Relative visual servoing: a comparison with hybrid intelligent models

摘要

In this paper, we compare four methods to perform eye-to-hand visual servoing using primitive tactile information. They are relative visual servoing, direct fuzzy servoing, fuzzy correction, and interpolation. These methods just need a monocular eye-to-hand camera and primitive tactile sensors. The fuzzy methods are tuned [1] by an Adaptive Neuro-Fuzzy Inference System (ANFIS). The set-up is part of a robotic platform named COERSU. Experimental results from COERSU are provided to validate these methods. An evolved visual perception system based on genetic algorithms was used to have the best possible combination of object segmentation and classification parameters [2]. As the experimental results show, relative visual servoing represents a significant improvement over the other methods in terms of reliability.