Autocovariance least-squares based measurement error covariance estimation for attitude determination of lunar lander

Abstract

In this paper, the autocovariance least-squares (ALS) method for a linear time-varying system is proposed for estimating both the process noise covariance and the measurement noise covariance associated with the measurement sensor to mitigate the performance degradation caused by an incorrect information of the sensor errors or by a large change of errors in sensor measurement. To verify the efficiency of the proposed method, simulations were performed for the attitude determination of the lunar lander which combines a gyro and a star tracker measurements assuming that the star tracker errors are increased by fault during the lunar descent and landing. The simulation results show that the attitude error of the proposed method is smaller than the conventional Kalman filter by adaptively tuning the noise covariance of the star tracker measurement errors. Also, the relationship between the ALS estimation accuracy and the innovation sample size and the time lag is discussed.