Desensitized Cubature Kalman Filter with Uncertain Parameter

TL;DR

This paper introduces a robust desensitized cubature Kalman filter designed for nonlinear systems with uncertain parameters, improving state estimation accuracy by penalizing sensitivities.

Contribution

It proposes a novel DCKF method that incorporates sensitivity analysis and a desensitized cost function to enhance robustness against parameter uncertainties.

Findings

Effective in numerical simulations with uncertain parameters

Reduces estimation error due to parameter uncertainties

Demonstrates improved robustness over traditional CKF

Abstract

A robust desensitized cubature Kalman filtering (DCKF) for nonlinear systems with uncertain parameter is proposed. Sensitivity matrices are defined as the integral form, and desensitized cost function is designed by penalizing the posterior covariance trace by a sensitivity-weighting sum of the posteriori sensitivities. The DCKF gain is obtained by minimizing the desensitized cost function to amend the state estimation. Then, the sensitivity propagation of the state estimate errors is described, and the sensitivity of the root square matrix is obtained by solving a special equation. The effectiveness of the proposed DCKF was demonstrated by two numerical simulations with uncertain parameters.