Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Linear and Neuro Control Strategies; Some Experimental Results

Olsson, Henrik (2000) In MSc Theses
Department of Automatic Control
Abstract
The variety of different controllers is today very big. Much research is done in the field of neural network control and more well established technologies are sometimes looked at as old fashioned. The work is much based on the identification of a process model. In this report some different controllers have been built to see if neural networks are a good alternative to classic controllers. The classical controllers are a pure linear controller and an adaptive linear controller
working with updating by a covariance matrix. The neural network controller is based on a process model made by a recurrent neural network. This model is used in an input-output feedback linearisation controller.
The identification with the neural network model is... (More)
The variety of different controllers is today very big. Much research is done in the field of neural network control and more well established technologies are sometimes looked at as old fashioned. The work is much based on the identification of a process model. In this report some different controllers have been built to see if neural networks are a good alternative to classic controllers. The classical controllers are a pure linear controller and an adaptive linear controller
working with updating by a covariance matrix. The neural network controller is based on a process model made by a recurrent neural network. This model is used in an input-output feedback linearisation controller.
The identification with the neural network model is very difficult to make. As often in the world of control the theory by far exceeds reality. It is possible to build good neural network controllers but they are today more another way of achieving the same thing as the conventional controllers. (Less)
Please use this url to cite or link to this publication:
author
Olsson, Henrik
supervisor
organization
year
type
H3 - Professional qualifications (4 Years - )
subject
publication/series
MSc Theses
report number
TFRT-5643
ISSN
0280-5316
language
English
id
8848343
date added to LUP
2016-03-20 18:15:11
date last changed
2016-03-20 18:15:11
@misc{8848343,
  abstract     = {{The variety of different controllers is today very big. Much research is done in the field of neural network control and more well established technologies are sometimes looked at as old fashioned. The work is much based on the identification of a process model. In this report some different controllers have been built to see if neural networks are a good alternative to classic controllers. The classical controllers are a pure linear controller and an adaptive linear controller
working with updating by a covariance matrix. The neural network controller is based on a process model made by a recurrent neural network. This model is used in an input-output feedback linearisation controller.
The identification with the neural network model is very difficult to make. As often in the world of control the theory by far exceeds reality. It is possible to build good neural network controllers but they are today more another way of achieving the same thing as the conventional controllers.}},
  author       = {{Olsson, Henrik}},
  issn         = {{0280-5316}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{MSc Theses}},
  title        = {{Linear and Neuro Control Strategies; Some Experimental Results}},
  year         = {{2000}},
}