AI Agent Centric Network: New Network Design Paradigm and Related Key Technologies

doi:10.12142/ZTECOM.202602002

ZTE Communications ›› 2026, Vol. 24 ›› Issue (2): 3-15.DOI: 10.12142/ZTECOM.202602002

• Special Topic • Previous Articles Next Articles

AI Agent Centric Network: New Network Design Paradigm and Related Key Technologies

Duan Xiangyang¹^,², Yang Li¹^,²(), Zhang Kangjie¹^,², Sun Wenwen¹^,², Xie Feng¹^,², Niu Li¹^,²

^1.State Key Laboratory of Mobile Network and Mobile Multimedia, Shenzhen 518055, China
^2.ZTE Corporation, Shenzhen 518057, China

Received:2026-04-11 Online:2026-06-16 Published:2026-06-16
About author:Duan Xiangyang is the Deputy General Manager of Technology Planning at ZTE Corporation and a professor-level senior engineer. He is responsible for technology pre-research planning and technical cooperation at ZTE. He has received one first prize of the National Science and Technology Progress Award, several first prizes of provincial and ministerial science and technology progress awards, and one first prize of the China Institute of Communications Technological Invention Award. His main research interests are in network communication systems technology.
Yang Li (yang.li8@zte.com.cn) is a professor-level senior engineer and serves as the chief engineer for 6G technology pre-research at ZTE Corporation. He is the head (senior expert) of the Algorithm Department’s Nanjing team, with 20 years of experience in wireless technology pre-research and standardization. He has authored and co-authored more than 30 technical papers, white papers, and technical monographs in the field of communications. His main research interests include strategic planning and system design for ODICT industry development, intelligent future wireless networks and digital twin applications, integrated communication, sensing, computing, intelligence, trust technologies, and vertical industry applications.
Zhang Kangjie is an engineer in wireless communications technology R&D at ZTE Corporation. He received his Ph.D. degree in communication engineering from Beijing University of Posts and Telecommunications (BUPT), China. He has published over 10 SCI-indexed papers. His research interests include 6G RAN architecture design and 6G AI agent communication protocols.
Sun Wenwen is a wireless technology pre-research engineer at ZTE Corporation. She received her master’s degree from Southeast University, China and her bachelor’s degree from Xi’an Jiaotong University, China. Her main research interests include 6G native intelligence and new service applications, such as digital twin and data plane technologies.
Xie Feng, PhD, is the head of 6G access network architecture and chief wireless architecture expert at ZTE Corporation, as well as the academic leader of the State Key Laboratory of Mobile Network and Mobile Multimedia Technology, China. He received his bachelor’s degree from Peking University, China and his Ph.D. degree from Nanyang Technological University, Singapore. He has conducted short-term research at BT Innovation Centre, UK and the University of Hong Kong, China. Since joining ZTE, he has served as subproject manager for WiMAX/LTE standard pre-research, head of 5G access network architecture pre-research, and chief designer for 5G RAN prototype and product protocol design.
Niu Li received her master’s degree from Beijing Jiaotong University, China in 2008. She is currently a pre-research engineer at ZTE Corporation in Beijing, China. Her research interests include IoT communications, satellite communications, and 5G/6G communications.
Supported by:
the National Science and Technology Major Project of China(2025ZD1304800)

Abstract

Abstract:

The diverse and heterogeneous terminal artificial intelligence (AI) agents and network-element AI agents are flourishing in a flywheel-like manner. The new characteristics of their capabilities and behaviors will reshape the service paradigm and traffic logic of future mobile information networks. This article first elaborates on the dynamics of the intertwined and integrated development of the AI agent/robot industry and the wireless communication industry. Then, based on an analysis of the new capabilities and behavioral characteristics of terminal AI agents and network-element AI agents, the article deduces a new design paradigm for future AI agent-centric (AA-Centric) networks, which includes seven core features: intent-driven, proactive service, distributed collaboration, efficient customization, deterministic guarantee, online evolution, and infinite generation. Guided by this new paradigm, the new architecture of 6G networks is further deduced, and the key supporting technologies are expounded. Finally, it is concluded that AI agents play a crucial role in driving the innovation of future network architectures and the ultimate expansion of capabilities.

Key words: 6G wireless system, AI agent, AA-Centric design, system capability characteristics, system behavior characteristics

Duan Xiangyang, Yang Li, Zhang Kangjie, Sun Wenwen, Xie Feng, Niu Li. AI Agent Centric Network: New Network Design Paradigm and Related Key Technologies[J]. ZTE Communications, 2026, 24(2): 3-15.

Figures/Tables 10

References 29

[1]	Xie F, Yang L, Chen L, et al. New requirements of 6G intelligent endogenous 2.0 and agent-based network architecture design [J]. Mobile communications, 2025, 49(1): 9–15
[2]	Pan J P, Cai L, Yan S, et al. Network for AI and AI for network: challenges and opportunities for learning-oriented networks [J]. IEEE network, 2021, 35(6): 270–277. DOI: 10.1109/MNET.101.2100118
[3]	Wang C X, You X H, Gao X Q, et al. On the road to 6G: visions, requirements, key technologies, and testbeds [J]. IEEE communications surveys & tutorials, 2023, 25(2): 905–974. DOI: 10.1109/comst.2023.3249835
[4]	Pennanen H, Hänninen T, Tervo O, et al. 6G: the intelligent network of everything [J]. IEEE access, 2025, 13: 1319–1421. DOI: 10.1109/access.2024.3521579
[5]	Yang Y, Ma M L, Wu H Q, et al. 6G network AI architecture for everyone-centric customized services [J]. IEEE network, 2023, 37(5): 71–80. DOI: 10.1109/mnet.124.2200241
[6]	Zheng Y T, Cheng X Z, Wang J Y. Research and application of AI agent empowering network operations [J]. ZTE technology journal, 2025, 31(5): 19–24. DOI: 10.12142/ZTETJ.202505004
[7]	Zhang P, Niu K, Wang X Y, et al. ComAI: the convergence of communication and artificial intelligence [J]. IEEE communications surveys & tutorials, 2026, 28: 2163–2197. DOI: 10.1109/COMST.2025.3608174
[8]	Duan X D, Huang Z L, Lu L, et al. AI agent communication network(ACN): a new network paradigm for 6G [J]. Journal on communications, 2025, 46(11): 332–346. DOI: 10.11959/j.issn.1000-436x.2025203
[9]	Chen Z Q, Sun Q, Li N, et al. Enabling mobile AI agent in 6G era: architecture and key technologies [J]. IEEE network, 2024, 38(5): 66–75. DOI: 10.1109/MNET.2024.3422309
[10]	Liu G Y, Chen T J, Cui Y P, et al. 6G native AI wireless network architecture for QoAIS guarantee [J]. Mobile communications, 2025, 49(1): 2–8. DOI: 10.3969/j.issn.1006-1010.20241125-0006
[11]	Zhang Y M, Guo D N. Multi-agent reinforcement learning for multi-cell spectrum and power allocation [J]. IEEE transactions on communications, 2025, 73(8): 5980–5992. DOI: 10.1109/TCOMM.2025.3534565
[12]	Yang L, Wang X Y, Duan X Y, et al. Key technologies and future outlook of agent-centric network technology framework [J]. Information and communications technologies, 2025, 19(2): 4–11. DOI: 10.3969/j.issn.1674-1285.2025.02.002
[13]	Zhao J, Pang R, Zhang S, et al. AI agent architecture for network digital twin [R]. IETF, 2026
[14]	Zhao J, Wang M X, Wu B, et al. AI based network management agent (NMA): concepts and Architecture [R]. IETF, 2026
[15]	Tong L, Ao S, Wang A J, et al. Network AI agent use cases and requirements in 6G [R]. IETF, 2025
[16]	Du Z P. Use cases and requirements of AI agent communication from 6G aspect [R]. IETF, 2025.
[17]	Boutouchent A, Mekrache A, Ksentini A, et al. 6G-INTENSE: intent-driven native artificial intelligence architecture supporting network-compute abstraction and sensing at the deep edge [J]. IEEE vehicular technology magazine, 2025, 20(1): 44–54. DOI: 10.1109/mvt.2024.3525001
[18]	Ling S, Xu C, Song J H. Network deep perception and adaptive communication method for robotic teleoperation, 2025, 49(5): 75–81. DOI:10.3969/j.issn.1006-1010.20250107-0001
[19]	Li K Y, Zhu Y P, Yu S X, et al. Ai-driven terminals: advances in intent recognition, multimodal interaction, and AI collaboration [C]//Proceedings of IEEE 6th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT). IEEE, 2025: 1–5. DOI: 10.1109/AINIT65432.2025.11035635
[20]	Liu E J, Li X, Tan Z, et al. Current status, trends, and policy recommendations for the AI-phone industry development [J]. Robot industry, 2025(6): 18–35. DOI:10.19609/j.cnki.cn10-1324/tp.2025.06.005
[21]	Liu M L, Ding Y H, Ma X Q, et al. Research on the development of embodied intelligence technology [C]//Proceedings of IEEE International Conference on Pattern Recognition, Machine Vision and Artificial Intelligence (PRMVAI). IEEE, 2025: 1–7. DOI: 10.1109/PRMVAI65741.2025.11108525
[22]	Liu Y K, Wang Q J, Zhu Z L, et al. Embodied artificial intelligent industrial robot: system architecture, key technologies and case study [J]. Computer integrated manufacturing systems, 2025, 31(12): 4513-4541. DOI:10.13196/j.cims.2024.Z21
[23]	Jin L M, Wang H C, Gu J C, et al. Research on low-altitude embodied artificial intelligence-enabled spectrum management and control technology [J]. Journal of data acquisition & processing, 2025, 40(1): 45–55. DOI: 10.16337/j.1004-9037.2025.01.004
[24]	Duan X D, Huang Z L, Lu L, et al. Agent communication network enabling new service provisioning and network architecture evolution [J]. Communications world, 2025(24): 5–6. DOI:10.13571/j.cnki.cww.2025.24.015
[25]	Duan X D, Sun T, Lu L, et al. Internet of agents: conception, architecture and key technologies [J]. Telecommunications science, 2025, 41(10): 1–10. DOI: 10.11959/j.issn.1000-0801.2025221
[26]	Sun F P, Qian Z T. Key technologies and applications of high level autonomous networks [J]. ZTE technology journal, 2024, 30(4): 77–82. DOI: 10.12142/ZTETJ.202404012
[27]	Gu H X, Zhao L Q, Han Z, et al. AI-enhanced cloud-edge-terminal collaborative network: survey, applications, and future directions [J]. IEEE communications surveys & tutorials, 2024, 26(2): 1322–1385. DOI: 10.1109/COMST.2023.3338153
[28]	Wang Z Q, Zhou J Z, Han K F. 6G AI-native network with edge-device collaboration [J]. ZTE technology journal, 2025, 31(4): 29–33. DOI: 10.12142/ZTETJ.202504005
[29]	Ehtesham A, Singh A, Gupta G K, et al. A survey of agent interoperability protocols: model context protocol (MCP), agent communication protocol (ACP), agent-to-agent protocol (A2A), and agent network protocol (ANP) [PP/OL]. V2. arXiv (2025-05-23)[2026-01-13].

Category	Intelligence	Planning Capability	Orchestration	Collaboration	Process Flexibility	Evolution & Updates
5G‑A status	External/built-in	Passive service	No	Weak	Predefined, rigid	Offline upgrade
6G new requirements	Natively embedded	Proactive service	On-demand	Strong	Flexible	Online evolution

Category	Intelligence	Planning Capability	Orchestration	Collaboration	Process Flexibility	Evolution & Updates
5G‑A status	External/built-in	Passive service	No	Weak	Predefined, rigid	Offline upgrade
6G new requirements	Natively embedded	Proactive service	On-demand	Strong	Flexible	Online evolution

Challenges	Current State	Key Challenges to solve
Communication-computation integration	Resource management and computation scheduling are decoupled	Joint optimization of inference and transmission latency; existing QoS mechanisms cannot enforce end-to-end determinism across the compute domain
Semantic-agnostic protocol stack	Protocol stack operates as a bitpipe; QoS defined solely by bandwidth, latency, and priority	No mechanism to differentiate semantic criticality; task-aware resource management is absent
Traffic burstiness	Scheduling assumes stationary traffic; GBR/non-GBR targets continuous flows	Agent coordination induces synchronous bursts; existing schedulers and buffers are ill-suited
Large-scale many-to-many access	5G NR scheduling is point-to-point; group communication support is limited	Many-to-many patterns are structurally mismatched with existing access mechanisms
Semantic transmission efficiency	Semantic communication remains pre-standardization	Token sequences impose high overhead; no compression or coding methods optimized for task-level metrics

Challenges	Current State	Key Challenges to solve
Communication-computation integration	Resource management and computation scheduling are decoupled	Joint optimization of inference and transmission latency; existing QoS mechanisms cannot enforce end-to-end determinism across the compute domain
Semantic-agnostic protocol stack	Protocol stack operates as a bitpipe; QoS defined solely by bandwidth, latency, and priority	No mechanism to differentiate semantic criticality; task-aware resource management is absent
Traffic burstiness	Scheduling assumes stationary traffic; GBR/non-GBR targets continuous flows	Agent coordination induces synchronous bursts; existing schedulers and buffers are ill-suited
Large-scale many-to-many access	5G NR scheduling is point-to-point; group communication support is limited	Many-to-many patterns are structurally mismatched with existing access mechanisms
Semantic transmission efficiency	Semantic communication remains pre-standardization	Token sequences impose high overhead; no compression or coding methods optimized for task-level metrics

AI Agent Centric Network: New Network Design Paradigm and Related Key Technologies

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 29

Related Articles 1

Recommended Articles

Metrics