Gradient-Based Parameter Estimation in Pairwise Linear Gaussian System

This technical note addresses the gradient-based parameter estimation problem for pairwise linear Gaussian systems. The new adaptive filtering scheme is based on gradient-based optimization methods for estimating the uncertain system parameters and a robust square-root variant of the pairwise Kalman...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on automatic control Vol. 62; no. 3; pp. 1511 - 1517
Main Author: Kulikova, M. V.
Format: Journal Article
Language:English
Published: IEEE 01.03.2017
Subjects:
Adaptation models
Computational modeling
Hidden Markov models
Kalman filters
Mathematical model
Pairwise Kalman filter
Pairwise linear Gaussian systems
Parameter estimation
Robustness
square-root algorithms
ISSN:0018-9286, 1558-2523
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This technical note addresses the gradient-based parameter estimation problem for pairwise linear Gaussian systems. The new adaptive filtering scheme is based on gradient-based optimization methods for estimating the uncertain system parameters and a robust square-root variant of the pairwise Kalman filter used for estimating the unknown states of the pairwise linear Gaussian systems. Hence, in the adaptive filtering techniques, the dynamic state and system parameters are estimated simultaneously. The method of the filter sensitivities computation required in gradient evaluation is derived in terms of square-root factors of covariance matrices.
ISSN:0018-9286
1558-2523
DOI:10.1109/TAC.2016.2579745