Neural network-based adaptive robust control of a class of nonlinear systems in normal form

Neural networks (NNs) and adaptive robust control (ARC) design philosophy are integrated to design performance oriented control laws for a class of n-th order nonlinear systems in a normal form in the presence of both repeatable and non-repeatable uncertain nonlinearities. Unknown nonlinearities can exist in the input channel also. All unknown but repeatable nonlinearities are approximated by outputs of multi-layer NNs. A discontinuous projection method with fictitious bounds is used to tune NN weights online with no prior information for a controlled learning process. Robust terms are constructed to attenuate model uncertainties effectively for a guaranteed output tracking transient performance and a guaranteed final tracking accuracy. If the unknown nonlinear functions are in the functional ranges of NNs and the ideal weights fall within the prescribed range, asymptotic output backing is also achieved. Furthermore, by choosing the prescribed range appropriately, the controller may have a well-designed built-in anti-integration windup mechanism.