Abstract：Aiming at the linear Gaussian state model of integrated navigation system with inaccurate system noise and measurement noise statistics, this paper proposes a modified variational Bayesian adaptive filter(MVBAKF) for integrated navigation system.Firstly, the Wishart distribution is selected as the conjugate priors of Gaussian covariance matrix with known mean value, and the joint probability distribution function of measuring noise variance, state vector and prediction error covariance matrix is given. Then, the formula of measurement noise variance and state vector prediction error covariance matrix is given by using the variable decibels method, and then the MVBAKF algorithm with iterative characteristics is proposed. Finally, the simulation experiment of GNSS/SINS integrated navigation system based on MVBAKF algorithm is carried out. Experimental results show that MVBAKF algorithm can accurately estimate the variance of measurement noise, and can effectively overcome the influence of inaccurate system noise statistics on the filtering accuracy compared with traditional VBAKF algorithm, and thus improve the filtering accuracy of integrated navigation system.
