Abstract A machine learning (ML) approach has been recently proposed to improve the orbit prediction accuracy of resident space objects (RSOs) through learning from historical data. Previous results have shown that the ML approach can successfully improve the point estimation accuracy. This paper extends the ML approach by introducing Gaussian Processes (GPs) which can generate uncertainty information about its point estimate. Both the simulation environment and the publicly available RSO catalogs are used to test the advanced ML approach. Numerical results demonstrate that the trained GP model can effectively improve the orbit prediction accuracy and generate uncertainty boundaries with high performance. Discussions and insights are also presented during the investigation using real data, including suggestions on designing learning variables and the possible causes for some unsatisfying results.
Highlights Gaussian processes (GPs) are used to generate both point estimates and uncertainties. Further extend the machine learning approach with GPs to improve orbit prediction. Both simulation environment and practical data are used to test the performance. Trained GP models reduce orbit prediction errors and provide reliable uncertainties.
Gaussian Processes for improving orbit prediction accuracy
Acta Astronautica ; 161 ; 44-56
2019-05-08
13 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Twin Gaussian Processes for Structured Prediction
British Library Online Contents | 2010
|Estimating Atmosphere Density Variations to Improve LEO Orbit Prediction Accuracy
British Library Conference Proceedings | 1998
|