UAV and insect vision by Thomas Netter, Aerospace and Robotics engineer [read resumé] A robot that flies with a neuromorphic eye

Download IROS'02 paper [A4], [US letter] PDF 212Kb
Watch it avoid obstacles: [AVI Video 6 Mbytes]
(Linux users: Codec = Radius Cinepak, viewable with xanim, xine, mplayer)

[Lire en français]

Flying insects rely on their panoramic eye and Optic Flow for guidance and obstacle avoidance. Furthermore, flies bear remarkable neural fusing of visual, inertial, and aerodynamic senses. Behavioural and neurophysiological studies of insects have inspired research in Artificial Intelligence and the construction of many autonomous robots. This flying robot was built within the Neurocybernetics Group of the Laboratory of Neurobiology, CNRS, Marseilles, France. Previous research on the visuo-motor system of the fly had led to the development of two mobile robots which feature an analogue electronic vision system based on Elementary Motion Detectors (EMD) derived from those of the fly.

A tethered Unmanned Air Vehicle (UAV), called Fania, was developed to study Nap-of-the-Earth (NOE) flight (terrain following) and obstacle avoidance using a motion sensing visual system. After an aerodynamic study, Fania was custom-built as a miniature (35 cm, 0.840 kg), electrically-powered, thrust-vectoring rotorcraft. It is constrained by a whirling-arm to 3 degrees of freedom with pitch and thrust control.

The robotic aircraft's 20-photoreceptor onboard eye senses moving contrasts with 19 ground-based neuromorphic EMDs. Visual, inertial, and tachymeter signals from the aircraft are scanned by a data acquisition board in the flight control computer which runs the Real-Time Linux operating system. A weighted average fusion of the visual inputs is used to command thrust. A PID controller regulates pitch. Flight commands are output via the parallel port to a microcontroller interfacing with a standard radio-control model transmitter.

Vision-based terrain following (a) and landing (b) were simulated. Automatic obstacle-avoiding flights at speeds between 2 m/s and 3 m/s were demonstrated within the laboratory. This UAV project is at the intersection of Neurobiology, Robotics, and Aerospace. It provides principles and technology to assist urban operations of Micro Air Vehicles (MAV).

Related projects for flight with insect vision

Insect vision and motion detection at Institute of Neuroinformatics