SELF-TUNING NEURAL CONTROLLER DESIGN WITH ORTHONORMAL ACTIVATION FUNCTIONS

Štefan Kozák

   This paper deals with the development of a class of neural networks, that have properties particularly attractive for identification and control of dynamic systems. These neural networks employ several types of orthogonal functions as neuron activation functions. It is shown that this modelling approach is beneficial for the robust identification and control context. Simulation results are provided to illustrate the usage and high performance of artificial neural networks in modelling and control of nonlinear dynamical systems.

Keywords: selftuning control, neural networks, identification, orthonormal series representation, orthonormal function model, robustness

[full-paper]