Application of Unscented Kalman Filter with Neural Network in the Polar Rapid Transfer Alignment

Due to the shortage of natural resources and the development of science and technology, the Arctic has been a strategic area for resource exploitation and military expansion. The grid inertial navigation system (INS) becomes a superior device to guarantee a safe voyage, and the polar rapid transfer alignment is pre-requisition for the initialization of INS. However, the polar harsh marine environment enlarges the observation noise, significantly impairing the nonlinear Kalman filter performance. The BP-assisted UKF is proposed here to aid polar rapid transfer alignment to anti-interfere the effect of marine noise. Initially, the mathematical model of the polar rapid transfer alignment was derived. Secondly, the learning sample set was modified to adapt to derived model. In the end, the semi-physical experiment was designed to validate the designed network and the proposed algorithm. The results demonstrated that the BP-assisted UKF could effectively estimate the misalignment under the condition of enlarged measurement noise.