Title: Perception neural network versus fuzzy neural network for controlling the inverted pendulum

Authors: Mohamed A. Belal

Addresses: Faculty of Computers and Information, Helwan University, Ein-Helwan, Cairo, Egypt

Abstract: Fuzzy neural networks (FNN) and other soft computing approaches provide robust and flexible solutions to the design of the industrial control systems. Their main feature is that they do not require a complete mathematical modelling of the controlled planet. However, there is still a problem in guaranteeing the convergence of the learning process as well as the reliability of the proposed solution when it is adapted to real operational situations. The advantages of the perception neural network (PNN) are its learning process which is guaranteed to converge. Besides, it provides multi-resolution approximation of the required function. Based on the sufficiency of the training data, the structure of the PNN is adjusted to represent each non-linear portion in the function surface. In this paper, we examine the capability of the PNN versus FNN in controlling the inverted pendulum. In comparing the performance of the two networks for controlling the inverted pendulum, results showed the accuracy and superiority of PNN over the FNN.

Keywords: fuzzy neural networks; FNN; inverted pendulum control; perception neural networks; learning.

DOI: 10.1504/IJAMECHS.2010.033044

International Journal of Advanced Mechatronic Systems, 2010 Vol.2 No.3, pp.192 - 203

Published online: 07 May 2010 *

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