Existing airborne defense systems integrate a wide variety of sensors necessary to provide operators with situational awareness across the visual, thermal, signals, and electromagnetic spectrums. To date, individual sensor systems have been largely stove-piped, as have Artificial Intelligence/Machine Learning (AI/ML) and advanced, Size, Weight, and Power (SWaP)-optimized data processing systems. The resulting challenge has largely precluded the reuse and feasibility of AI/ML solutions across platforms and sensors, resulting in fewer capabilities being deployed across SOCOM and DoD sensors at a higher cost per capability. To fill this gap, SOCOM requires portable, open architecture solution for orchestration of AI/ML at the edge that consists of three primary pillars: 1) an open architecture computing environment to provide computing resources to mange AI/ML from multiple sensors simultaneously; 2) a container-based software architecture to run AI/ML orchestrators and inference models against sensor data; and 3) a scalable AI/ML development and training pipeline and edge orchestration mechanism to transform cross-sensor data into actionable information.
