AF241-0004: Context-Aware RF Electromagnetic Surveying for Exploiting Signals of Opportunity

Vulcan Ventura will develop a cutting-edge Radio Frequency (RF) geolocation and signal identification system that integrates 3D environmental models and RF propagation patterns, producing data for low-latency Artificial Intelligence/Machine Learning (AI/ML) models. Our Solution will be optimized for deployment on low Size, Weight, and Power (SWaP) RF Electronic Spectrum Monitoring (ESM) antenna arrays, specifically designed for Class I and Class II UAS airframes. The final system will operate within the 0.8 GHz - 6 GHz frequency range and be capable of geolocating RF sources in complex multipath environments, including scenarios where signals reach the sensor via up to two reflections, i.e. Non-Line-of-Sight (NLOS) propagation. Our software pipeline and integration to hardware, starting with the construction of context-specific RF propagation models to training and deploying AI/ML models to a payload, will be completed within a 16-hour window. To fulfill this stringent time constraint, we will leverage GPU-accelerated simulation software and deploy our models to RFSoC hardware such as Zynq Ultrascale+ for real-time processing. To establish accurate locks on RF sources, we will create an end-to-end pipeline that (1) generates synthetic RF training data, (2) develops AI algorithms for signal detection, geolocation, and identification, and (3) integrates algorithms with low-SWaP UAS payloads for real-time inference in pre-selected mission areas. At the end of Phase 2, we will deliver a functional prototype demonstrating effective RF geolocation and signal-type identification within complex, real-world propagation environments.