@article{M2098BBDA,
title = "Reinforcement Learning-Based Low Earth Orbit Satellite Beam Hopping Algorithm Considering Traffic Distribution in South Korea",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2025",
issn = "1226-4717",
doi = "10.7840/kics.2025.50.3.432",
author = "Taehan Moon, Jaeyeol Lee, Tae-Yoon Kim, Youngpo Lee, Dongwook Kim, Takki Yu, Jae-Hyun Kim",
keywords = "Non-Terrestrial Network, Low Earth Orbit Satellite, Beam Hopping, Deep Q-network Algorithm",
abstract = "Low Earth orbit (LEO) satellites, unlike terrestrial networks, are not constrained by geographical limitations and have the advantage of providing data services to multiple regions. In multi-beam satellite communication systems, efficient resource management of spectrum, power, and capacity is essential, highlighting the importance of beam hopping (BH) technology. This paper proposes an Earth-fixed beam hopping algorithm based on a deep Q-network (DQN) for a multi-beam LEO satellite scenario over the South Korea. The proposed algorithm is designed to optimize beam hopping by efficiently managing the satellite’s limited capacity, accounting for channel conditions, and accommodating the random traffic distribution of ground cells.
Simulation results confirmed that the proposed algorithm improves the efficiency of satellite resources compared to existing heuristic algorithms, offering enhanced performance in maximizing the handling of cell traffic within South Korea."
}