Wide-Field Integration Methods for Autonomous Navigation in 3-D Environments

Abstract

Wide-Field Integration (WFI) of optic ∞ow, a method for information extraction modeled on the spatial decompositions of specialized interneurons in the insect visuomotor system, can provide useful information about the proximity and relative speed of a vehicle with respect to objects in the environment. In this paper, WFI methods are extended to a 6-DOF UAV and used to demonstrate simultaneous stabilization of obstacle avoidance and terrain following with 2-D optic ∞ow. Measurements are taken in 3 planes (yaw, pitch, roll) of the spherical ∞ow fleld and decomposed spatially into Fourier harmonics. By writing a mathematical description of the 3-D environments that are likely to be encountered, the explicit linearized state data contained in each harmonic is found. By selecting harmonics whose information content is robust to changes in obstacle environment a control law was written that specifles how the optic ∞ow data is decomposed into input for each actuator. LQR techniques were used to produce feedback gains that were ultimately employed in an output feedback structure. Stability is proved via closed-loop eigenvalues and the UAV was simulated in an urban canyon-like environment. The aircraft was able to navigate the obstacle fleld, including a 90 bend, and maintain a flxed altitude above ground.