多移动终端轻量化感-算-策协同增强方法  

Lightweight sensing-computing-decision collaboration enhancement for multi-mobile terminals

作　　者：高元 刘思聪 郭斌[1] 徐祥瑞 卞浩羽 郝静怡 徐王锦 於志文[1,3] Yuan GAO;Sicong LIU;Bin GUO;Xiangrui XU;Haoyu BIAN;Jingyi HAO;Wangjin XU;Zhiwen YU(School of Computer Science,Northwestern Polytechnical University,Xi'an 710000,China;China Shipbuilding Research and Design Center,Wuhan 430000,China;College of Computer Science and Technology,Harbin Engineering University,Harbin 150000,China)

机构地区：[1]西北工业大学计算机学院,西安710000 [2]中国舰船研究设计中心,武汉430000 [3]哈尔滨工程大学计算机学院,哈尔滨150000

出　　处：《中国科学：信息科学》2024年第9期2136-2156,共21页Scientia Sinica(Informationis)

基　　金：国家杰出青年科学基金(批准号:62025205);国家自然科学基金(批准号:62032020,62102317)资助项目。

摘　　要：近年来,随着物联网和人工智能技术的融合,智能物联网(AI+IoT,AIoT)逐渐成为备受关注的新兴前沿领域.在这一背景下,深度学习驱动的智能应用逐渐渗透到智慧城市、公共安全等多个领域.为了实现智能计算从云端向物联网终端和边缘端延伸,智能物联网的多移动终端设备协同工作需要面对的挑战包括可用资源受限和环境动态变化等方面.在智能物联网中,多移动终端具备泛在感知、智能计算与自主决策能力,并参与到感知、计算、学习和决策的过程中.本文提出了多移动终端轻量化感-算-策协同增强方法,旨在克服单终端的视野、资源和性能局限,提升系统的感知覆盖和计算效率,提高在多种应用场景下的任务性能.In recent years,with the integration of the Internet of Things(IoTs)and artificial intelligence(AI)technologies,the concept of artificial intelligence in IoT(AIoT)has gradually emerged as a prominent frontier.Against this backdrop,deep learning-driven intelligent applications are increasingly permeating various domains such as smart cities and public safety.To extend intelligent computing from the cloud to IoT terminals and edge devices,the collaborative efforts of multiple mobile terminal devices in AIoT face challenges including limited available resources and dynamic environmental changes.In AIoT,multiple mobile terminals possess ubiquitous perception,intelligent computing,and autonomous decision-making capabilities,participating in the processes of perception,computation,learning,and decision-making.This paper proposes a collaborative enhancement method for lightweight perception,computation,and decision-making among multiple mobile terminals,aiming to overcome the limitations of single-terminal perspectives,resources,and performance,improve perception coverage and computational efficiency,and enhance task performance in various application scenarios.

关 键 词：智能物联网 数据融合感知 深度模型伸缩卸载 大小模型互馈决策 异构系统跨层优化 

分 类 号：TN929.5[电子电信—通信与信息系统] TP393[电子电信—信息与通信工程]