Please use this identifier to cite or link to this item: http://repositorio.unicamp.br/jspui/handle/REPOSIP/92773
Type: Artigo de periódico
Title: Convergent Lmi Relaxations For Quadratic Stabilizability And H∞ Control Of Takagi-sugeno Fuzzy Systems
Author: Montagner V.F.
Oliveira R.C.L.F.
Peres P.L.D.
Abstract: This paper investigates the quadratic stabilizability of Takagi-Sugeno (T-S) fuzzy systems by means of parallel distributed state feedback compensators. Using Finsler's lemma, a new design condition assuring the existence of such a controller is formulated as a parameter-dependent linear matrix inequality (LMI) with extra matrix variables and parameters in the unit simplex. Algebraic properties of the system parameters and recent results of positive polynomials are used to construct LMI relaxations that, differently from most relaxations in the literature, provide certificates of convergence to solve the control design problem. Due to the degrees of freedom obtained with the extra variables, the conditions presented in this paper are an improvement over earlier results based only on Pólya's theorem and can be viewed as an alternative to the use of techniques based on the relaxation of quadratic forms. An extension to cope with guaranteed ℋ∞ attenuation levels is also given, with proof of asymptotic convergence to the global optimal controller under quadratic stability. The efficiency of the proposed approach in terms of precision and computational effort is demonstrated by means of numerical comparisons with other methods from the literature. © 2009 IEEE.
Editor: 
Rights: fechado
Identifier DOI: 10.1109/TFUZZ.2009.2016552
Address: http://www.scopus.com/inward/record.url?eid=2-s2.0-68849121488&partnerID=40&md5=930f3317bb8d7cce1f5fe9b05f0b9f5e
Date Issue: 2009
Appears in Collections:Unicamp - Artigos e Outros Documentos

Files in This Item:
File Description SizeFormat 
2-s2.0-68849121488.pdf408.95 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.