Multi-scale Windowing over Trajectory Streams

Abstract

Many modern monitoring applications collect massive volumes of positional information and must readily respond to a variety of continuous queries in real-time. An important class of such requests concerns evolving trajectories generated by the streaming locations of moving point objects, like GPS-equipped vehicles, commodities with RFID’s etc. In this paper, we suggest an advanced windowing construct that enables online, incremental examination of recent motion paths at multiple levels of detail. This spatiotemporal operator can actually abstract trajectories at progressively coarser resolutions towards the past, while retaining finer features closer to the present. We explain the semantics of such multi-scale sliding windows through parametrized functions that can effectively capture their spatiotemporal properties. We point out that window specification is much more than a powerful means for efficient processing of multiple concurrent queries; it can be also used to obtain concrete subsequences from each trajectory, thus preserving continuity in time and contiguity in space for the respective segments. Finally, we exemplify window utilization for characteristic queries and we also discuss algorithmic challenges in their ongoing implementation.