Title: A new neural networks based adaptive model predictive control for unknown multiple variable non-linear systems

Authors: Daohang Sha

Addresses: Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, 525 Blockley Hall, 423 Guardian Drive, Philadelphia, PA 19104, USA

Abstract: In this paper, a novel adaptive model predictive control (AMPC) based on neural networks for unknown MIMO non-linear systems was proposed. Firstly, a new recursive second order online learning algorithm with a forgetting factor was developed for the training of the neural network model which is used to identify the unknown non-linear system. Secondly, a new cost function based on the system output error and control inputs with a forgetting factor was adopted. Thus, a new adaptive MPC control law was obtained based on the extreme value theory. The effectiveness of the proposed algorithm was demonstrated by simulation experiments to a two-input and two-output non-linear system. The algorithms both for identification and control were implemented in SIMULINK, which are ready for the real-time control. The results achieved confirmed the validity and suitability of the algorithms proposed in the paper.

Keywords: neural networks; NN; model predictive control; adaptive MPC; multiple variable systems; adaptive control; nonlinear control; identification; simulation.

DOI: 10.1504/IJAMECHS.2008.022013

International Journal of Advanced Mechatronic Systems, 2008 Vol.1 No.2, pp.146 - 155

Published online: 16 Dec 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article