Distributed, Continuous and Real-time Trajectory Similarity Search

Salman Ahmed Shaikh, Hiroyuki Kitagawa, Akiyoshi Matono

Abstract: Moving objects’ trajectory data is becoming increasingly available with the omnipresent GPS-equipped devices, e.g., smart phones, smart watches, etc. Identification of similar trajectories is a fundamental requirement of many real-world applications, for instance, car pooling, road planning, epidemic contact tracing. There exist a number of studies focusing on trajectory similarity search, where given a query trajectory, similar trajectories are searched from a trajectory store. GPS devices generate trajectories as a continuous data stream, highlighting the critical need for real-time and continuous trajectory similarity search. Thus, given a trajectories stream ($S_{\Gamma}$) and a trajectory store (R), we propose a distributed, continuous and real-time similarity search between $S_{\Gamma}$ and R. The proposed method employs an index-based strategy to facilitate effective similarity search, with the added capability of execution in distributed environments to ensure scalable processing. A comprehensive empirical study is provided to demonstrate the efficiency and pruning capability of the proposed approach.