Seminar - Nelson Eduardo Castaño Giraldo11:00 am | POLYTECH ANGERS |Room 18 | 62, avenue Notre-Dame du Lac | ANGERS

The November 17, 2022

Subject : Identification of upper bound covariance matrices for min-sup  Robust Kalman Filter

Abstract

The estimation of the noise covariance matrices (CMs) of the systems that lead to the random behavior of the state space (SS) models is a fundamental task for the optimal estimation of the state. The robust Kalman filter (RKF) of the min-sup type introduced in Azhmyakov [2002] guarantees robust estimation in uncertain linear dynamic systems. In particular, the joint observation-state noise is assumed to have some unknown joint probability distribution functions (pdf) of some class of centered distributions with covariances bounded by some known positive-definite matrix S which is used for estimating the states of the  system in the worst case sense. In this study, an adequate estimation of the matrix S is developed from the available observations of the system. Its performance compares with other recent robust techniques based on covariance least squares (RALS).