知识驱动的6G网络资源调度综述  

Knowledge-Driven Resource Management for 6G Networks:A Survey

作　　者：孙瑞锦 文添圣 杨浩 黄蕾 承楠 李长乐 SUN Ruijin;WEN Tiansheng;YANG Hao;HUANG Lei;CHENG Nan;LI Changle(School of Telecommunications Engineering,Xidian University,Xi’an 710071,China)

机构地区：[1]西安电子科技大学通信工程学院,陕西西安710071

出　　处：《无线电通信技术》2022年第4期630-637,共8页Radio Communications Technology

基　　金：国家重点研发计划(2020YFB1807700)。

摘　　要：随着网络覆盖的立体化、应用场景的多元化和服务需求的个性化,6G网络面临着资源维度高、网络动态性强、资源调度复杂度大等诸多问题。传统单独基于模型驱动或数据驱动的资源调度方法难以同时满足6G网络资源调度的精确性、时效性和稳健性等需求。为解决上述问题,提出将基于理论模型和专家经验的网络资源调度知识与神经网络方法深度融合,设计知识驱动的资源调度方法。首先,梳理了网络资源调度知识的定义和表征方式;其次,从协议层次和功能模块两个方面设计了知识定义的6G网络架构,分别引入了知识子层和网络资源决策知识库,并综述了现有基于知识驱动的资源调度方法;最后,针对现有方法的不足,提出了未来的研究方向,包括基于本体论的6G全场景知识图谱构建、基于6G全场景知识图谱的场景服务识别和基于多知识聚合的6G网络资源按需调度。Featuring three-dimensional network coverage,diversified application scenarios,and personalized service requirements,the sixth generation(6G)networks face several issues such as high resource dimensionality,dynamics of the network and resource allocation complexity.Traditional model-driven or data-driven resource allocation methods can hardly meet the demands of high accuracy,low latency and robustness of 6G network at the same time.To solve these problems,this paper proposes a knowledge-driven method,which combines network knowledge of theoretical models and expert experience with deep neural network.In this paper,a survey on knowledge-driven resource allocation for communications networks is presented.Firstly,the definition and classification of the knowledge in network domain are introduced.Then,a knowledge-defined architecture of 6G networks is designed from two aspects:protocol stack and functional modules.In specific,the knowledge sub-layer and the network knowledge base are respectively added.Furthermore,current knowledge-driven resource allocation algorithms are summarized.Finally,future research directions of knowledge-driven resources scheduling for 6G networks are proposed,including the construction of ontology-based knowledge graph(KG)for 6G full scenarios,KG-based service identification for 6G full scenarios and on-demand resource scheduling in 6G networks based on multi-knowledge aggregation.

关 键 词：知识驱动 神经网络 多维资源调度 知识图谱 

分 类 号：TN929.5[电子电信—通信与信息系统]