Iterative learning fuzzy inference system

This paper presents a learning fuzzy controller which can adapt with changing performance requirements. During the past decade we have witnessed a rapid growth in the number and variety of applications of fuzzy logic ranging from consumer electronics and industrial process control to decision support system and financial systems. The fuzzy controller designer faces the challenge of choosing the appropriate membership functions, minimum rule base and the most suitable fuzzifier and defuzzifier. Having made these choices, the fuzzy controller has to be tuned to deliver the desired response. Multiple simultaneous adjustments (rules, membership functions and gains) make the optimum tuning even more difficult. It is now realized that complex real world problems require intelligent systems that combine knowledge, techniques and methodologies from various sources. These intelligent systems are supposed to possess humanlike expertise within a specific domain, adapt themselves and learn to do better in changing environment. In this paper we combine fuzzy techniques with iterative learning to formulate a scheme that can automatically find the appropriate fuzzy controller to meet our design requirements. The scheme is adaptive and can handle the uncertainties arisen by difference in perception about a concept. Extensive literature survey shows that designing fuzzy controllers with desired performance specifications is not a trivial task. Even the specification of linguistic variables, key concept in fuzzy system design, can be different from different experts. This scheme tries to fill this gap. The results show that the scheme is robust, cost effective and relatively simple to implement. It makes use of the non linearity inherent in the fuzzy systems. This scheme has the potential to make consumer electronics, decision support systems and all other countless number of areas where fuzzy has made in roads, perform better.