An Adaptation of Deep Learning Technique In Orbit Propagation Model Using Long Short-Term Memory

Abstract

The orbit propagation model is used to predict the position and velocity of the satellites. It is crucial to obtain accurate predictions to ensure that satellite operation planning is in place and detects any possible disasters. However, the model's accuracy decreases as the propagation span increases if the input data are not updated. Therefore, to minimize these errors while still maintaining the model accuracy, a study is conducted. The Simplified General Perturbations-4 (SGP4) model and two-line elements (TLE) data are selected to perform this study. The problem is analyzed, and the deep learning technique is the proposed method to solve the issue. Next, the enhanced model is validated. The study aims to produce a reliable orbit propagation model and assist the satellite's operational planning. Also, the improved model can provide vital information for space-based organizations and anyone who may be affected.