Geometrically-Based Adaptive Kalman Filtering for Attitude Estimation

Accurate and reliable attitude estimation is essential for precise motion control in micro-/nano-satellites. This paper presents a geometrically intuitive adaptive multiplicative extended Kalman filter (AMEFK). The filter integrates Earth sensor measurements for roll/pitch correction and an auxiliary sensor for yaw estimation. By deriving an error quaternion based on geometric relationships, the proposed method simplifies the filter structure, directly utilizing the error quaternion as the measurement input. To improve robustness against external disturbances and sensor faults, the filter adaptively updates the noise covariance using a sequential sum algorithm. Numerical simulations across different scenarios demonstrate the effectiveness and reliability of the proposed filter, validating its suitability for micro- and nano-satellite applications.