Uncalibrated Fuzzy Visual Servo Control
Prof. Paulo Jorge Sequeira
Goncalves, Electrotechnical and Industrial Engineering Scientific Unit School of Technology -
Polytechnic Institute of Castelo Branco (Portugal)
DEI - Seminars Room
January 19th, 2012
at 4.00 p.m.
Visual Servoing approaches that can be applied to robot manipulators, are presented and compared, while the corresponding control laws are derived. The first visual control laws presented, describe the model of the interaction between the object (seen by the camera) and robot motions, i.e., visual motor interaction. Approaches where the visual motor interaction is estimated are also presented.
The first part presents Visual Servoing approaches, supported in screw theory, and the mathematical formulation of visual servoing will start from this background.
Within the Visual Servoing framework there are two main categories to obtain the Visual Motor Interaction: based on a model already known; based on an estimated model. The first category can be divided by how the image features are obtained: Image Based; Position Based; Hybrid.
The second category can be divided by how the model is estimated: analytically, using classical approaches like Broyden's update; by learning using fuzzy modeling or neural networks.
A brief review of the applications developed worldwide, will be discussed, both in industry and academia. Experimental results of Learning approaches implementation on real robotic manipulators are presented, allowing to identify real implementation issues regarding the visual frame rate, the error introduced by the visual data, and its implications to robot control.
Control, automation and measurement