A Hinfinity Loop Shaping Framework for Bio-Inspired Sensorimotor Control

Abstract

The insect visuomotor system combines a lightweight and high bandwidth sensor with fast processing algorithms for efficient information extraction that enables autonomous navigation in complex, obstacle laden environments. In this dissertation, a H&infin loop shaping controller synthesis framework is introduced to couple the dynamic controller with an information extraction approach based on the processing of optic flow patterns by using wide-field motion-sensitive interneurons in the insect visuomotor system. Local proximity and velocity estimates are obtained with an optic flow model that is based on parameterization of typical three-dimensional urban environments. The insect inspired visual navigation technique developed in the dissertation combines optic flow outputs with a H&infin controller to provide robust stability in a cluttered environment while mitigating measurement noise and gusts. Simulation-based validation studies are undertaken and the loop shaping approach is used to overcome limitation in optic flow-based navigation for planar applications as well as demonstrate safe obstacle avoidance and terrain following behavior on an autonomous rotary wing miro-air-vehicle (MAV) for an urban-like environment subjected to gusts for both planar and 3D navigation applications. In addition, an alternate approach to the H&infin loop shaping framework is considered, based on using hair mechanosensory arrays in conjunction with optic flow outputs for enabling safe reflexive navigation. The hair sensor array outputs are combined with optic flow outputs within a biomimetic control framework and simulation-based studies are carried out to investigate their impact on the dynamics of a fixed wing MAV in an urban environment. The use of hair sensor arrays is found to augment stability and improve gust rejection performance resulting in safe obstacle avoidance behavior in the urban environment.