Robust variational Bayesian method-based SINS/GPS integrated system

SINS/GPS integrated systems are influenced by non-Gaussian noise and unknown measurement noise due to exogenous disturbances and inaccurate noise statistics. To overcome this problem, a robust variational Bayesian method-based SINS/GPS integrated system is designed. First, the variational Bayesian-based Kalman filter is selected to estimate unknown measurement noise covariance. Second, the maximum correntropy criterion is introduced to the nonlinear robust filter to handle interference from non-Gaussian noise. Finally, the robust variational Bayesian method is designed based on the interacting multiple model, which not only fuses the variational Bayesian-based Kalman filter and the robust filter but also avoids non-Gaussian noise interference to the estimation result of measurement noise covariance. The robustness and adaptivity of the robust variational Bayesian method are verified by numerical simulation. Furthermore, the flight test results show improved performance of the SINS/GPS integrated system using the proposed method.

Automated summary

This paper proposes a robust variational Bayesian method-based SINS/GPS integrated system to overcome the influence of non-Gaussian noise and unknown measurement noise. The method estimates the unknown measurement noise covariance using a variational Bayesian-based Kalman filter, and handles interference from non-Gaussian noise using the maximum correntropy criterion. Additionally, the robust variational Bayesian method, based on the interacting multiple model, is designed to avoid interference from non-Gaussian noise to the estimation result of measurement noise covariance, and its robustness and adaptivity are verified through numerical simulation.