Grid-based k-Nearest Neighbor Queries over Moving Object Trajectories with MapReduce

Abstract

k-Nearest Neighbor Trajectory (k-NNT) Query is a basic and important spatial query operation widely used in many fields, such as intelligent transportation and urban planning. However, with the rapid increase of trajectory data volume, traditional k-NNT query algorithms for centralized environment are not effective and scalable enough, because the computational complexity increases dramatically when the spatial continuity of trajectories is considered. To address this problem, we propose a distributed grid index for trajectory data which partitions the trajectory into grids under MapReduce framework. Furthermore, a parallel query approach MR-GB-KNNT is proposed based on the proposed grid index to improve the efficiency and scalability of the k-NNT query. The experiment demonstrates that MR-GB-KNNT could perform well in cloud computing environment and improve the querying performance of the k-NNT.