Highly Integrated and Distributed Recognition Architecture for Anti-Ballistic Missile (HYDRA-ABM) System

American GNC Corporation (AGNC) and the University at Albany - State University of New York (UAlbany) propose the Highly Integrated and Distributed Recognition Architecture for Anti-Ballistic Missile (HYDRA-ABM) System. This technology is designed considering: (i) Missile Defense Systems requirements; (ii) a variety of networked sensors; and (iii) networked communications. High level system components are: (1) set of nodes, (2) parameter server, and (3) communication channels. The system design considers the effects of communication channel constraints related with the (i) sensor system communication module and (ii) characteristics of the networked communication to interconnect the set of nodes. The core benefit is a flexible and distributed object classification system optimized by Federated Learning and designed for networked, multimodal sensors with non-IID data. Specific innovations are: (1) a classification system with performance optimization by Federated Learning where the classification knowledge of each node is incorporated in the global learning process (and shared among nodes) while training data requirements are relaxed; (2) system architecture and software framework for integration of a variety of disparate missile defense sensors considering networked and multimodal sensors and data communication module and channels constraints; and (3) a novel system implementation for missile defense systems with global optimization using an advanced distributed learning process specifically tailored to meet DoD needs. Approved for Public Release | 21-MDA-11013 (19 Nov 21)