Comparing sensitivity and robustness of fuzzy and neuro-fuzzy controllers

The main goal of this paper is to investigate the reliability of fuzzy systems, analyzing the sensitivity of two different models: a conventional fuzzy controller and a neuro-fuzzy controller. The environment used was to guide a simulated robot through a virtual world populated with obstacles. In order to analyze the sensitivity we studied the robustness of the controllers when rules were removed from the system. The idea was to simulate situations when the designer does not have complete knowledge about the problem, or an out of order controller. The results obtained so far suggest a higher sensitivity of the hybrid system in a knowledge extraction process. Our experiments indicate that fuzzy systems can perform reasonably under adverse conditions.