Accurate and efficient orbit prediction through improved drag force modelling of GRACE and PROBA-V satellites

Abstract

In recent years, with the increasing number of man made objects in space the need for accurate satellite orbit prediction has increased tremendously. Prediction of satellite trajectories is important to plan collision avoidance manoeuvres between space assets and debris, to autonomously maintain formation flying missions and to plan manoeuvres for ground-track maintenance of Earth-observation missions. For satellites in very low LEO, aerodynamic drag is the largest and the most difficult force to model because of the changing nature of atmospheric density. This report describes the efforts made towards improving the orbit prediction of SAOCOM-CS with a focus on drag force modelling. This is accomplished by orbit determination using GPS state vector measurements and precise deterministic force models, during periods of high and low solar activity. Drag scale factors are estimated with different resolutions. Different methods are used to choose the estimated drag scale factors for orbit prediction. GRACE-A and PROBA-V satellites are used as test cases. For a prediction arc length of one day, the best prediction strategy results in maximum position errors (3D) of 243.5 m and 24.1 m for GRACE-A \& PROBA-V, respectively during high solar activity. Based on the prediction results of GRACE-A \& PROBA-V, a rule of thumb analysis is used to derive the maximum position error in the orbit prediction of SAOCOM-CS, which lies between 40 and 75 m. Changes in the mean estimated drag scale factors of the satellites are observed between high and low solar activity which might indicate deficiencies in the NRLMSISE-00 density model. The report also provides the effect of the space weather forecast errors on the best prediction strategy. Introducing a 10 \% error in the solar activity index resulted in mean maximum along-track prediction errors of 393 m and 16 m for GRACE-A \& PROBA-V, respectively during high solar activity. Similarly, including random errors in the geomagnetic activity index resulted in mean maximum along-track prediction errors of 443 m and 15 m for GRACE-A \& PROBA-V, respectively during high solar activity. Finally, the optimization of the estimation and prediction methods for the computational efficiency of PROBA-V is presented. A six hour estimation arc length with the force model comprising Earth gravity field of degree and order 30 of the model ITU\_GRACE 16 with luni-solar perturbations and devoid of atmospheric drag, solar radiation pressure and tidal forces is the most computationally efficient combination for a prediction arc length of one day.