Comparing Robustness and Performance of Hybrid and Non-hybrid Fuzzy Controllers aimed to guide a simulated robot

In this paper we investigate the reliability of fuzzy system by an analysis of sensitivity using two different models: Mamdani-type controller and a Takagi-Sugeno-type controller. The main idea concerns the control of a simulated robot walking thought a virtual environment populated with obstacles. We remove some rules from the controllers in order to analyze the sensitivity and investigate the robustness of them. We are interested in situations where the system designer does not have all knowledge concerning the problem or an out of order controller. We measured the success of the controllers -to reach its goal -in order to compare the hybrid and the non-hybrid models. Our results suggest a higher sensitivity of the Takagi-Sugeno system in a knowledge process extraction and they indicate that fuzzy systems can perform properly even under adverse conditions.