Efficient analog CMOS implementation of fuzzy rules by direct synthesis of multidimensional fuzzy subspaces

A novel way of evaluating the condition part of fuzzy rules is introduced. It requires only operations that can be efficiently performed with analog circuits. The degree of membership of an input vector to some fuzzy subspace is defined using a measure of distance between the input vector and some central point in the subspace. Since this approach avoids implementing operators like MIN or MAX, very dense analog hardware may be realized. A class of analog building blocks is presented, which allows a modular synthesis of arbitrary types of fuzzy rules. A circuit based on these blocks was integrated and tested successfully