Improving Accuracy and Precision through Machine Learning Fusion using Two-Line Element Sets
Hao Peng, Xiaoli Bai · AIAA SCITECH 2022 Forum · 2022
View Video Presentation: https://doi.org/10.2514/6.2022-0863.vid The machine learning (ML) approach can improve both the orbit prediction accuracy and precision with a properly designed fusion strategy. ML models are trained based on historical TLE sets and then can generate ML-corrections to the future state and uncertainty predictions. The ML-correction will be first given a pseudo-physical interpretation as pseudo-measurements of the true orbit prediction error and then fused with the conventional orbital state and uncertainty propagation results. The regularized particle filter (PF) with progressive correction and systematic sampling methods is used to accomplish the uncertainty propagation task using the SGP4 model. Comprehensive experiments on different resident space objects (RSOs) with TLE sets in different orbit types are carried out and analyzed. The results reveal that the fusion strategy enables the ML approach to work with the advanced PF prediction method, which can enhance both the prediction accuracy and precision of PF.