectively in each phase to carry out the corresponding operations.In the updating phase,the data structures used in the framework are designed.Moreover,in the execution phase a kNN query strategy is proposed which includes an off-line pre-computation and an on-line query algorithm.The computational complexity of the algorithm and the speedup of the framework are analyzed theoretically.Experimental results show that,under the frequent update environment,the proposed query algorithm when serially executed has better performance than the traditional algorithms,however,the multi-threading based parallel execution is better in all kinds of parameter setups;what's more,the multi-threading based parallel execution maintains good performance scalability,and the speedup can reach to 1.51~1.7.