An adaptive split and merge unscented Gaussian sum filter for initial alignment of SINS

摘要

In order to improve the performance of the unscented Kalman filter with uncertain or time-varying noise statistic, a novel adaptive split and merge unscented Gaussian sum filter is proposed for the initial alignment on the swaying base. The novel filter makes use of the output measurement information to online update the covariance of the process noise. A split technique is used to estimate the mean of the process noise. The updated mean and covariance are further feed back into the unscented Gaussian sum filter. The simulation results demonstrate that the novel filter is superior to the unscented Kalman filter.