Tactical AI Counter Swarm System (TACOS): A Hybrid Machine Learning Approach for Identifying, Classifying, and Defeating Enemy Drone Swarming Attacks

We propose to develop, demonstrate, and commercialize Tactical AI-based Counter Swarm (TACOS) system, a hybrid yet synergistic fusion of expert system and deep learning algorithms in our friendly swarm to identify counter-swarm strategies, track them autonomously with no human control, and defeat them (both kinetically and non-kinetically) with high precision. Current swarm and counter-swarm technologies are experiencing an explosion in the variety of airframes, sensors, weapons (both kinetic and non-kinetic), and control methodologies available for rapid integration and use. As such, deep learning/reinforcement learning approaches to counter-swarm tactics may be invalidated by the rapid integration of new friendly or adversary capabilities. Thus, the technical problem of current swarm and counter-swarm systems is primarily its lack of integration of human expertise (logical reasoning in the form of expert system and ontology) and deep learning; therein exists an unprecedented opportunity to advance the field of counter-swarming technology that exploits the best of both expert system and machine learning. TACOS, with its fusion of expert system and deep learning integration as its core brain for detecting, tracking, and defeating even the most advanced adversary swarms, provides a highly effective, reliable, and explainable system, allowing for assured and bounded behavior and reliable command and control, even with limited, degraded, or intermittent communications. In addition to real-time tactical data (e.g., swarm flight paths), we also consider the swarm behaviors at the operational level which takes into account weapon characteristics, command & control structures, communication types, battery/power-source characteristics, and overall kinetic capabilities. A naïve, pure deep learning-based algorithms will likely fail to establish trustworthy relationship with human operators due to the black-box nature of deep learning algorithms. For Phase I, we will develop and demonstrate our prototype TACOS system as N-vs-M swarm-vs-swarm in a modular, open-system fashion that is capable of: (1) Detecting and classifying adversary swarm behaviors using a combination of swarm flight pattern classifiers and finite state predicate automata (FSPA) which detects swarm tactics; (2) Adapting TACOS responses autonomously to adversary swarm behaviors from (1); (3) Prototyping rapidly and demonstrating TACOS “agents” that actively counter adversary swarm with dynamic, expert-level actions based on ongoing data-stream at both tactical and operational levels, using high-fidelity simulation tools. Specifically, our core Tactical AI technology is designed to allow rapid integration of human-based expert knowledge in order to speed training and convergence to an optimally performing swarm. We anticipate a strong commercialization opportunity by utilizing our SwarmSense product line to readily add TACOS capabilities for both defense and non-defense markets.