Neural Network Approaches For Feedback Linearization

Abstract

In this paper a recent approach is reported, which performs feedback linearization of uncertain
nonlinear systems using Artificial Neural Networks (ANNs). Also, a new ANN approach is presented,
which tackles a special case of feedback linearization using ANNs. Instead of using the ANN as an
estimator of the uncertain system dynamics the ANN is used as a compensator to the effects of the
model uncertainties, which appear in the linearizing control law. The updating of the neural
weights is carried out on-line using a conventional back propagation scheme, where the error to be
minimized is chosen such that it ensures the stability of the tracking error system. The proposed
method is tested on a well known nonlinear system and its application on a fermentation process is
reported.