Moving base initial alignment method for improving robustness

The invention relates to a moving base initial alignment method for improving robustness, which is technically characterized by comprising the following steps: abstracting an initial alignment model; comparing the innovation covariance P(img file=' DDA0000469718250000011.TIF' wi='50'he='66'/) in the filtering algorithm with the calculated innovation covariance P(img file='DDA0000469718250000012.TIF' wi='53 he='50'/) to obtain the change condition of the model noise variance, and adaptively scaling the model noise w; and adaptively judging the threshold of adaptive matrix traces after each filtering according to the residual between the observed value and the filtered value so as to correct the state estimation. According to the method provided by the invention, the design is reasonable, and the UKF (unscented Kalman filter) algorithm is improved from two aspects of model noise adaptation and observation interference adaptation; on one hand, an aim of stabilizing the model noise by use of a scaling factor is achieved by monitoring the predicted variance in UKF; and on the other hand, a more accurate filtering result can be obtained by monitoring the traces of the adaptive matrix based on the innovation feature and performing real-time correction to achieve an aim of inhibiting the observation interference; and the method has relatively high robustness and can realize quick stabilization.