面向6G核心网的内生智能体架构与意图驱动子网生成机制

李紫燕; 谷肖飞; 李慧欣; 王亚鹏; 艾明

doi:10.11959/j.issn.1000-0801.DXKX260132

您当前的位置：

首页 >

文章列表页 >

面向6G核心网的内生智能体架构与意图驱动子网生成机制

更新时间：2026-04-09

- 面向6G核心网的内生智能体架构与意图驱动子网生成机制
- Endogenous Intelligent Agent Architecture and Intent-Driven Subnetwork Generation Mechanism for 6G Core Networks
- 电信科学 2026年
- 作者机构：
  
  1.中信科移动通信技术股份有限公司创新中心，北京 100085
  2.无线移动通信国家重点实验室（电信科学技术研究院有限公司），北京，100191
  3.大唐移动通信设备有限公司，北京，100083
- 作者简介：
  
  [ "李紫燕（2001-），女，电信科学技术研究院在读硕士研究生，中信科移动通信技术股份有限公司预研工程师，主要研究方向为网络架构、人工智能、意图驱动网络等。" ]
  谷肖飞，guxiaofei@cictmobile.com
- 基金信息：
  
  移动信息网络国家科技重大专项(2026ZD1307000)
- DOI：10.11959/j.issn.1000-0801.DXKX260132
  中图分类号： TN929.5;TP18
- 收稿：2026-02-28，
  
  修回：2026-03-20，
  
  录用：2026-04-09，
- 稿件说明：
移动端阅览
李紫燕, 谷肖飞, 李慧欣, 等. 面向6G核心网的内生智能体架构与意图驱动子网生成机制[J/OL]. 电信科学, 2026.

LI Ziyan, GU Xiaofei, LI Huixin, et al. Endogenous Intelligent Agent Architecture and Intent-Driven Subnetwork Generation Mechanism for 6G Core Networks[J/OL]. Telecommunications Science, 2026.
李紫燕, 谷肖飞, 李慧欣, 等. 面向6G核心网的内生智能体架构与意图驱动子网生成机制[J/OL]. 电信科学, 2026. DOI： 10.11959/j.issn.1000-0801.DXKX260132.

LI Ziyan, GU Xiaofei, LI Huixin, et al. Endogenous Intelligent Agent Architecture and Intent-Driven Subnetwork Generation Mechanism for 6G Core Networks[J/OL]. Telecommunications Science, 2026. DOI： 10.11959/j.issn.1000-0801.DXKX260132.

摘要

为实现IMT-2030（6G）普惠智能愿景，提出一种内生智能（AI-Native）核心网架构，该架构具有多智能体协同特征。基于服务化架构，通过定义意图受理（AIEF）、智能控制（ACF）与资源管理网络功能（ARMF）等核心网络功能，构建了面向6G场景的内生智能网络架构。为验证架构可行性，选取意图驱动子网生成任务作为典型任务，搭建整体系统验证架构所提注册、编排、UE接入基本功能。针对6G逻辑子网构建需求，设计了由ACF全程协调保障的两阶段多智能体协同的子网生成机制，并在单卡RTX4090D下开展仿真实验与性能评估工作。在意图解析阶段，利用大模型常识推理能力将非结构化意图解析收敛为标准化策略契约，推理时延平均为1744.9ms；在子网规划与生成阶段，由两个智能体协同推理完成子网自动化规划与生成，子网生成时延平均为185.7ms，1000条不同意图的子网生成成功率为96.6%。验ss证表明，所提架构与机制可实现基于自然语言的意图表述到子网实例的生成闭环，验证了未来网络引入AI智能体实现内生的可行性。

Abstract

To realize the IMT‑2030 (6G) vision of ubiquitous intelligence

this paper proposes an AI‑native core network architecture featuring multi‑agent collaboration. Built upon a service‑based framework

the architecture introduces three core network functions: Intent Awareness (AIEF)

Intelligent Control (ACF)

and Resource Management (ARMF). To validate the architecture

an intent‑driven subnet generation task is selected as a representative use case

and a two‑stage multi‑agent collaborative mechanism coordinated by ACF is designed. Experiments conducted on a single RTX 4090D show that in the intent parsing stage

a large language model leverages commonsense reasoning to convert unstructured intents into standardized policy contracts

achieving an average inference latency of 1744.9 ms. In the subsequent subnet planning and generation stage

two AI agents collaboratively automate the subnet generation process

with an average latency of 185.7 ms. Across 1

000 diverse intents

the subnet generation success rate reaches 96.6%. The results confirm that the proposed architecture enables a closed loop from natural-language intent to subnet instance generation

validating the feasibility of embedding AI agents as native components in future networks.

关键词

Keywords

references

Recommendation ITU-R M . 2160-0 (11/2023): Framework and overall objectives of the future development of IMT for 2030 and beyond [S ] . Geneva : ITU , 2023 .

崔琪楣 , 尤肖虎 , 倪巍 , 等 . 面向6G网络的AI与通信融合探析：基础、挑战与未来研究机遇 [J ] . 中国科学:信息科学 , 2026 , 56 ( 01 ): 241 - 242 .

朱威 , 马田丰 , 金友兴 , 等 . 智能体互联网潜在关键技术展望 [J ] . 信息通信技术 , 2025 , 19 ( 05 ): 41 - 54 .

中国移动通信研究院 . 智能体通信网络（ACN）白皮书（ 2024 年）[R/OL ] . 2024 .

Wang, L., Ma, C., Feng, X. et al . A survey on large language model based autonomous agents . Front. Comput. Sci. 18 , 186345 ( 2024 ).

Yao S , Zhao J , Yu D , et al . ReAct: Synergizing Reasoning and Acting in Language Models [C ] // Proceedings of the 11th International Conference on Learning Representations (ICLR 2023) .

Wang G , Xie Y , Jiang Y , Mandlekar A , Xiao C , Zhu Y , Fan L , Anandkumar A . Voyager: An Open-Ended Embodied Agent with Large Language Models [J ] . Transactions on Machine Learning Research , 2024 .

Li G , Hammoud H A A K , Itani H , et al . CAMEL: Communicative Agents for “Mind” Exploration of Large Language Model Society [C ] // Proceedings of the 37th International Conference on Neural Information Processing Systems (NeurIPS 2023) .

Xu M , Niyato D , Kang J , Xiong Z , Mao S , Han Z , Kim D I , Letaief K B . When Large Language Model Agents Meet 6G Networks: Perception, Grounding, and Alignment [J ] . IEEE Wireless Communications , 2024 , 31 ( 6 ): 63 - 71 .

3GPP. Study on Architecture for 6G System; Stage 2 (Release 20): 3GPP TR 23.801-01 V0.3.0 (2025-11) [R ] . Sophia Antipolis : 3GPP, 2025 - 11 .

Huawei ; SK Telecom ; Google . KI#18 - AI for 6G Architecture (Moderator Merged tdoc): 3GPP TSG-WG SA2#173, TDoc S2-2601162r5-cl, Goa, India, 9–13 February 2026 [R ] . 3GPP , 2026 - 02 .

China Mobile ; Apple ; Samsung . KI#19: Solution variants: 3GPP TSG-WG SA2#173, TDoc S2-2601625, Goa, India, 9–13 February 2026 [R ] . 3GPP , 2026 - 02 .

刘海涛 , 谌丽 , 康绍莉 , 等 . 面向6G网络融合的以用户为中心关键技术 [J ] . 无线电通信技术 , 2024 , 50 ( 03 ): 453 - 460 .

CATT . CATT initial views on 6G Core [C ] // 3GPP workshop on 6G . Incheon, Korea, March 10 – 11 , 2025 . pp. 1 – 7 .

段晓东 , 黄正磊 , 陆璐 , 等 . 智能体通信网络赋能新型服务供给和网络架构演进 [J ] . 通信世界 , 2025 ,( 24 ): 5 - 6 .

段晓东 , 黄正磊 , 陆璐 , 等 . 智能体通信网络（ACN）：面向6G的网络发展新范式 [J ] . 通信学报 , 2025 , 46 ( 11 ): 332 - 346 .

张培龙 , 马云潇 , 李华 , 等 . 大语言模型驱动的多智能体网络意图识别框架 [J/OL ] . 小型微型计算机系统 , 1 - 10 [ 2026-01-16 ] .

Ma X , Zhao R , Liu Y , et al . Design of a large language model for improving customer service in telecom operators [J ] . Electronics Letters , 2024 , 60 ( 10 ): e13218 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

面向5G-Advanced与6G的智能无线电接入网：关键标准技术与未来演进

面向6G空天地一体化系统的OTA测试方法与关键技术展望

面向实时通信的边缘智能关键技术研究

基于标准空间光传送网协议的天地一体化网络构建研究