Intelligent Tuning of Fuzzy Controllers by Learning and Optimization - Fuzzy Modeling and Control: Theory and Applications - page 128

Chapter 7
Intelligent Tuning of Fuzzy Controllers
by Learning and Optimization

Rodolfo Haber, Raúl Mario del Toro, Jorge Godoy and Agustín Gajate

7.1 Introduction

Fuzzy Logic Control (FLC) emerged as one of the most outstanding control
techniques in the middle of 80s. The great amount of literature on FLC that has
appeared is the main evidence of the increasing importance that fuzzy controllers
have been given in the control systemdesign field. Nowadays, there are several indus-
trial applications of fuzzy control systems in automotive, aeronautic and manufactur-
ing sectors, and home appliances for mentioning only some of them (Sugeno 1985 ;
Bonissone et al. 1995 ). However, the transfer of fuzzy control technology is often
limited due to inadequate and insufficient methods for tuning fuzzy control systems.
From academic viewpoint, since the pioneering works in the middle of 80s, a wide
range of techniques and methods have been proposed in the literature with dozens of
papers published in the field of tuning and self-tuning strategies for fuzzy controllers
(Maeda and Murakami 1992 ; Sugeno and Kang 1988 ; Bonissone et al. 1996 ).
The tuning of a fuzzy controller aims at achieving a suitable closed-loop response
according to certain design specifications that are usually dictated by the system
performance required. Last decades reflect the enormous effort of the scientific com-
munity towards clear and effective tuning strategies for fuzzy control systems. One
of the main design difficulties is the setting of initial fuzzy systems (membership
functions, number of rules, etc.). When some input-output data are available, the use
of neuro-fuzzy techniques and fuzzy clustering strategies becomes powerful compu-
tational tools to accomplish an appropriate design. Another crucial issue is the setting
of fuzzy control parameters such as scaling factors, which is not a trivial operation. If
a performance indices or figure of merits are well defined from a control and compu-
tational viewpoints, and a rough mathematical description of the process dynamics