An ontology for active and passive aerial drone threat automatic plan recognition
Ronald Prescott Loui, Josh K. Smith · 2016
This paper initiates a discussion on the design of terms, features, and descriptors that would support machine learning for automated plan recognition of drone and drone swarms engaged in threatening activity. A few prototype aerial missions for drones are discussed and semantic markers, such as distance and line of sight to potential targets, mirrored motion, path and position optimality, coordination, and formation, are noted. This semantic description of motion in terms of objectives and capabilities contrasts with a naïve description of motion in a 3d coordinate system without reference to targets; terminology is the first step in automated anomaly detection analytics. The paper further discusses active plan recognition, which selects interventions in order to force the drone or swarm to reveal its intentions. Analogies to, and distinctions from, two-dimensional active plan discernment, e.g., stalking, tailing, pursuing, and intercepting, are given.