Adaptation and learning factor for neural controllers

The definition of a passive state for a system composed of a plant and a neural network controller (NNC), using a multi-layer artificial neural network, is presented. The term transition time is defined using the concept of the passive state. The experimental results obtained for the on-line determination of the adaptation and learning factors for a direct and indirect neural controller are presented. The transition time versus the adaptation and learning factors curves are used to study these factors. An heuristic algorithm, using this curves, for on line determination of the values of learning and adaptation factors is proposed.