Time relaxed spatiotemporal trajectory joins

Many spatiotemporal applications store moving object data in the form of trajectories. Various recent works have addressed interesting queries on trajectorial data, mainly focusing on range queries and Nearest Neighbor queries. Here we examine another interesting query, the Time Relaxed Spatiotemporal Trajectory Join (TRSTJ) which effectively finds groups of moving objects that have followed similar movements in different times. We first attempt to address the TRSTJ problem using a symbolic representation algorithm, which we have recently proposed for trajectory joins. However we show experimentally that this solution produces false positives that grow rapidly with the increase of the problem size. As a result, it is inefficient for TRSTJ queries as it leads to large query time overhead. In order to improve query performance, we propose two important heuristics that turn the symbolic represenation approach effective for TRSTJ queries. Our first improvement, allows the use of multiple origins when processing strings representing trajectories. The experimental evaluation shows that the multiple-origin approach drastically reduces query performance. We then present a ``divide and conquer'' approach to further reduce false positives through symbolic class separation. The proposed solutions can be combined together, which leads to even better query performance. We present an experimental study revealing the advantages of using these approaches for solving Time Relaxed Spatiotemporal Trajectory Join queries.