Reusable software neurofuzzy controller model

This paper introduces a method for designing a reusable software controller model. The proposed controller model is designed for a specific class of systems (e.g. zero order, fast order, second order, ...), and it can be reused to develop a specific controller to control any system within this specific class. The basic idea of this reusable controller model is to design a general adaptive control system in which knowing the plant parameters is not important, as long as they are within a prespecified class. The proposed controller model is composed of two submodel parts: (1) Model Identifier which is an on-line fuzzy model identifier represented as an adaptive fuzzy rule based system which takes a sample from the input and output signals, tries to approximate the plant model behavior, and, as long as it takes more samples, gives a more precise model for the plant; (2) Neurofuzzy Controller which is a fuzzy rule based system stored in a feedforward neural network. This feedforward network is trained using the information coming from the fuzzy model identifier, from the desired output, and from the actual output.