Novel internal units for a neural network based adaptive fuzzy inference systems

Novel internal computing units are discussed for carrying out the 'OR' operation in fuzzy logic based controllers. A simplified form of the Tageki-Sugeno controller, implemented by a form of (generalised) radial basis function neural network, is used throughout. As well as extending the scope of possible rule bases, a development of the new type of node is shown to provide an alternative to conventional defuzzification methods.