Adaptive filtering and neural networks for realisation of internal model control

Abstract

The authors show that adaptive inverse control is a member of the class of control design techniques with an internal model control structure. By implication, therefore, adaptive inverse control is supported by the firm analytical foundation on which internal model control is now based. They present artificial neural network architectures for the implementation of nonlinear internal model control. This approach can be viewed as a nonlinear analogue of adaptive inverse control; the network models used are nothing more than nonlinear adaptive filters. The authors use two separate networks in the implementation of nonlinear IMC; one network models the plant, and the second network models the plant inverse. They conclude with a simulation example demonstrating nonlinear IMC using neural networks. >