Comparison of Sigma-point Update Framework in Cubature Kalman Filter for Tightly Coupled GNSS/INS

Abstract

New sigma-point update frameworks are investigated and compared in this paper, where the nonlinear approximation residuals represented by the posterior sigma points error are transformed directly to construct new sigma points. However, whether it is favorable to include likelihood function error in the update of sigma points for tightly coupled GNSS/INS is still need to be answered. The 3rd-degree and 5th-degree cubature rule for cubature Kalman filter (CKF) are adopted to verify the effectiveness of different sigma-point update algorithms. Numerical simulation indicates that the usage of sigma points error from prior probability density function (pdf) and likelihood function can speed up the convergence rate of CKF, and in terms of complexity and ease of use it is better to select the transformation based on prior pdf only.