星地融合网络中基于无监督图学习的卫星切换方法研究  

Research on Satellite Handover Methods in Satellite-Terrestrial Integrated Networks Based on Unsupervised Graph Learning

作　　者：刘人鹏 涂文康 胡博[1] 王恒 齐文 LIU Renpeng;TU Wenkang;HU Bo;WANG Heng;QI Wen(State Key Laboratory of Networking and Handover Technology,Beijing University of Posts and Telecommunications,Beijing 100876,China;China Telecom Research Institute,Beijing 102209,China)

机构地区：[1]北京邮电大学网络与交换技术全国重点实验室,北京100876 [2]中国电信股份有限公司研究院,北京102209

出　　处：《移动通信》2024年第1期33-39,共7页Mobile Communications

基　　金：国家自然科学基金“支持快速移动的密集接入网关键技术”(61931005);国家重点研发计划项目“天地融合的6G卫星通信立体智能组网技术研究”(2020YFB1807900)。

摘　　要：随着未来6G移动通信研究的快速发展,星地融合通信将是满足全域覆盖、泛在接入的重要技术。在星地融合网络中切换决策直接影响网络性能与服务持续性,现有切换算法主要通过实时监测候选卫星状态,计算权重进行切换判决,难以适应高动态星地网络中时空周期异步的切换决策。提出一种基于无监督图学习的卫星切换方法,引入卫星覆盖时空图表征网络连接关系,设计前向蚂蚁和后向蚂蚁发现并反馈的信息素感知机制,建立图学习模型,挖掘卫星覆盖时空图中卫星节点、用户节点与连接关系之间的隐式特征信息和拓扑信息,将切换决策问题转化为求解节点和边之间的连接关系概率,同步切换决策周期。仿真结果验证了所提卫星切换方法具有较快的收敛速度,切换时延降低20%。As 6G mobile communication research rapidly progresses,satellite-terrestrial integrated communication emerges as a crucial technology for achieving universal coverage and ubiquitous access.In these networks,handover decisions critically influence network performance and service continuity.Existing algorithms,primarily based on real-time monitoring and weight calculation of candidate satellites,struggle to adapt to the spatio-temporally asynchronous decision-making required in dynamic satellite-terrestrial environments.This paper introduces an unsupervised graph learning-based satellite handover method.It incorporates a spatio-temporal graph of satellite coverage to characterize network connections and designs a pheromone perception mechanism facilitated by forward and backward ants for discovery and feedback.The graph learning model developed mines implicit features and topological information of satellite and user nodes within the spatio-temporal graph,transforming the handover decision problem into solving for connection probabilities between nodes and edges,thereby synchronizing the decision-making cycle.Simulation results confirm that the proposed method achieves faster convergence,reducing handover delay by 20%.

关 键 词：星地融合网络 移动性管理 切换决策 图神经网络 

分 类 号：TN929.5[电子电信—通信与信息系统]