Model reference adaptive control of nonlinear dynamical systems using multilayer neural networks

Abstract
A multilayer discrete-time neural net (NN) controller is presented for the model reference adaptive control of a class of MIMO dynamical systems. No initial learning phase is needed and the tracking error between the output of the nonlinear plant and a linear model converges within a very short time. This weight tuning paradigm is based on the well-known delta rule but includes a modification to the learning rate parameter plus a correction term. It guarantees tracking as well as bounded NN weights in non-ideal situations, so that a persistency of excitation condition on the internal signals is not needed. Simulation results are presented in order to verify the theoretical conclusions.<>