大规模卫星网络的立体可重构资源管控架构与方法  

Spatio-configurable Resource Management Architecture andMethod for Mega Satellite Networks

作　　者：郝琪 周笛[1] 盛敏[1] 史琰[1] 李建东[1] HAO Qi;ZHOU Di;SHENG Min;SHI Yan;LI Jiandong(The State Key Laboratory of Integrated Service Networks,Xidian University,Xi’an 710071 China)

机构地区：[1]西安电子科技大学空天地一体化综合业务网全国重点实验室,陕西西安710071

出　　处：《无线电通信技术》2024年第6期1169-1176,共8页Radio Communications Technology

基　　金：国家自然科学基金(62371360,62422114,62121001);青年人才托举工程(2022QNRC001)。

摘　　要：大规模卫星网络(Mega Satellite Networks,MSNs)是由位于不同轨道高度、不同功能的卫星组成的多层卫星网络,可突破现有地面网络的覆盖限制,实现未来6G无缝覆盖的高质量服务需求。然而,MSNs中各类卫星高动态移动,致使网络资源异构且拓扑大时空尺度持续变化,对海量资源自主管控和保障多样化业务需求的实时响应提出了严峻的挑战。针对该挑战,设计了面向MSNs的立体可重构资源管控架构,通过在空间部署不同功能等级的卫星管控节点实现对MSNs异构资源的层次化管理,并提出基于深度强化学习(Deep Reinforcement Learning,DRL)的多维资源调度策略对多元业务需求动态重构不同管控节点的资源,以提高资源利用率和保障服务质量。仿真结果验证了基于所提架构的资源管控方法相较于传统启发式算法,可提升11.64%的资源利用率和46.4%的任务完成率。Mega Satellite Networks(MSNs)are multi-layer satellite networks composed of satellites located at different orbital altitudes and with different functions,which can break through the coverage limitations of existing terrestrial networks and achieve high-quality service requirements for future 6G seamless coverage.However,high dynamic movement of various satellites in MSNs results in heterogeneous resources and continuous changes in topology on a large spatio-temporal scale,posing a severe challenge to autonomous management of massive resources and ensuring real-time response to diversified service demands.For this challenge,a spatio-reconfigurable resource management architecture is designed for MSNs,and it realizes hierarchical management of heterogeneous resources by deploying satellite management nodes with different functional levels in space.Moreover,a multi-dimensional resource scheduling strategy based on Deep Reinforcement Learning(DRL)is proposed to dynamically reconstruct resources of different management nodes for multiple services,which can enhance resource utilization and guarantee service quality.Simulation results verify that the resource management approach based on the proposed architecture can raise the resource utilization rate by 11.64% and the task completion rate by 46.4% compared with traditional heuristic algorithms.

关 键 词：大规模卫星网络 软件定义网络 可重构管控 深度强化学习 

分 类 号：TN919.23[电子电信—通信与信息系统]