Future mobile communication services and applications demand real-time and ubiquitous AI support,with ITU extending 6G scenarios to encompass pervasive intelligence.To meet these demands,6G must integrate AI as a native capability and service.Current 5G networks rely on"use-case-driven,add-on or grafted"approaches to support intelligent applications,which face limitations in performance,versatility,and efficiency.This paper proposes an native AI architecture for 6G wireless networks,incorporating AI service orchestration,AI task management,multidimensional resource integration,and network digital twin functionalities.These features enable the on-demand,coordinated mapping of AI service requirements to underlying network resources,ensuring end-to-end,differentiated AI service quality.This architecture offers flexible and efficient high-quality AI services to both the network itself and a wide range of industries.Furthermore,potential research directions are outlined,including distributed AI model deployment and the integration of communication,sensing,AI and control,to adapt to emerging trends in AI foundation models and AI agents.
关键词
6G/智慧内生/AI服务质量/网络数字孪生/智能体
Key words
6G,native AI/quality of AI service/network digital twin/AI agents
维普
万方数据