Reinforcement Learning-Based Dynamic Routing for Robust Optimization of Low Earth Orbit (LEO) Satellite Communication Networks

Recently, satellite communication has garnered significant attention as a novel industry capable of providing global internet access in conjunction with the next-generation communication system, 6G. Notably, low-Earth orbit satellites, operating at comparatively lower altitudes, offer an advantage in communication system configuration due to their closer proximity to Earth. The inherent characteristics of LEO satellites, such as their high orbital speed and deployment of numerous satellites in the same orbit, necessitate research into inter-satellite routing technology for enhanced communication performance. Consequently, this study presents a routing algorithm aimed at optimizing the LEO satellite communication network by employing reinforcement learning, a machine learning technique. By applying various reinforcement learning algorithms to satellite topologies that may arise in space environments, the superiority of the algorithm is assessed, and simultaneously, the feasibility of implementing inter-satellite routing in space is demonstrated.