GEO卫星多波束动态频率-功率联合分配研究OA
Research on Joint Allocation for Multi-beam Dynamic Frequency-power of GEO Satellite
针对地球同步轨道(GEO)多波束卫星面临的频率、功率、干扰约束,建立以总服务率最大化为目标的系统优化模型,提出一种基于深度强化学习(DRL)的分支深度 Q网络(DQN)算法,实现多波束动态频率-功率联合分配.其中,将联合动作拆分为起始频率槽、频率槽数量和功率档位3个子决策,以有效降低所提DQN动作空间维度.仿真结果表明:分支DQN算法服务率达86.5%,较传统近端策略优化(PPO)算法提升约5.0%,因此可应用于 GEO卫星的频率规划与功率调整,以提高系统服务满足度.
To address the spectrum,power,and interference constraints of geostationary orbit(GEO)multi-beam satellite,a system optimization model with the objective of maximizing the total service ratio is established,and a split-branch deep Q network(DQN)algorithm based on deep reinforcement learning(DRL)is proposed to realize joint allocation for multi-beam dynamic frequency-power.The joint action is decomposed into three sub-decisions-starting frequency slot,frequency slot number,and power level-effectively to reduce the action-space dimension.Simulation results show that split-branch DQN algorithm achieves a service ratio of 86.5%,which is about 5.0%higher than that of the traditional proximal policy optimization(PPO)algo-rithm.Therefore,it can be applied to frequency planning and power regulation of GEO satellite to improve the system service satisfaction.
朱浩宁;李静涛;董赞扬;聂高峰
中国空间技术研究院通信与导航卫星总体部,北京 100094中国空间技术研究院,北京 100094中国空间技术研究院通信与导航卫星总体部,北京 100094北京邮电大学,北京 100876
信息技术与安全科学
地球同步轨道卫星多波束频率-功率联合分配深度Q网络
GEO satellitemulti-beamfrequency-power joint allocationDQN
《航天器工程》 2026 (4)
25-32,8
评论