Visual Algorithms for Navigation and Guidance of UAVs with Autonomous Relational Decisions (VANGUARD)

Unmanned systems play a critical role in military operations across a wide range of missions. The DoD’s Unmanned Systems Integrated Roadmap identifies leader-follower tactics, swarming capabilities, sensor advancements, collision avoidance, and GPS-denied solutions as key technologies to support autonomy. Advances in these areas are needed to support coordinated multi-aircraft maneuvers and swarm tactics, and robust visual relative navigation algorithms are needed to enable the successful use of heterogeneous robotic swarms. We propose to develop a lightweight, affordable, all-passive optical system onboard UAVs to visually navigate and maneuver relative to other non-cooperative aircraft in real-time with Visual Algorithms for Navigation and Guidance of UAVs with Autonomous Relational Decisions (VANGUARD). VANGUARD enables tactics and capabilities that support new CONOPS such as airborne launch and recovery, enemy airspace penetration, and UAV swarming/formation flying in GPS denied environments without inter-aircraft communication links. VANGUARD uses a bio-inspired, neuromorphic, dynamic vision sensor with ultra-low response latency and optical flow capabilities, our advanced Object Detection Framework (ODF), and deep-learning driven aircraft pose estimation to rapidly identify and respond to subtle maneuvers by the aircraft being followed. Our GeMFIRE flight simulator provides a simulation test bed for closed-loop system performance evaluation.