Machine learning approach

Atmospheric density estimation by machine learning approach
Very low Earth orbit (VLEO) which is 400-100 km in altitude is the area where many satellites will stay in the future. However, its atmospheric characteristics are still not fully understood. Especially, because atmospheric density has a critical impact on a space mission using satellites, it is necessary to establish a method for estimating the atmospheric density. Our laboratory is reconstructing a methodology to estimate atmospheric density in VLEO using GPS positioning data from the nanosatellite “EGG,” which was demonstrated in 2017. Machine learning approaches, which are Gaussian process regression and Bayesian optimization, are adopted to clarify the atmospheric density in VLEO by the data acquired by the nanosatellite.

Mission sequence of the nanosatellite EGG
Trajectory analysis result
Trajectory reconstruction by machine learning approach

Paper related

  • Y. Takahashi, M. Saito, N. Oshima, and K. Yamada, “Trajectory reconstruction for nanosatellite in very low Earth orbit using machine learning,” Acta Astronautica, vol. 194, pp. 301-308, 2022, DOI : 10.1016/j.actaastro.2022.02.010