Article ID: 2026EAL2007
To address the inaccurate process noise statistics caused by unknown disturbances in satellite relative navigation, this work proposes a novel adaptive Kalman filter. The method is developed based on the linearized Clohessy-Wiltshire equations and derives an analytical relationship between the process noise covariance and the power spectral density, enabling online covariance reconstruction. Compared with conventional adaptive approaches, the proposed structured noise modeling significantly reduces the number of estimated parameters, mitigates reliance on prior noise statistics, and incurs no additional computational burden. Numerical simulation results demonstrate that the novel filter achieves excellent estimation accuracy, stability, and adaptability.