物联网学报2026，Vol.10Issue(1)：81-98,18.DOI:10.11959/j.issn.2096-3750.2026.00503
面向6G的跨域知识驱动网络元智能算法框架
Cross-domain knowledge-driven meta-intelligent network algorithm framework for 6G
摘要
Abstract
A cross-domain knowledge-driven meta-intelligent network algorithm framework was proposed in this paper to address the limitations of existing automated operation and maintenance models in supporting multi-scenario and real-time intelligent management in 6G networks.Traditional approaches often rely on static rules or single-domain optimiza-tion,which are insufficient for 6G demands such as heterogeneous perception,dynamic policy adaptation,and multi-objective scheduling.The framework modeled network states across environmental,network,and user behavior domains,leveraging lightweight models and graph neural networks for high-level intent parsing and online knowledge fusion.A global knowledge base was dynamically updated via knowledge distillation.Multi-layer meta-intelligent agents formed a closed-loop control process of perception,reasoning,knowledge generation,decision issuance,validation,and memory re-trieval.Self-supervised learning,reinforcement learning,and meta-learning techniques were integrated to support rapid policy adaptation and continual optimization.Centered on a low-altitude traffic control scenario,the framework was evaluated through three tasks:knowledge-driven networking,intent-guided agent management,and swarm path planning.Experimental results show that the proposed method consistently outperforms baseline approaches in throughput,failure recovery time,traffic prediction accuracy,decision latency,execution success rate,resource fairness,path efficiency,and task success rate.关键词
网络知识/知识驱动网络/网络元智能/智能网络Key words
network knowledge/knowledge-driven network/network meta-intelligence/intelligent network分类
信息技术与安全科学引用本文复制引用
林佳琦,钱琪杰,钟旭东,冯涛,高先明,葛嘉鑫,彭木根,任保全..面向6G的跨域知识驱动网络元智能算法框架[J].物联网学报,2026,10(1):81-98,18.基金项目
国防重点实验室基金重点资助项目(No.6142006240401) Foundation Item:The Key Laboratory Fund of National Defense(No.6142006240401) （No.6142006240401）
