Abstract :
A design method of adaptive control of processes, which include nonlinear-parametrized models such as three-layered neural networks or wavelet networks, is proposed. That control scheme is derived as a solution of a particular nonlinear H/sub /spl infin// control problem, and is shown to be robust to various errors contained in the nonlinear-parametrized models, and the stability of overall system is assured. Additionally, in the resulting control schemes, the control errors can be made arbitrarily small, even if the nonlinear-parametrized models are not good approximators.
