In-Memory Distance Threshold Similarity Searches on Moving Object Trajectories

The need to query spatiotemporal databases that store trajectories of moving objects arises in a broad range of application domains. In this work, we focus on in-memory distance threshold searches which return all moving objects that are found within a given distance d of a fixed or moving object over a time interval. We propose algorithms to solve such searches efficiently, using an R-tree index to store trajectory data and two methods for filtering out trajectory segments so as to reduce segment processing time. We evaluate our algorithms on both real-world and synthetic in-memory trajectory datasets. Choosing an efficient trajectory splitting strategy to reduce index resolution increases the efficiency of distance threshold searches. Moreover, we demonstrate that distance threshold searches can be performed in parallel using a multithreaded implementation and we observe that high parallel efficiency (72.2%-85.7%) can be obtained. Interestingly, the traditional notion of considering good trajectory splits by minimizing the volume of hyperrectangular minimum bounding boxes (MBBs) so as to reduce index overlap is not well-suited to improve the performance of in-memory distance threshold searches. Keywords-query optimization; query parallelization; spatiotemporal databases; trajectory searches.