Mobile robot navigation in unstructured environments using machine vision

Abstract

The aim of this investigation is to develop and evaluate a stereo vision system specifically for navigation and position estimation. The system must be capable of automatically modeling an environment and providing an accurate estimate of its own position within the environment. Specific areas in current vision algorithms are identified as candidates for improvement. The application of recent image processing tools such as wavelets and neuro-fuzzy logic to the proposed vision system is investigated. A combination of stereo matching techniques, photogrammetry and position estimation is used in an attempt to produce a system ideally suited to gathering detailed information about an unfamiliar environment for the purpose of navigating an autonomous mobile robot.