Voronoi-Neighboring Regions Tree for Efficient Processing of Location Dependent Queries

As a data management technique, indexing aims to judicious organization of data allowing efficient query processing. In the context of location based services (LBSs), indexing techniques are, substantially, affected by location dependency and modes of data access. This paper focuses on the processing of location dependent queries (nearest neighbors and range queries) based on indexing structures with respect to the context of LBSs. In fact, traditional indexes (R-trees, kd-trees …) adopt either overlapped partitioning schemes or backtracked search algorithms. These features increase response times for an in-memory index within an on-demand data access mode. They, also, make such indexes impractical for a broadcast data access mode. As a solution, we propose a novel index called VNR-tree (Voronoi-Neighboring Regions tree), based on the Delaunay Triangulation. VNR-tree adopts a non-overlapping partitioning scheme, supports backtracking-free search algorithms and generates a special clustering of the Voronoi-neighboring regions. This clustering is adequate to the processing of several types of location dependent queries based on the exploration of the mobile client’s Voronoi-neighborhoods. We conduct various experiments to compare VNR-tree with R*-tree as in-memory indexes in the on-demand data access mode. Main results show that VNR-tree outperforms R*-tree significantly.