Probabilistic nearest neighbor queries of uncertain data via wireless data broadcast

Abstract

Most existing location-dependent query processing methods are based on the client-server model. However, due to the increasing nubmer of smart mobile devices, there can be a large volume of data being processed on the server side and the server can be system performance bottleneck. This paper takes the first step towards processing probabilistic nearest neighbor queries of uncertain data objects via wireless data broadcast (BPNN). Our method leverages the key properties of Voronoi Diagrams for Uncertain Data (UV-Diagram). To preserve the good properties of UV-Diagram, according to the property of Hilbert curve, UV-Hilbert-Partition is proposed to partition the UV-Diagram into several grid cells, called Hilbert-Cells, which have good locality-preserving behavior. Then a special organizing method is proposed. For a certain UV-Diagram, the CellFrame structure, which can be located based on the coordinates of a query client, is used to efficiently minimize the broadcast cycle and keep the probabilistic nearest neighbor results. Based on the sequence of the CellFrames, a distributed index, called UVHilbert-DI, is proposed to support BPNN query processing. Finally, the efficient BPNN algorithms based on UVHilbert-DI is presented and extensive experiments have been conducted to demonstrate the performance of our approaches.