Towards Unified Spatio-Temporal Index for Hybrid Trajectory Search

Tianyao Wen, Shengkun Zhu, Yiming Wang, Sheng Wang

Abstract: Trajectory search benefits many real-world applications, such as transport planning and congestion prediction. It is typically divided into real-time search and historical search. Existing works mainly design indexes for efficient historical trajectory search. However, their indexes require frequent reconstruction for streaming data, greatly impacting search efficiency. In this paper, we introduce a unified spatio-temporal index to support efficient hybrid trajectory search. We design real-time grids and spatio-temporal cubes to index streaming and historical data rapidly. Then, we develop multiple lightweight key-value posting lists to facilitate filtering and pruning, improving trajectory search efficiency. Furthermore, we introduce a novel search operator to efficiently answer range queries and propose a new trajectory similarity measure to speed up kNN queries. Experiments on two real-world datasets show that our approach achieves high efficiency, with an average query time of 10 ms, and significantly outperforms three state-of-the-art methods.