A comparative analysis of two self-learning based strategies for fuzzy controller design

Two self-learning based methodologies for building the rule-base of a fuzzy logic controller (FLC) are presented in a comparative study. These methodologies aim to bring intelligent characteristics to controller design stage by simulating designer's actions as learning and adapting. An iterative learning algorithm, also called self-learning, is used to gather useful and trustful control data, which can be subsequently used to extract fuzzy rules. Both methods focus on the better use of data obtained by iterative self-learning algorithm. The methodologies have been proven reliable in different cases - practical applications and simulations. Although results are largely dependent on the actual implementation, satisfactory results were obtained without process identification or modeling. So far, no comparative studies were presented, as there are many parameters that may strongly influence the results. By intensively analyzing their reliability, both procedures can be further used on more complex or ill-defined processes.