Online reverse subpattern matching for reproducing trajectories from sub-paths

Abstract

In this paper, we present an online method for reverse subpattern matching (RSM). RSM is a special kind of subpattern matching, whereby the query trajectory is longer than the patterns for matching. This is the case, if trajectories have to be reproduced from given sub-paths. In particular, we use as a showcase an anomaly detection method for Self-Organizing Industrial Systems (SOIS) where the movements of the robots in the factory have to be reproduced online by given moving patterns. Assuming network-constrained trajectories, we introduce edge lists as suitable data structure for indexing the pattern dictionary. Based on them candidate lists are built and updated each time a new edge of the query trajectory is recorded. Patterns of the candidate lists which are completely contained in the trajectory are shifted to the matched pattern list. In addition, we propose the covering rate as a quality indicator to discover trends about the reproducibility of the trajectory as soon as possible. The work-flow of the presented method is illustratively evaluated based on a SOIS scenario where anomalies are detected by reproducing the trajectories from given patterns online.