context-aware RF EM

Rapid identification and geolocation of electromagnetic emitters is required for many military and civilian applications. Recent progress in deep learning and signal processing has enabled AI-based inference for real-time signal detection, geolocation, and classification, and highly realistic 3D models have become available for nearly all locations around the globe. RF simulations in combination with these 3D geometric models provide novel pathways for generating training data required for these inference algorithms. We will leverage these 3D model generation technologies and develop novel methods in radar processing for locating and exploiting RF sources within a complex geometric environment. Our approach will establish geolocation locks on RF sources using You Only Look Once inference methods. Our algorithms will be deployed on RF-surveying payloads for Class I and Class II UAS airframes. Our end-to-end model development and deployment time will meet the required 16-hour specification window, and performance will be compared against similar state of the art approaches.