Design of fuzzy logic controllers based on generalized T-operators (Q809958)

The aim of the paper is to investigate different triangular norms (t- and s-norms) as logical operators used in designing fuzzy controllers. The control rules studied there consist of two conditions (E and DE) and a single action (U) and are expressed as conditional statements ``if \(E_ i\) and \(DE_ i\) then \(U_ i'' \). All the variables in the rules are treated as fuzzy sets. The two basic steps in the design involve (i) aggregation of ruls with the aid of Cartesian product; at this stage a fuzzy relation R is formed: \(R(e,de,u)=^{N}_{i=1}R_ i(e,de,u)\) where T stands for t-norm; each \(R_ i\) is again a result of aggregation of \(E_ i\), \(DE_ i\), and \(U_ i\) with the use of t-norms (or a combination of t-norms and complements) (ii) inference of fuzzy control U computed for E and DE provided. This step is carried out by performing sup-t composition of the available fact and the fuzzy relation of the controller R, \[ U(u)=\sup_{e,de}[E(e)tDE(de)tR(e,de,u)]. \] By performing computer simulations different relationships between various t-norms and the performance of the controller (expressed in terms of characteristics of time response of the system under control) were revealed. It has been found that the t-norms usually significantly affect the performance of control, however no universal and define design suggestions could be derived.