智能物联网:概念、体系架构与关键技术  

AIoT:The Concept,Architecture and Key Techniques

作　　者：郭斌[1] 刘思聪 刘琰 李志刚[1] 於志文[1] 周兴社[1] GUO Bin;LIU Si-Cong;LIU Yan;LI Zhi-Gang;YU Zhi-Wen;ZHOU Xing-She(School of Computer Science,Northwestern Polytechnical University,Xi’an 710072;School of Computer Science,Peking University,Beijing 100871)

机构地区：[1]西北工业大学计算机学院,西安710072 [2]北京大学计算机学院,北京100871

出　　处：《计算机学报》2023年第11期2259-2278,共20页Chinese Journal of Computers

基　　金：国家自然科学杰出青年基金(No.62025205);国家自然科学基金(62032020,62102317,62302017);中国博士后科学基金(No.2023M730058)资助。

摘　　要：智能物联网是当前人工智能与物联网技术相融合的产物,正成长为一个具有广泛发展前景的新兴前沿领域,实现从“万物互联”到“万物智联”的演进.在人工智能、边缘计算、物联网、移动嵌入式硬件等技术发展背景下,本文系统性地介绍智能物联网这一新兴方向.它对物联网感知、通信、计算和应用通过人工智能技术赋能,呈现泛在智能感知、云边端协同计算、分布式机器学习、人机物融合等新特征,具有更高灵活性、自组织性、自适应性.本文首先介绍了智能物联网的基本概念特质;其次阐述了智能物联网的体系架构;进一步详细介绍了智能物联网中的研究挑战与关键技术,包括泛在智能感知、群智感知计算、智能物联网通信、终端适配深度计算、物联网分布式学习、云边端协同计算、安全与隐私保护;最后,基于最新研究动态展望了极具潜力的未来研究方向,包括软硬协同终端智能、面向AIoT的智能演进、新一代智能物联网络、动态场景模型持续演化、人机物融合群智计算和通用AIoT系统平台.Artificial Intelligence of Things(AIoT)is an emerging research field with broad development prospects,realizing the evolution from“Internet of Everything”to“Intelligent Connection of Everything.”With the technological development such as artificial intelligence,edge computing,the Internet of Things,and mobile/embedded devices,AIoT aims to build a self-organizing,self-learning,self-adaptive,and con-tinuous-evolving smart IoT system based on the deep fusion of these advanced technologies.AIoT is the combination of Artificial Intelligence(AI)technology and Internet of Things(IoT)infrastructure to achieve more intelligent IoT applications and provide more efficient services.Artificial intelligence models are good at analyzing and mining the potential patterns and strategies from massive amounts of data,while IoT has the ability to establish extensive connectivity for hundreds of millions of physical devices.AIoT em-powers the perception,communication,computing,and application of the Internet of Things through vari-ous artificial intelligence technologies,and draws new features such as ubiquitous intelligent sensing,cloud-edge-end collaborative computing,distributed machine learning,and human-machine-things fusion.It has higher flexibility, self-organization, and self-adaptability. Specifically, AIoT enables the collection of multi-modal real-word data in real-time, and then utilizes machine learning approaches on the end devices, edge clusters or cloud servers for intelligent processing and decision making. This paper systematically in-troduces the direction of AIoT. Specifically, this paper firstly introduces the essential conceptual character-istics of AIoT, and then elaborates its architecture. Furthermore, we detail the research challenges and key technologies in AIoT, including ubiquitous intelligent sensing, mobile crowd sensing and computing, AIoT communication, terminal-adapted deep computing, AIoT distributed learning, cloud-edge- end collaborative computing, as well as security and privacy protectio

关 键 词：智能物联网 群体智能 深度模型 边缘智能 人机物融合群智计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]