A novel robust Kalman filter for SINS/GPS integration

Abstract

A robust filtering technique based on Student's t distribution is proposed for the characteristics that the traditional Kalman filtering algorithm cannot apply for measurement and process which with noise non-gaussian distribution. In this paper, A reasonable approach is introduced to construct a new Student's t-based hierarchical Gaussian state-space model and then using variational Bayesian approach to get the jointly estimated PDF of parameters in the constructed model. The proposed algorithm is verified mainly combined with SINS/GPS integrated navigation system. At last, the simulation results show that the proposed method can restrain the non-Gaussian noise in process and measurement well and improve the system precision.