Identification of state and measurement noise covariance matrices using nonlinear estimation framework

Abstract

The paper deals with identification of the noise covariance matrices affecting the linear system described by the state-space model. In particular, the stress is laid on the autocovariance least-squares method which belongs into to the class of the correlation methods. The autocovariance least-squares method is revised for a general linear stochastic dynamic system and is implemented within the publicly available MATLAB toolbox Nonlinear Estimation Framework. The toolbox then offers except of a large set of state estimation algorithms for prediction, filtering, and smoothing, the integrated easy-to-use method for the identification of the noise covariance matrices. The implemented method is tested by a thorough Monte-Carlo simulation for various user-defined options of the implemented method.