Free-model based neural network for dynamic systems

An approach to identify and control an unknown nonlinear dynamic system using the free-model based neural network is discussed. The free model for an unknown system is further derived and approximation errors between real and approximated outputs are briefly discussed. Applying the free-model preprocessed inputs to the neural network, the free-model based neural network is trained to approximate the output for the unknown system. Since the free-model input preprocessing can be viewed as a high pass filter, the free-model based neural network converges faster than the conventional neural network representation. The free-model based neural networks are implemented in a model reference adaptive inverse control scheme which is applied to a nonlinear plant with satisfactory performance