Precomputable Kalman-based filter for Markov Jump Linear Systems

Eduardo F. Costa; Benoite de Saporta

doi:10.1109/SYSTOL.2016.7739782

Retour

Acte de colloque

Precomputable Kalman-based filter for Markov Jump Linear Systems

Eduardo F. Costa et Benoite de Saporta

Conference on Control and Fault-Tolerant Systems (Print), Vol.2016-, pp.393-398

09/2016

DOI: https://doi.org/10.1109/SYSTOL.2016.7739782

Résumé

Convergence

Fault tolerance

Linear systems

Markov processes

Quantization (signal)

Symmetric matrices

Trajectory

This paper presents a new efficient numerical method for state estimation of Markov Jump Linear Systems (MJLSs). It is based on the selection of a finite set of typical trajectories of an underlying piecewise deterministic Markov process (PDMP) related to the gain matrices of the optimal Kalman-Bucy filter and allows for pre-computations. The trajectories are optimally selected using quantization of the post jump locations of the PDMP. The performance of this approach is evaluated both theoretically and on a numerical example of a magnetic suspension system subject to failures.

Indicateurs

1 Consultations de la notice

Détails

Titre: Precomputable Kalman-based filter for Markov Jump Linear Systems
Créateurs - sans rôle: Eduardo F. Costa - Univ. São Paulo - Instituto de Ciencias Matemáticas e de Computacao, C.P. 668, 13560-970, São Carlos, SP, Brazil
Benoite de Saporta - University of Montpellier, F-34095 Montpellier, France, CNRS, IMAG, UMR 5149, F-34095 Montpellier, France and Inria Bordeaux Sud Ouest, team CQFD, F-33400 Talence, France
Détails de publication: Conference on Control and Fault-Tolerant Systems (Print), Vol.2016-, pp.393-398
Éditeur: IEEE
Nombre de pages: 6
Identifiants: 99139978309311
Unité académique: Institut Montpelliérain Alexander Grothendieck - IMAG
Langue: English
Type de ressource: Conference proceeding