Conference Paper
To read the full-text of this research, you can request a copy directly from the authors.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

Full-text available
The primary goal of this dissertation is to use machine learning and pattern recognition techniques to generate robot behaviour. This thesis presents different contributions in the fields of image processing for robots and self-localization. Robot localization is a key challenge in making truly autonomous robots. In order to localize itself, a robot has access to two sources of information: odometry and measurements. Odometry works by integrating incremental information relative to the motion of the robot over time. Measurements or observations pro- vide information about the location of the robot. Measurements come from the sensors the robot is fitted with and the most common sensors are visual cameras. Therefore, images are the most common measurements and information about the environment can be provided through image processing techniques. In this thesis we discuss the use of machine learning techniques in the development of image processing and localization algorithms. Regarding image processing, we propose genetic algorithms for real-time object detection as well as several classifiers to estimate the quality of the images. The localization problem is tackled in this thesis using two approaches: topological localization and visual place classification. Topological localization assumes that an environment map is provided by an external source and our proposals explore the use of the quality of the images for integrating odometry and observations. Visual place classification is the problem of classifying images depending on semantic areas or rooms. The proposals in this thesis for solving visual place classification are based on the use of clustering techniques and support vector machines. The image processing for these proposals consists of using invariant features and image descriptors.
ResearchGate has not been able to resolve any references for this publication.