Přístupnostní navigace
E-application
Search Search Close
Publication detail
DOKOUPIL, J.
Original Title
Adaptivní regulátory s principy umělé inteligence a jejich porovnání s klasickými přístupy
English Title
Adaptive controllers with principles of artificial inteligence and its comparison with classical identifications methods
Type
conference paper
Language
Czech
Original Abstract
Článek se se zaměřuje na komparaci rekurzivních metod identifikací postavených na bázi koncepčních metod nejmenších čtverců a umělé inteligence.
English abstract
Presented article is focused on on-line parametric identification methods based on least-squares algorithms and neural networks. The strategy for preventing the problem known as estimator windup which is typical of exponential forgetting algorithm is solved here as well as the observance of positive definition of information matrix. Some ways how to improve accuracy of both methods are also proposed. These properties make them suitable for adaptive control, but a little supremacy is still on the side of methods based on least-squares algorithms.
Key words in English
adaptive system identification, least-squares, artificial intelligence, restricted forgetting, U-D decomposition, Cholesky decomposition, RIV, RPLR, RPEM
Authors
Released
23. 4. 2009
Publisher
NOVPRESS s.r.o., nám. Republiky 15, 614 00 Brno
ISBN
978-80-214-3870-5
Book
Proceedings of the 14th conference, STUDENT EEICT 2008, Volume 3
Edition number
1
Pages from
98
Pages to
100
Pages count
3
BibTex
@inproceedings{BUT31287, author="Jakub {Dokoupil}", title="Adaptivní regulátory s principy umělé inteligence a jejich porovnání s klasickými přístupy", booktitle="Proceedings of the 14th conference, STUDENT EEICT 2008, Volume 3", year="2009", number="1", pages="98--100", publisher="NOVPRESS s.r.o., nám. Republiky 15, 614 00 Brno", isbn="978-80-214-3870-5" }