Research on development of mobile AI for 6G

doi:10.12267/j.issn.2096-5931.2026.10.002

Abstract

Abstract:

This paper analyzes the connotation and core vision of Mobile AI, summarizes the core requirements and constructs a hierarchical indicator system for ToC smart life, ToB smart production and ToG urban governance scenarios, and further proposes key technologies of Mobile AI from the perspectives of network and AI. Finally, the development path and suggestions of Mobile AI are presented, providing reference for the high-quality development of 6G intelligence-communication integration.

Key words: mobile AI, 6G, agent, device-edge-network-cloud collaboration, Token communication

CLC Number:

TN929.5

DU Liutong, LI Yang, WANG Huanli, LIU Yujuan, LI Zhuofang, DU Ying, XU Fei, ZHANG Jing. Research on development of mobile AI for 6G[J]. Information and Communications Technology and Policy, 2026, 52(10): 9-15.

Add to citation manager EndNote|Ris|BibTeX

URL:

http://ictp.caict.ac.cn/EN/10.12267/j.issn.2096-5931.2026.10.002

http://ictp.caict.ac.cn/EN/Y2026/V52/I10/9

Figures/Tables 3

References 14

[1]	ITU-R. Framework and overall objectives of the future development of IMT for 2030 and beyond:recommendation ITU-R M.2160-0[R]. 2023.
[2]	工业和信息化部. 2026年前7个月通信业经济运行情况[EB/OL]. (2026-08-28)[2026-08-28]. https://www.miit.gov.cn/gxsj/tjfx/txy/art/2026/art_d78fc2fbd1d9467286f9a71986bdafe7.html.
[3]	新华网. Token经济加速起势彰显智能经济新活力[EB/OL]. (2026-08-06)[2026-08-06]. https://www.news.cn/tech/20260806/eb5c18bf7944459abc6aad82f836ec4f/c.html.
[4]	OpenAI. GPT-4 technical report[PP/OL]. V1. arXiv (2023-03-15)[2026-08-06]. https://arxiv.org/abs/2303.08774.
[5]	Howard A G, Zhu M, Chen B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[PP/OL]. V1. arXiv (2017-04-17)[2026-08-06]. https://arxiv.org/abs/1704.04861.
[6]	IMT-2030(6G)推进组. 6G总体愿景与潜在关键技术白皮书[R]. 2021.
[7]	ITU-R. Minimum requirements related to technical performance for IMT-2030 radio interface(s)[R]. 2026.
[8]	全国机器人标准化技术委员会. 具身智能机器人和移动通信技术研究报告(2025版)[R]. 2026.
[9]	3GPP. Study on 6G use cases and service requirements: TR 22.870 (Release 20)[R]. 2026.
[10]	GB/Z 177—2026 人工智能终端智能化分级[S].
[11]	Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]// Advances in Neural Information Processing Systems 30. Long Beach:Curran Associates, 2017: 5998-6008.
[12]	IMT-2030(6G)推进组. 6G网络架构愿景与关键技术展望白皮书[R]. 2023.
[13]	TS 28. 105(Release 18) Management and orchestration; artificial intelligence/machine learning (AI/ML) management[S].
[14]	3GPP. Study on artificial intelligence (AI)/machine learning (ML) for NR air interface: TR 38.843 (Release 18)[R]. 2024.

关键指标	指标定义	近期参考指标
首Token交互时延/ms	用户发起请求后,系统输出首个有效响应内容的时间,反映系统“是否够快地开始响应”的能力(参考Humanbenchmark)	≤400
Token间时延/ms	系统连续生成响应过程中,相邻Token间的平均输出间隔,反映交互连续性与流畅性	≤50
指令识别准确率	模型在面对复杂、明确、可验证的指令约束时,能够一次性生成完全符合要求的回答的比例(参考IF Bench)	>95%
推理准确率	AI模型预测正确的样本占总样本的比例(参考AA—LCR)	>95%

关键指标	指标定义	近期参考指标
首Token交互时延/ms	用户发起请求后,系统输出首个有效响应内容的时间,反映系统“是否够快地开始响应”的能力(参考Humanbenchmark)	≤400
Token间时延/ms	系统连续生成响应过程中,相邻Token间的平均输出间隔,反映交互连续性与流畅性	≤50
指令识别准确率	模型在面对复杂、明确、可验证的指令约束时,能够一次性生成完全符合要求的回答的比例(参考IF Bench)	>95%
推理准确率	AI模型预测正确的样本占总样本的比例(参考AA—LCR)	>95%

类别	关键指标	指标定义	近期参考指标
通信类	泛在网络上行速率	超过95%的场景区域下,网络可以支持的上行实时速率大小	密集城区室内:50 Mbit/s;^[7] 城区室外:20 Mbit/s
	用户面时延	指一个数据包从发送端的无线协议栈入口,经过空口传输,到接收端的无线协议栈出口的单向传输时间	毫秒级
	端到端时延抖动	同一业务类型的数据包,先后到达接收端的端到端时延差值的变化量的范围	≤50 ms^[8]
	可靠性	在预定的时间段内以高成功概率传输一定量业务流量的能力	99.999%^[7] [1ms内传输32字节]
感知类	水平定位精度^[7]	设备在水平方向上的定位误差的累积分布函数(CDF)的90%点	室内≤ 0.75 m 室外≤6 m
感知类	检测概率^[7]	检测概率是指当感知目标存在时正确检测到该目标存在的概率	≥95%
智能类	关键Token传输可靠性^[9]	多维感知信息Token化后再上传,关键Token的传输正确率需求	>99.999%
智能类	非关键Token传输可靠性	多维感知信息Token化后再上传,非关键Token的传输正确率需求	>80%

类别	关键指标	指标定义	近期参考指标
通信类	泛在网络上行速率	超过95%的场景区域下,网络可以支持的上行实时速率大小	密集城区室内:50 Mbit/s;^[7] 城区室外:20 Mbit/s
	用户面时延	指一个数据包从发送端的无线协议栈入口,经过空口传输,到接收端的无线协议栈出口的单向传输时间	毫秒级
	端到端时延抖动	同一业务类型的数据包,先后到达接收端的端到端时延差值的变化量的范围	≤50 ms^[8]
	可靠性	在预定的时间段内以高成功概率传输一定量业务流量的能力	99.999%^[7] [1ms内传输32字节]
感知类	水平定位精度^[7]	设备在水平方向上的定位误差的累积分布函数(CDF)的90%点	室内≤ 0.75 m 室外≤6 m
感知类	检测概率^[7]	检测概率是指当感知目标存在时正确检测到该目标存在的概率	≥95%
智能类	关键Token传输可靠性^[9]	多维感知信息Token化后再上传,关键Token的传输正确率需求	>99.999%
智能类	非关键Token传输可靠性	多维感知信息Token化后再上传,非关键Token的传输正确率需求	>80%

[1]	WANG Yanmin, FENG Wei, DONG Chao, REN Baoquan, DUAN Xiangyang. Sensing-communication-computing-control closed loop for 6G unmanned operations: model, metric, and network design [J]. Information and Communications Technology and Policy, 2026, 52(9): 2-8.
[2]	LIU Guangyi, HE Zhimin, WANG Xinyao, CAI Qing, WANG Kaiyue, CHEN Tianjiao, JIN Jing. 6G communication-computing-intelligence convergence empowering embodied intelligence applications [J]. Information and Communications Technology and Policy, 2026, 52(9): 29-37.
[3]	ZHANG Wenjing, CHAO Junxian, WU Zijian, CHENG Baoping. Evolution directions and development pathways of embodied artificial intelligence [J]. Information and Communications Technology and Policy, 2026, 52(9): 38-44.
[4]	LIU Jun, DONG Ji, WANG Wei. Research on security protection and governance paths of terminal agents from perspective of risk management and control [J]. Information and Communications Technology and Policy, 2026, 52(8): 18-22.
[5]	WU Hanbing, FENG Yizhuo. Research on user authorization mechanisms and identity forgery risks in agent interconnection scenarios [J]. Information and Communications Technology and Policy, 2026, 52(8): 2-10.
[6]	HAN Han, WANG Jingyao, CHANG Tianen. A novel agent security software testing scheme driven by large language model [J]. Information and Communications Technology and Policy, 2026, 52(8): 74-81.
[7]	XU Ming. A review of cross-domain data security governance for low-altitude one-network unified flight platform [J]. Information and Communications Technology and Policy, 2026, 52(6): 70-78.
[8]	WANG Shiji, WANG Liwen, ZHANG Yan, CAO Chang, WANG Runyan, HE Wanxian, PEI Pei. Service semantic routing technology for application-level computing power services [J]. Information and Communications Technology and Policy, 2026, 52(5): 80-88.
[9]	KANG Shaoli, WANG Ke, QIN Haichao, FENG Yunjiu, WANG Dan, YANG Qixin. A study on use cases for low-altitude intelligent networked systems towards 6G [J]. Information and Communications Technology and Policy, 2026, 52(4): 2-14.
[10]	CHENG Ying, LIU Zhipeng, ZHANG Meng. Research on development pathways, potential impacts, and governance strategies of terminal intelligent agents [J]. Information and Communications Technology and Policy, 2026, 52(3): 2-7.
[11]	YANG Lei, WANG Feng, ZHANG Xin, LIN Jun. The ecological game and comparison of cooperative modes between intelligent terminals and internet services in the large model era [J]. Information and Communications Technology and Policy, 2026, 52(3): 8-13.
[12]	GUO Zhenming, CHEN Li, CHEN Feng. AI-empowered wireless positioning technologies in the 6G era [J]. Information and Communications Technology and Policy, 2026, 52(10): 16-25.
[13]	ZHANG Jing, WANG Huanli, CUI Can, LI Zhuofang, DU Liutong. Research on standardization progress of mutual empowerment between 6G networks and AI [J]. Information and Communications Technology and Policy, 2026, 52(10): 2-8.
[14]	XU Mingfeng, LI Yang, XU Weihua, ZHOU Jizhe, JIANG Jiamo. Key technologies, challenges and prospects for 6G enabled edge-end collaboration [J]. Information and Communications Technology and Policy, 2026, 52(10): 40-46.
[15]	LI Jiahui, SUN Yanran, LONG Biao. Multi-agent collaboration and AI capability lifecycle management architecture for 6G core networks [J]. Information and Communications Technology and Policy, 2026, 52(10): 47-54.