Atmospheric density estimation in very low Earth orbit based on nanosatellite measurement data using machine learning

The atmospheric density in the very low Earth orbit (VLEO), the lower region of the thermosphere is a critical factor for satellite missions on prediction of orbit and lifetime. However, accurately determining atmospheric density remains challenging owing to difficulties in acquiring data within the VLEO. The re-Entry satellite with Gossamer aeroshell and GPS/Iridium (EGG) nanosatellite mission, conducted in 2017, obtained sparse coordinate data through the iridium network and Global Navigation Satellite System (GNSS). Using these data, we developed a methodology based on machine learning and special perturbations to estimate atmospheric density. The originally obtained atmospheric density predicted by NRLMSISE-00 was corrected during trajectory reproduction from GNSS data by special perturbation. The prediction performance of the method was validated using the generated trajectory data. Subsequently, we estimated the density at altitudes of approximately 200 km, particularly in the lower region of the VLEO. Density values that accurately reproduced trajectories of the EGG were 0.5 to 0.64 times greater than that predicted by the NRLMSISE-00 model. This methodology is effective for measuring atmospheric density in the lower VLEO region using GNSS data from low-cost and high-frequency nanosatellite missions. By increasing the availability of accurate atmospheric density data, this methodology enhances our understanding of the atmosphere in the VLEO. •A new approach for VLEO density estimation by machine learning is developed.•Estimated density suggests potential overestimation in the reference model.•The density accuracy based on nanosatellite GPS data is successfully evaluated.