What kind of hardware is necessary for a fuzzy rule based system?

For a fast evaluation of fuzzy rule based systems (FRBS's) some special fuzzy hardware is available, but in most of the realized applications only standard hardware or specific non fuzzy ASICs are used. In this paper we discuss some of the reasons of this gap and give hints for a fast realization of FRBS's especially on digital standard processors. For the realiza tion a fast algorithm with an optimized calculation of the antecedent of the fuzzy rules is introduced which minimises the number of operations. This is possible because the FRBS is an universal approximator. Furthermore some implementation possibilities are discussed depending on their complexity. II. BASIC TERMS OF FUZZY RULE BASED SYSTEMS As we have mentioned above it is proved that FRBS's are universal approximators [5,6,7], that is: For each a,b with a<b let a,b: be a membership function such that a,b(x) 0 if x [a,b]. Let furthermore T1 and T2 be t-norms and I a fuzzy implication verify ing I(a,0) = 0 if a 0 (for example a Rimplication or a t-norm implication I(a,b)=T 3(a,b) ). Let G be a t-conorm and S a defuzzification method.