AI computing technologies for intelligent terminals: paths and challenges

doi:10.12267/j.issn.2096-5931.2026.03.003

Abstract

Abstract:

With the migration of artificial intelligence (AI) technologies toward the edge, intelligent terminals are evolving from cloud service access devices into computing nodes with local intelligence capabilities. This paper systematically sorts out the development path of intelligent terminal AI computing technologies, analyzing its evolutionary trajectory from cloud-based inference to edge-side inference, from lightweight models to the deployable large models, and from single-point optimization to system-level collaboration. It summarizes core optimization technologies including model compression, compilation optimization, hardware acceleration and system-level energy efficiency management, and discusses key challenges such as edge-friendly model paradigms, unified compilation and runtime systems, AI-native operating systems, and local intelligent agent construction, providing a useful reference for research and industrial practices in this field.

Key words: intelligent terminals, on-device artificial intelligence, edge computing, model compression, low-power computing

CLC Number:

TN929.11

BI Ran, LI Fengqiu, GE Jian, HUANG Yongtao. AI computing technologies for intelligent terminals: paths and challenges[J]. Information and Communications Technology and Policy, 2026, 52(3): 14-21.

Add to citation manager EndNote|Ris|BibTeX

URL:

http://ictp.caict.ac.cn/EN/10.12267/j.issn.2096-5931.2026.03.003

http://ictp.caict.ac.cn/EN/Y2026/V52/I3/14

Figures/Tables 5

References 12

[1]	中国信息通信研究院. 新一代智能终端蓝皮书(2024年)[R], 2024.
[2]	云晴. 智能终端发展观察与研究[J]. 通信世界, 2025(15):18-20.
[3]	侯伟胜. 智能终端市场洞察[J]. 商业观察, 2025, 11(17):1-4.
[4]	FRIHA O, FERRAG M A, KANTARCI B, et al. LLM-based edge intelligence: a comprehensive survey on architectures, applications, security and trustworthiness[J]. IEEE Open Journal of the Communications Society, 2024, 5: 5799-5856. doi: 10.1109/OJCOMS.2024.3456549 URL
[5]	陆寒熹, 蒋长献, 李雨晴. 智能终端数据中心性能优化技术研究[J]. 电工电气, 2025(10):71-76.
[6]	黄丰雨, 黄云霞, 傅蓉蓉, 等. 人工智能终端产业的发展现状研究[J]. 信息通信技术与政策, 2025, 51(12):92-96. doi: 10.12267/j.issn.2096-5931.2025.12.013
[7]	米援军. 智能终端与多模融合技术在现代指挥调度中的应用实践[J]. 信息系统工程, 2025(10):53-56.
[8]	HAN B, ZHANG X, QI Y, et al. A surrounding-adaptive QoE adjustment approach for optimization of capacity and quality of air interface[C]// 2012 19th International Conference on Telecommunications (ICT). Jounieh: IEEE, 2012:1-5.
[9]	ZHAO W, JING W, LU Z, et al. Edge and terminal cooperation enabled LLM deployment optimization in wireless network[C]// 2024 IEEE/CIC International Conference on Communications in China (ICCC Workshops). Hangzhou: IEEE, 2024:220-225.
[10]	ZHONG N Z, WANG Y, XIONG R, et al. CASIT: Collective intelligent agent system for internet of things[J]. IEEE Internet of Things Journal, 2024, 11(11):19646-19656. doi: 10.1109/JIOT.2024.3366906 URL
[11]	IGNATOV A, TIMOFEEV A, CHENG Z, et al. AI benchmark: running deep neural networks on android smartphones[C]// Proceedings of the European Conference on Computer Vision (ECCV) Workshops. Munich: Springer, 2018:288-306.
[12]	黎芷杉. 鸿蒙操作系统在分布式智能终端管理中的应用[J]. 信息与电脑(理论版), 2024, 36(17):90-93.

分类维度	类型	主要特征	典型应用场景
应用领域	消费类智能终端	面向个人与家庭用户,强调交互体验与生态兼容性	智能手机、可穿戴设备、智能家居
应用领域	行业应用类智能终端	面向专业场景,强调可靠性、安全性与专用功能	工业终端、医疗终端、车载系统
设备形态	便携式智能终端	具备高度移动性,运行于严格功耗与算力约束下	手机、平板、智能手表
	固定式智能终端	部署于固定场所,具备相对充足的算力与能源条件	智能电视、智能音箱、中控屏
	嵌入式智能终端	嵌入其他设备或系统中,承担特定智能功能	车载系统、工业控制模块
技术架构	传统智能终端	在既有终端形态基础上引入AI能力,实现功能增强	AI手机、AI个人计算机
技术架构	新一代智能终端	采用新架构与交互方式,具备更强自主感知与决策能力	AI眼镜、服务机器人

分类维度	类型	主要特征	典型应用场景
应用领域	消费类智能终端	面向个人与家庭用户,强调交互体验与生态兼容性	智能手机、可穿戴设备、智能家居
应用领域	行业应用类智能终端	面向专业场景,强调可靠性、安全性与专用功能	工业终端、医疗终端、车载系统
设备形态	便携式智能终端	具备高度移动性,运行于严格功耗与算力约束下	手机、平板、智能手表
	固定式智能终端	部署于固定场所,具备相对充足的算力与能源条件	智能电视、智能音箱、中控屏
	嵌入式智能终端	嵌入其他设备或系统中,承担特定智能功能	车载系统、工业控制模块
技术架构	传统智能终端	在既有终端形态基础上引入AI能力,实现功能增强	AI手机、AI个人计算机
技术架构	新一代智能终端	采用新架构与交互方式,具备更强自主感知与决策能力	AI眼镜、服务机器人

终端处理器	厂商	年份	准确率(INT8)	准确率(FP16)	功耗(INT8)	功耗(FP16)
Snapdragon 8 Elite	高通	2024	69.9%	95.4%	60.9	13.2
Dimensity 8400	联发科	2024	77.5%	97.8%	36.6	12.8
Google Tensor G2	谷歌	2022	83.8%	90.1%	12.6	5.1
Exynos 2100	三星	2021	74.7%	95.9%	2.5	3
Kirin ggo 5G	海思	2019	74.7%	95.7%	2.5	2.6

终端处理器	厂商	年份	准确率(INT8)	准确率(FP16)	功耗(INT8)	功耗(FP16)
Snapdragon 8 Elite	高通	2024	69.9%	95.4%	60.9	13.2
Dimensity 8400	联发科	2024	77.5%	97.8%	36.6	12.8
Google Tensor G2	谷歌	2022	83.8%	90.1%	12.6	5.1
Exynos 2100	三星	2021	74.7%	95.9%	2.5	3
Kirin ggo 5G	海思	2019	74.7%	95.7%	2.5	2.6

[1]	ZHAO Hui. Research on RISC-V architecture empowering innovation and development of intelligent terminals [J]. Information and Communications Technology and Policy, 2026, 52(3): 22-29.
[2]	TIAN Wenling, ZHAO Zhenyu, HUANG Rong, ZHANG Tiankui. Research on design for converged system architecture of MEC and CDN [J]. Information and Communications Technology and Policy, 2024, 50(2): 89-96.
[3]	GONG Geyong, LIU Yang. Design of AI smart elderly care service platform based on 5G technology [J]. Information and Communications Technology and Policy, 2023, 49(5): 91-96.
[4]	CAO Xiaobo, CHENG Qiang, YANG Yueqiang, SHI Hang, LIU Zishan. The deployment and key technologies of broadband optical access network and customer premises network supporting edge computing [J]. Information and Communications Technology and Policy, 2023, 49(10): 80-86.
[5]	XU Muhong, ZHANG Rui. Analysis and research on MEC industry development [J]. Information and Communications Technology and Policy, 2021, 47(7): 76-81.
[6]	LI Xiaomin. A deep learning framework based on 5G edge computing with its applications [J]. Information and Communications Technology and Policy, 2021, 47(4): 93-96.
[7]	SONG Ping, BI Libo. A scalable network I / O framework for edge computing [J]. Information and Communications Technology and Policy, 2021, 47(3): 33-40.
[8]	ZHANG Xiaoming, YAN Kai, CUI Dingxin. Research of 5G in VOCs monitoring and traceability of a chemical enterprise [J]. Information and Communications Technology and Policy, 2021, 47(10): 34-.
[9]	WU Xiaoyu, YU Bingyan, WANG Jing. [J]. Information and Communications Technology and Policy, 2020, 46(8): 45-49.
[10]	ZHANG Yuanjing, WANG Yao, XIE Jun, BI Ran. Analysis on security of 5G network and suggestions for security scheme [J]. Information and Communications Technology and Policy, 2020, 46(4): 47-53.
[11]	KONG Ling, ZHANG Yanjie, FAN Dian. Research on the development and application of mobile edge computing [J]. Information and Communications Technology and Policy, 2020, 46(3): 81-85.
[12]	TANG Xiongyan, WANG Youxian, CHEN Gao, HUANG Rong. Industrial status and development suggestions of Mobile Edge Computing [J]. Information and Communications Technology and Policy, 2020, 46(2): 1-5.
[13]	XU Muhong, WANG Xingyan. MEC technology and application in the video surveillance field [J]. Information and Communications Technology and Policy, 2020, 46(2): 87-91.
[14]	GE Hantao, LU Yeye. 5G applications and trends in the field of smart medicine [J]. Information and Communications Technology and Policy, 2020, 46(12): 15-20.
[15]	XU Muhong, YU Li, LI Lanfei. Research on MEC technology development [J]. Information and Communications Technology and Policy, 2020, 46(12): 86-92.