Bio-inspired optical flow vision sensors for visual guidance of autonomous robots

Abstract

Visual perception -the ability to interpret the environment by sensing light -is an impressive ability of biological organisms.Vision allows animals and human beings to explore and to interact with their environment.It is a daily task biological organisms seem to perform with very little effort.Even smallest insects with a body size in the range of millimeters depend on vision as a main source of feedback for their behavior control which includes impressive flight and landing skills.Despite the efforts of many researchers in mimicking these skills in artificial systems, biological vision systems are still more powerful, versatile and efficient than any artificial systems developed so far.Nature has evolved integrated and compact solutions where the vision system is welladjusted to the properties of the sensory-motor control system and the physical body of the biological organism -properties that artificial systems still lack.Optical flow is a major visual sensor cue used by animals in navigating their environment.Optical flow is due to the relative motion of the vision system and the objects in its environment.It encodes information about the shape and position of objects as well as the self-motion of the animal.This information is used extensively in particular by insects such as flies and bees in flight control.Because of the impressive properties of biological systems, various groups have designed and implemented sensors in analog Very Large Scale Integrated (aVLSI) technology that mimic biological optical flow processing on various levels of abstraction.In parallel with recent efforts to construct miniature aerial platforms, optic flow information from camera sensors together with gyroscope outputs are used on such platforms to demonstrate properties such as autonomous flight.At the EPFL Laboratory of Intelligent Systems, Jean-Christophe Zufferey and others developed a 10-gram indoor airplane and demonstrated autonomous wall avoidance.The airplane uses an off-the-shelf one-dimensional image sensor from which optical flow caused by the relative motion of objects on the left and right side of the airplane is extracted.Optical flow is computed on a microcontroller that is also running the control ix x