The present work aims at the development of adaptive schemes for a Kalman filter processing GPS data to perform an optimal estimate of the attitude and angular velocities of the user satellite. An extended Kalman filter with continuous states and discrete measurements has been used to represent the attitude determination process. An adaptive filter updating at each time step the measurement covariance matrix has been designed. To optimize the performances the selection of the measurement covariance matrix has been based on the analysis of the residuals, intended as differences between the actual and predicted measurements. Five different algorithms have been tested, and their performances have been compared to those of a previously designed non-adaptive Kalman filter, in terms of accuracy of the estimated attitude and increased computational cost. In three different cases, the accuracy of the estimated attitude is increased with a limited added computational cost.
展开▼
