Neural network controller with flexible structure based on feedback-error-learning approach

Mohammad Teshnehlab, Keigo Watanabe

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

In practice, the back-propagation algorithm often runs very slowly, and the question naturally arises as to whether there are necessarily intrinsic computation and difficulties with training neural networks, or better training algorithms might exist. Two important issues will be investigated in this framework. One establishes a flexible structure, to construct very simple neural network for multi-input/output systems. The other issue is how to obtain the learning algorithm to achieve good performance in the training phase. In this paper, the feedforward neural network with flexible bipolar sigmoid functions (FBSFs) are investigated to learn the inverse model of the system. The FBSF has changeable shape by changing the values of its parameter according to the desired trajectory or the teaching signal. The proposed neural network is trained to learn the inverse dynamic model by using back-propagation learning algorithms. In these learning algorithms, not only the connection weights but also the sigmoid function parameters (SFPs) are adjustable. The feedback-error-learning is used as a learning method for the feedforward controller. In this case, the output of a feedback controller is fed to the neural network model. The suggested method is applied to a two-link robotic manipulator control system which is configured as a direct controller for the system to demonstrate the capability of our scheme. Also, the advantages of the proposed structure over other traditional neural network structures are discussed.

Original languageEnglish
Pages (from-to)367-387
Number of pages21
JournalJournal of Intelligent and Robotic Systems: Theory and Applications
Volume15
Issue number4
Publication statusPublished - Apr 1996
Externally publishedYes

Fingerprint

Flexible structures
Neural networks
Feedback
Controllers
Learning algorithms
Backpropagation algorithms
Feedforward neural networks
Backpropagation
Manipulators
Dynamic models
Teaching
Robotics
Trajectories
Control systems

ASJC Scopus subject areas

  • Artificial Intelligence
  • Control and Systems Engineering

Cite this

Neural network controller with flexible structure based on feedback-error-learning approach. / Teshnehlab, Mohammad; Watanabe, Keigo.

In: Journal of Intelligent and Robotic Systems: Theory and Applications, Vol. 15, No. 4, 04.1996, p. 367-387.

Research output: Contribution to journalArticle

@article{ca421c0091ed4afdb94a8b8fa62834b3,
title = "Neural network controller with flexible structure based on feedback-error-learning approach",
abstract = "In practice, the back-propagation algorithm often runs very slowly, and the question naturally arises as to whether there are necessarily intrinsic computation and difficulties with training neural networks, or better training algorithms might exist. Two important issues will be investigated in this framework. One establishes a flexible structure, to construct very simple neural network for multi-input/output systems. The other issue is how to obtain the learning algorithm to achieve good performance in the training phase. In this paper, the feedforward neural network with flexible bipolar sigmoid functions (FBSFs) are investigated to learn the inverse model of the system. The FBSF has changeable shape by changing the values of its parameter according to the desired trajectory or the teaching signal. The proposed neural network is trained to learn the inverse dynamic model by using back-propagation learning algorithms. In these learning algorithms, not only the connection weights but also the sigmoid function parameters (SFPs) are adjustable. The feedback-error-learning is used as a learning method for the feedforward controller. In this case, the output of a feedback controller is fed to the neural network model. The suggested method is applied to a two-link robotic manipulator control system which is configured as a direct controller for the system to demonstrate the capability of our scheme. Also, the advantages of the proposed structure over other traditional neural network structures are discussed.",
author = "Mohammad Teshnehlab and Keigo Watanabe",
year = "1996",
month = "4",
language = "English",
volume = "15",
pages = "367--387",
journal = "Journal of Intelligent and Robotic Systems: Theory and Applications",
issn = "0921-0296",
publisher = "Springer Netherlands",
number = "4",

}

TY - JOUR

T1 - Neural network controller with flexible structure based on feedback-error-learning approach

AU - Teshnehlab, Mohammad

AU - Watanabe, Keigo

PY - 1996/4

Y1 - 1996/4

N2 - In practice, the back-propagation algorithm often runs very slowly, and the question naturally arises as to whether there are necessarily intrinsic computation and difficulties with training neural networks, or better training algorithms might exist. Two important issues will be investigated in this framework. One establishes a flexible structure, to construct very simple neural network for multi-input/output systems. The other issue is how to obtain the learning algorithm to achieve good performance in the training phase. In this paper, the feedforward neural network with flexible bipolar sigmoid functions (FBSFs) are investigated to learn the inverse model of the system. The FBSF has changeable shape by changing the values of its parameter according to the desired trajectory or the teaching signal. The proposed neural network is trained to learn the inverse dynamic model by using back-propagation learning algorithms. In these learning algorithms, not only the connection weights but also the sigmoid function parameters (SFPs) are adjustable. The feedback-error-learning is used as a learning method for the feedforward controller. In this case, the output of a feedback controller is fed to the neural network model. The suggested method is applied to a two-link robotic manipulator control system which is configured as a direct controller for the system to demonstrate the capability of our scheme. Also, the advantages of the proposed structure over other traditional neural network structures are discussed.

AB - In practice, the back-propagation algorithm often runs very slowly, and the question naturally arises as to whether there are necessarily intrinsic computation and difficulties with training neural networks, or better training algorithms might exist. Two important issues will be investigated in this framework. One establishes a flexible structure, to construct very simple neural network for multi-input/output systems. The other issue is how to obtain the learning algorithm to achieve good performance in the training phase. In this paper, the feedforward neural network with flexible bipolar sigmoid functions (FBSFs) are investigated to learn the inverse model of the system. The FBSF has changeable shape by changing the values of its parameter according to the desired trajectory or the teaching signal. The proposed neural network is trained to learn the inverse dynamic model by using back-propagation learning algorithms. In these learning algorithms, not only the connection weights but also the sigmoid function parameters (SFPs) are adjustable. The feedback-error-learning is used as a learning method for the feedforward controller. In this case, the output of a feedback controller is fed to the neural network model. The suggested method is applied to a two-link robotic manipulator control system which is configured as a direct controller for the system to demonstrate the capability of our scheme. Also, the advantages of the proposed structure over other traditional neural network structures are discussed.

UR - http://www.scopus.com/inward/record.url?scp=0030121150&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030121150&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:0030121150

VL - 15

SP - 367

EP - 387

JO - Journal of Intelligent and Robotic Systems: Theory and Applications

JF - Journal of Intelligent and Robotic Systems: Theory and Applications

SN - 0921-0296

IS - 4

ER -