Design of Transfer Alignment Optimization Algorithm Based on Modified Adaptive Kalman Filter

The transfer alignment performance of ship strapdown inertial navigation is easily affected by the system error model and filtering algorithm. This paper firstly proposes an integrated error model of dynamic deflection and lever effect in which the rigid lever acceleration is compensated, and the random process error model is extended to the filter state variables for preferable estimation. Secondly, a simplified Sage-Husa adaptive Kalman filter algorithm is modified through decomposing the state estimation covariance matrix by UD. On the premise of ensuring the semi-positive quality of the noise estimator, the calculation error during filtering iteration is reduced and the numerical stability is further improved. In addition, the divergent filtering suppression method based on covariance matching is introduced to improve the timeliness of filtering results. Simulation results show that when the prior statistical characteristics of system noise are difficult to accurately determine, the modified adaptive filtering algorithm can maintain the numerical stability and realize the adaptive estimation of the noise. The installation misalignment angle can be estimated more quickly and effectively, and the fast transfer alignment of SINS can be achieved.