Trajectory-aware Load Adaption for Continuous Traffic Analytics Full text

Kostas Patroumpas, Serafeim Papadias
Symposium on Spatial and Temporal Databases (SSTD 2019)
2019
Conference/Workshop
Abstract. We introduce a framework for online monitoring of moving objects, which takes into account their evolving trajectories and copes smoothly with fluctuating demands of multiple continuous queries for limited system resources. This centralized scheme accepts streaming positional updates from numerous objects, but it only examines recent trajectory segments with expectedly higher utility in query evaluation, shedding the rest as immaterial. We focus on adaptive processing under extreme load conditions, opting to retain salient trajectory segments and possibly sacrifice smaller, frequently observed paths in favor of longer, distinctive routes. We propose heuristics for incremental, yet approximate, query evaluation in order to provide up-to-date traffic analytics using windows that abstract particular regions and time intervals of interest. Finally, we conduct a comprehensive experimental study to validate our approach, demonstrating its benefits in result accuracy and efficiency for almost real-time response to trajectory-based aggregates.