Stability of a neural predictive controller scheme on a neural model

In recent papers [4],[7],[8],[11], [12],[14] different forms of neural network based predictive controllers have been proposed.The main emphasis in these papers is on the implementation aspects of the controller, i.e. the development of a robust optimization algorithm for the controller, which will be able to perform in real time. Rowever, the stability issue has not been addressed specifically for these controllers. On the other hand a number of results concerning the stability of receding horizon controllers on a non-linear system exist 121, (lo] and 191. In this paper we present a proof of stability for a predictive controller controlling a newal network model. The resulting controller is tested on a non-linear pneumatic servo system.