An efficiency algorithm on Gaussian mixture UKF for BDS/INS navigation system

Abstract

To further improve the performance of UKF (Unscented Kalman Filter) algorithm used in BDS/SINS (BeiDou Navigation Satellite System/Strap down Inertial Navigation System), an improved GM-UKF (Gaussian Mixture Unscented Kalman Filter) considering non-Gaussian distribution is discussed in this paper. This new algorithm using SVD (Singular Value Decomposition) is proposed to alternative covariance square root calculation in UKF sigma point production. And to end the rapidly increasing number of Gaussian distributions, PDF (Probability Density Function) re-approximation is conducted. In principle this efficiency algorithm proposed here can achieve higher computational speed compared with traditional GM-UKF. And simulation experiment result show that, compared with UKF and GM-UKF algorithm, new algorithm implemented in BDS/SINS tightly integrated navigation system is suitable for handling nonlinear/non-Gaussian integrated navigation position calculation, for its lower computational complexity with high accuracy. Keywords: Gaussian mixture, UKF, Singular Value Decomposition, Integrated navigation, Time complexity