Design of Noise Covariance Adaptive Federated Filter Based on Variational Bayesian Theory
Xiao Hua Ding, Xiuyun Meng, Shusen Zhang · 2023
In order to solve the problem that the measurement noise covariance of the navigation measurement components in the integrated navigation system of unmanned aerial vehicle (UAV) in close formation flight is time-varying or difficult to detect accurately, which leads to the decline of filtering accuracy, the zero-reset federated filter is adopted. Different filter calculation periods and fusion periods are designed to fuse the navigation information with unequal intervals. Based on the idea of variational Bayesian inference in the subfilter, the real posterior distribution is approximated by a simple distribution, and the unknown measurement noise covariance is estimated adaptively. The mathematical simulation shows that the algorithm can effectively improve the relative navigation accuracy of UAV formation flight, and can better adapt to the situation of measurement noise covariance mutation.