Deep Reinforcement Learning Based Interference Avoidance Beam-Hopping Allocation Algorithm in Multi-beam Satellite Systems

Abstract

Efficiently utilizing beam resources in multi-beam satellite systems is crucial for improving system performance. Existing beam-hopping allocation algorithms focus solely on Quality of Service (QoS) objectives, such as time delay, transmission rate, and system capacity, while overlooking the interference caused by large Low Earth Orbit (LEO) constellations on Geostationary Earth Orbit (GEO) satellite systems. Although beam shutdown and power reduction are common methods to mitigate interference, they can adversely impact the quality of service and communication coverage in areas. To address this issue, we propose a deep reinforcement learning based interference avoidance beam-hopping allocation algorithm(DRL-IABHA). The article proposes an IABHA optimization approach to maximize system throughput in a multi-beam satellite system under interference avoidance conditions. To handle the time-dependent nature of this problem, it is modeled as a Markov decision process (MDP), which is commonly used in deep reinforcement learning (DRL) analysis. The MDP state is transformed into an image and features are extracted using convolutional neural networks. Simulation results indicate that the DRL-IABHA algorithm can enhance system capacity by 22.94% and prevent interference with GEO ground stations more effectively than other beam assignment techniques