Skip to Main content Skip to Navigation
Theses

Contributions à l'identification enesembliste ellipsoïdale

Abstract : This work deals with the membership set estimation techniques based upon an unknown but bounded error. Only the ellipsoidal approach has been considered here. In this particular context, a realistic evaluation of the bound based upon an analysis of the kurtosis number of the error sequence has been proposed. A unified approach of the different algorithms has been given, which leads to a numerically stable implementation. Note that nearly all published algorithms do not possess this fundamental property. The convergence analysis of these estimation techniques requires the well-known persistent excitation property. A algorithmic evaluation of this persistency hypothesis of the input has been proposed and a way to built such an input in this particular context is exhibited. Lastly, the compromise between a suboptimal sequential algorithm and an optimal global but numerically unrealistic has been studied ; the different results are illustrated thanks to some simulated data but also with some real data from industrial applications.
Document type :
Theses
Complete list of metadatas

Cited literature [49 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00168416
Contributor : Patricia Reynier <>
Submitted on : Friday, September 14, 2007 - 3:57:40 PM
Last modification on : Friday, November 6, 2020 - 3:49:26 AM
Long-term archiving on: : Friday, November 25, 2016 - 5:34:13 PM

Identifiers

  • HAL Id : tel-00168416, version 1

Collections

Citation

Khoi Quoc Tran Dinh. Contributions à l'identification enesembliste ellipsoïdale. Automatique / Robotique. Institut National Polytechnique de Grenoble - INPG, 2005. Français. ⟨tel-00168416⟩

Share

Metrics

Record views

532

Files downloads

497