On Computational Intelligence Tools for Vision Based Navigation of Mobile Robots

Abstract

This PhD Thesis has dealt with two types of sensors, studying their usefulness for the development of navigation systems: optical cameras and 3D range finder cameras. Both were used to attack the problems of selflocalization, egomotion and topological and metric map building. Good results have been obtained using Lattice Computing, Evolution Strategies and Artificial Neural Network techniques. Lattice Computing has been applied to self-localization in qualitative maps and to visual landmark detection using optical cameras. For metric localization tasks with 3D range finder cameras, Competitive Neural Networks and Evolution Strategies have been used to estimate the transformations between 3D views, which will provide the estimation of the robot's movement. Finally, we have performed a proof-of-concept physical experiment on the control of Linked Multi-Component Robotic Systems with visual feedback.

Date of Award	2009
Original language	English
Awarding Institution	Universidad del País Vasco (UPV/EHU)