Functional Identification and Nonlinear Control via a Perceptron Network

Tracking control of a general class of nonlinear systems using a Perceptron Neural Network (PNN) is presented. The basic structure of the PNN along with the conditions for its exponential convergence under a suitable training law are first derived. A novel discrete-time control strategy is introduced that employs the PNN for direct on-line estimation of the feedforward control input. A Lie-algebraic formalism is used to compute the gradient information demanded by the network´s training law. Unlike most of the existing direct adaptive or learning schemes, the nonlinear plant is not assumed to be feedback linearisable. An application of the developed controller to the navigation a ground vehicle, which is a nonlinear nonholonomic system, is also presented.