Initialization of SINS/GNSS Error Covariance Matrix Based on Error States Correlation

The traditional Strapdown Inertial Navigation System (SINS)/Global Satellite (GNSS) integrated system uses standard Kalman Filter (KF) to estimate the error states, which weakens correlation between different components directly initialed Error Covariance Matrix (ECM) into diagonalization. This method is also widely applied in State Transformation Extended (STEKF) and Invariant (IEKF), results state estimation failing achieve optimal performance. To solve this problem, paper first analyses transformed relationship from linear nonlinear redefined STEKF IEKF, strong found components, namely ECM or IEKF no longer appears as a diagonal matrix. Then, aiming at models of are derived respectively, establishes theoretical basis for initialization based on states correlation. Finally, accuracy, feasibility general applicability proposed verified by boat-mounted field trial.