Abstract
The sigma-point Kalman filtering (SPKF) uses a set of sigma points to completely capture the first and second order moments of the apriori random variable. The sigma-points can be mapped into the state space or the measurement space through the nonlinear functions of the system directly, instead of linearization via the Jacobian matrices. Using the SPKF, a tightly-coupled GPS/INS integration can be designed in either the full state space or the error state space. This paper uses the error state space. It blends the INS error model and the nonlinear range and range-rate equations. The computational loads between SPKF and EKF are compared for the tightly integrated system through this design. Theoretical analysis and experiment results show that for GPS/INS integration the second-order effect on the accuracy of the EKF solution is so small that the EKF and SPKF give almost the same solutions in terms of accuracy.