Inter-Satellite Link Prediction for Non-Terrestrial Networks Using Supervised Learning

Abstract

Distributed Space Systems (DSS) are becoming increasingly popular in the space industry as they integrate advancements in 6G and Non-Terrestrial Networks concepts to offer innovative and efficient solutions for satellite communication and data transmission. In those DSS where communication be-tween heterogeneous satellites is required, achieving autonomous cooperation while minimizing energy consumption is crucial (especially in sparse constellations with nano-satellites). This work proposes an autonomous and scalable solution based on a Supervised Learning model that enables heterogeneous satellites in circular polar Low Earth Orbits to predict their encounters with other satellites given the orbital elements and assuming isotropic antenna patterns. The proposed solution obtains an accuracy of around 90 % when evaluated with realistic data from real Celestrak satellites. This work could be considered the first stage of a promising and alternative approach in the field of DSS.