Credible Gaussian sum cubature Kalman filter based on non-Gaussian characteristic analysis

Abstract

Toward the pose estimation problem of unmanned aerial vehicles (UAVs) flight over water, a credible Gaussian-sum cubature Kalman filter (CGSCKF) based on non-Gaussian characteristic analysis for nonlinear non-Gaussian system is proposed. In the framework of the proposed scheme, a data characteristic analysis based on multi-peak distribution degree, skewness and kurtosis is applied to evaluate the non-Gaussian intensity of noise, and a Gaussian mixture reduction (GMR) method based on the improved monarch butterfly optimization (IMBO) is proposed to improve the filtering accuracy in the case of weak non-Gaussian noise (Gaussian-like noise). In addition, considering the problem of inconsistency between the practical and the theoretic system model, the process noise and measurement noise are estimated online, and the credibility of filtering result, which is obtained by Gaussian-sum cubature Kalman filter (GSCKF), is measured by comparing the filtering error covariance with the true error covariance. Some comparison simulations and experiments with classical algorithms are provided, justifying that the proposed scheme has better performance in the pose estimation of UAV.