An Empirical Atmospheric Density Calibration Model Based on Long Short-Term Memory Neural Network

The accuracy of the atmospheric mass density is one most important factors affecting orbital precision spacecraft at low Earth orbits (LEO). Although there are a number empirical models available to use in orbit determination and prediction LEO spacecraft, all them suffer from errors various degrees. A practical way reduce error particular model calibrate using precise data or tracking data. In this paper, long short-term memory (LSTM) neural network proposed NRLMSISE-00 model, which densities derived spaceborne accelerometer main input. resulted LSTM-NRL calibrated Challenging Minisatellite Payload (CHAMP) satellite, extensively experimented evaluate calibration performance. With month train shown effectively root mean square outside training window by more than 40% time spans space weather environment. also have remarkable transferring performance when it applied along GRACE satellite orbits.