Home // COGNITIVE 2010, The Second International Conference on Advanced Cognitive Technologies and Applications // View article

Learning Approaches to Visual Control of Robotic Manipulators

Authors:
Paulo Goncalves
Pedro Torres

Keywords: Fuzzy Modeling, Neural Networks, Support Vector Machines, Computer Vision, Robotic Manipulators

Abstract:
This paper presents learning approaches to model the interaction between a robotic manipulator and its working environment. The approaches used are fuzzy modeling, neural networks and support vector machines. The interaction tackled in this paper is between the robot visual perception of the work environment and its actuators, while performing positioning tasks. This interaction, e.g., model, is obtaining only on measurements. This fact allows to obtain an uncalibrated model of the interaction, minimizing the setup time of the robotic system, not requiring calibrated robot kinematic and camera models. The input-output sample data used to learn the model are visual features from the work environment and the robot joint velocities, respectively. Experimental data, obtained from a IR52C robot and a visual stereo system, was used to validate the obtained models. Due to its accuracy and lower computational complexity, when compared to the other three, the off-line fuzzy model was used to control the robot, which clearly shows the effectiveness of the approach.

Pages: 103 to 108

Copyright: Copyright (c) IARIA, 2010

Publication date: November 21, 2010

Published in: conference

ISSN: 2308-4197

ISBN: 978-1-61208-108-3

Location: Lisbon, Portugal

Dates: from November 21, 2010 to November 26, 2010