Satellite orbit prediction based on the deep neural network error model

摘要

In order to improve the accuracy of orbit prediction, a deep neural network is designed to model the error model which is caused by the differences between the satellite's real dynamical model and the existing dynamical model. An energy-based learning algorithm is used to train the deep neural network in order to avoid a situation where the trained network parameter is stuck in the local minimum. Short-term and long-term simulations are done with GPS satellites, the simulation results proved the effectiveness of deep belief network applied in orbit prediction.