基于深度强化学习的低轨卫星跳波束资源分配方法OA

With the rapid development of digital economy,low earth orbit satellite communication has emerged as a crucial complement to terrestrial networks and an effective means of bridging the digital divide,owing to its wide coverage and relatively low latency compared to medium-and high-orbit satellites.However,the uneven spatial distribution of users,together with the bursty and tidal characteristics of service demands,leads to inefficiencies in conventional static resource allocation methods.To address the mismatch between user demand and resource supply in low earth orbit satellite Beam Hopping(BH)systems,a resource allocation scheme based on Deep Reinforcement Learning(DRL)is proposed.By formulating a joint optimization problem of BH patterns and power allocation,a Deep Q-network(DQN)algorithm integrated with Convolutional Neural Networks(CNN)is designed to achieve joint decision-making of beam switching and discrete power control.Simulation results demonstrate that the proposed method significantly improves system throughput and user satisfaction under various conditions of power constraints,user scales,and the number of active beams,thereby providing an efficient solution for intelligent resource management in LEO satellite communication.