Summary: II MODEL AND FEEDBACK LINEARIZING CONTROLLER 1
A Multilayer Perceptron Replaces a Feedback
Linearization Controller in a Nonlinear
Jos'e F. Haffner, Ney T. Meyrer, Jos'e N. Amaral and Lu'is F. A. Pereira
Abstract--- A Feedback Linearizing Controller (FLC) is
used to train a multilayer perceptron control a DC motor.
After training, the multilayer perceptron replaces the FLC
and yields significantly better performance in the presence
of state≠measurement noise, load disturbances and param≠
eter variations. Simulation results also indicate that the
neural network based controller is better able to cope with
input saturation resulting from an overly demanding refer≠
ence model specification.
Keywords--- Multilayer Perceptron, Neural Network Con≠
trol, DC motor control, Feedback Linearization.
N ONLINEAR systems whose dynamics are described
by smooth functions can be well controlled using in≠
put/output feedback linearizing control techniques. These