无线算力网络:架构与关键技术  

The Architecture and Key Technologies of Radio Computing Power Networks

作　　者：郭凤仙 闫实 彭木根[1] 刘亮[2] 王尚广 石川[3] GUO Fengxian;YAN Shi;PENG Mugen;LIU Liang;WANG Shangguang;SHI Chuan(School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China;School of Artificial Intelligence,Beijing University of Posts and Telecommunications,Beijing 100876,China;School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]北京邮电大学信息与通信工程学院,北京100876 [2]北京邮电大学人工智能学院,北京100876 [3]北京邮电大学计算机学院,北京100876

出　　处：《移动通信》2025年第3期2-9,共8页Mobile Communications

基　　金：国家自然科学基金项目“移动边缘算力网络协同调度理论与方法”,“面向体征检测的无源射频计算协同感知理论与方法”(62201073,6243000390)。

摘　　要：在智能时代,无线网络引入了边缘计算、雾计算等技术,并向云化、开放化、智能化等方向持续演进,网络的计算能力逐渐增强。在未来,无线网络的作用将发生根本性变革,如支持内生计算、原生AI、通算融合等。无线算力网络旨在构建通算一体化的智能服务系统,实现无线接入网络内生算力对外开放以及原生AI支持,协同云、边、站、端多级算力,保障未来智能应用的端到端需求。首先从核心网、承载网、接入网三个方面概述了通算融合的研究现状。然后介绍了无线算力网络的体系架构和关键技术,包括算力基站、通算融合调度、移动性管理、云边端协同调度等技术,阐述了其原理和相关方法,旨在充分利用无线网络通算融合优势,以支持原生AI和未来智能应用的需求。最后总结了无线算力网络面临的技术挑战,并对未来的发展方向进行了展望。In the intelligent era,wireless networks have incorporated edge computing,fog computing,and other technologies,continuously evolving toward cloud-based,open,and intelligent directions,with gradually enhanced computing capabilities.In the future,wireless networks will undergo fundamental transformations,supporting endogenous computing,native AI,and communication-computing integration.Wireless computing power networks aim to construct an integrated intelligent service system that enables endogenous computing power within wireless access networks to be externally accessible with native AI support,coordinating cloud,edge,station,and terminal multi-level computing resources to ensure end-to-end requirements for future intelligent applications.This paper first outlines the current research status of communication and computing integration from three aspects:core network,transport network,and access network.It then introduces the architecture and key technologies of wireless computing power networks,including computing power base stations,communication-computing integrated scheduling,mobility management,and cloud-edge-end collaborative scheduling technologies,explaining their principles and related methods.These approaches aim to fully utilize the advantages of communication-computing integration in wireless networks to support native AI and future intelligent application requirements.Finally,the paper summarizes the technical challenges facing wireless computing power networks and provides an outlook on future development directions.

关 键 词：无线算力网络 通算融合 原生AI 云边端协同 

分 类 号：TN929.5[电子电信—通信与信息系统]