INSTITUTE OF INFORMATION TECHNOLOGIES - BAS

Cybernetics and Information Technologies
Volume 3, No 2. Sofia, 2003, Bulgarian Academy of Sciences


Model Reference Adaptive Neural Control of a Variable Structure System

Moises Bonilla*, Ieroham Baruch*, Jose Martin Flores*, Maria M. Goire*, Boyka Nenkova**

* Department of Automatic Control, CINVESTAV-IPN, 07360 Mexico D.F., MEXICO
** Institute of Information Technologies, 1113 Sofia


Abstract: The aim of this paper is to propose a model reference adaptive neural control of a variable structure plant, described by an implicit realization with variable order and parameters, using only output feedback. The neural control scheme proposed is composed by two recurrent neural networks, named: neuro-identifier and neuro-controller. A variable structure plant model together with the realized adaptive neural control are simulated by means of the MatLab-Simulink and the obtained simulation results are compared with those obtained by the use of an ideal implicit control, applying the true descriptor variable. The simulation results show a great similarity of the obtained graphics for both control schemes, which demonstrated the applicability of the proposed adaptive neural control.

Keywords: Model reference adaptive control, variable structure plant, recurrent neural networks, backpropagation learning.