Estimates of spacecraft attitude and sensor parameters using adaptive filtering: Meteosat-1 transfer orbit application

Abstract

In practical applications of Kalman filtering theory often inconsistencies between real estimation errors and computed error variances occur due to modelling errors. An adaptive filtering technique has been developed at NLR to deal with this filter divergence problem. Special attention is paid to divergence detection. When divergence is detected, adaptation is applied with an intensity governed directly by the intensity of the divergence detected. This technique has been applied to METEOSAT-I transfer orbit flight data. Modelling errors are due to simple modelling of spacecraft dynamics and sensors. Spacecraft attitude and earth sensor delays have been estimated. Results, based on different combinations of sun aspect angle, earth chord and midscan rotation angle measurements show that in all these cases reliable estimates can be found.