Adaptively tuning sampling weights of the unscented Kalman filter in starlight refraction navigation

The objective of this study was to improve the performance of starlight refraction navigation using the unscented Kalman filter. In previous studies, the uncertainty and nonlinearity of the starlight refraction model limited the advancement of navigation accuracy. To overcome this limitation, an adaptive unscented Kalman filter is proposed in which the sampling weights are selected as the adaptive parameters. During the filtering process, the sampling weights are adaptively tuned based on the gradient descent method to increase the accuracy of the unscented transform predicted value. Simulations were conducted to verify the proposed filter. Results showed that it notably improved the navigation accuracy, and the position error was reduced to 68.7 m.