Using extended Kalman filter and least squares method for spacecraft attitude estimation

Abstract

Herein, the purpose is to compare the attitude estimation results between the least squares method and the extended Kalman filter of an artificial satellite using real data of on-board attitude sensors. These estimation methods are applied for nonlinear problems, where t he first is an alternative for the estimation criterion of minimum variance, and yields instantaneous attitude determination, by processing the attitude sensors data. However, the extended Kalman filter carries out the processing of such sensors measurements in real time, and its formulation accounts for the dynamic noise of the states, yielding a kinematic attitude determination, using additionally the gyro measurements. It is observed that the averages of the values of the attitude estimated by the least squares method are very close to the results for the extended Kalman filter. In this way it can be concluded that the algorithm of the extended Kalman filter converges to the least squares solution when fed with real data supplied by the attitude sensors.