Abstract : We develop adaptive frequency block allocation schemes to mitigate the interference between intelligent Low Earth Orbit (LEO) satellites. As satellite networks attract increasing attention, the demand for limited frequency resources is expected to surge, creating a need for more efficient frequency utilization techniques. In particular, intelligent and dynamic frequency allocation methods will be more popular, which underscores the necessity for novel frequency resource allocation algorithms that take these considerations into account. In this work, we introduce two resource allocation strategies that exploit multi-agent reinforcement learning: the Unmodified Terrestrial-to-Satellite (UTS) strategy that extends previous terrestrial method to the satellite environment, and the Adapted Satellite Specific (ASS) strategy that is tailored to satellite communication systems. Through simulations in both controlled and interference prone environments, we evaluate and compare their performance, showing that, compared to the UTS strategy, the proposed ASS strategy improves throughput by up to 38% and reduces collision rate by up to 89% across different interference scenarios. Our findings highlight the effectiveness of customized resource allocation strategies in dynamic LEO satellite environments, paving the way for more efficient and scalable satellite communication systems in 6G networks.
Index terms : Interference management, Radio spectrum management, Resource allocation, Satellite communication, Multi-agent reinforcement learning