A robotic aircraft that follows terrain using a neuromorphic eye

Abstract

Future Unmanned Air Vehicles (UAV) and Micro Air Vehicles (MAV) will fly in urban areas and very close to obstacles. We have built a miniature (35 cm, 0.840 kg) electrically-powered aircraft which uses a motion-sensing visual system to follow terrain and avoid obstacles. Signals from the 20-photoreceptor onboard eye are processed by 19 custom Elementary Motion Detection (EMD) circuits which are derived from those of the fly. Visual, inertial, and rotor RPM signals from the aircraft are acquired by a flight computer which runs the real-time Linux operating system. Vision-guided trajectories and landings were simulated and automatic terrain-following flights at 2 m/s were demonstrated with the aircraft tethered to a whirling-arm. This UAV project is at the intersection of neurobiology, robotics, and aerospace. It provides technologies for MAV operations.