Vision Based Mobile Robotics: Mobile Robot Localization using vision sensors and active probabilistic approaches

The use of vision in mobile robotics in one of the main goal of this thesis. In particular novel appearance based approaches for image matching metric are introduced. These approaches are applied to the problem of mobile robot localization. Similarity measures between robot’s views are used in probabilistic methods for robot pose estimation. In this field of probabilistic localization active approach are proposed allowing the robot to faster and better localize. All methods have been extensively tested using a real robot in an indoor environment.