车联网边缘智能:概念、架构、问题、实施和展望  

Edge intelligence empowered internet of vehicles: concept,framework, issues, implementation, and prospect

作　　者：江恺 曹越[1,2] 周欢 任学锋[2] 朱永东 林海 JIANG Kai;CAO Yue;ZHOU Huan;REN Xuefeng;ZHU Yongdong;LIN Hai(Wuhan University,Wuhan 430072,China;Huali SmartWays Technology Co.,Ltd.,Wuhan 430056,China;China Three Gorges University,Yichang 443002,China;Zhejiang Lab,Zhejiang 310100,China)

机构地区：[1]武汉大学,湖北武汉430072 [2]华砺智行(武汉)科技有限公司,湖北武汉430056 [3]三峡大学,湖北宜昌443002 [4]之江实验室,杭州浙江310100

出　　处：《物联网学报》2023年第1期37-48,共12页Chinese Journal on Internet of Things

基　　金：湖北省国际科技合作计划项目(No.2022EHB002);教育部中国高校产学研创新基金支持项目(No.2021LDA07005);湖北省重大调研课题基金项目(No.2022KT03-2)。

摘　　要：作为一项新兴交叉学科领域,边缘智能通过将人工智能推送至靠近交通数据源侧,并利用边缘算力、存储资源及感知能力,在提供实时响应、智能化决策、网络自治的同时,赋能更加智能、高效的资源调配与处理机制,从而实现车联网从接入“管道化”向信息“智能化”使能平台的跨越。然而,当前边缘智能于车联网领域的成功实施仍处于起步阶段,迫切需要以更为广阔的视角对这一新兴领域进行全面综述。为此,面向车联网应用场景,首先介绍边缘智能的背景、概念及关键技术;然后,对车联网应用场景中基于边缘智能的服务类型进行整体概述,同时详细阐述边缘智能模型的部署和实施过程;最后,分析边缘智能于车联网中的关键开放性挑战,并探讨应对策略,以推动其潜在研究方向。As an emerging inter discipline field, edge intelligence pushes AI to the side close to the traffic data source. Edge intelligence makes use of the computing power, storage resources, and perception ability of edge to provide a more intelligent and efficient resource allocation and processing mechanism while providing a real-time response, intelligent decision-making and network autonomy, realizing the critical leap for internet of vehicles from access “pipelining” to the intelligent enabling platform of information. However, the successful implementation of edge intelligence in internet of vehicles is still in its infancy, and there exists a demand for a comprehensive survey in this young field from a broader perspective.Based on this context of internet of vehicles, the background, concepts and key technologies of edge intelligence were introduced. Then, a holistic overview of service types based on internet of vehicles was taken, and the entire processes of model training and inference in edge intelligence were elaborated. Finally, to promote the potential research directions, the key open challenges of edge intelligence in the internet of vehicles were analyzed, and the coping strategies were discussed.

关 键 词：人工智能 车联网 边缘智能 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]