边缘智能:研究进展及挑战

Edge Intelligence:Research Progress and Challenges

近年来,物联网的普及让数以亿计的移动设备连接到互联网上,在网络边缘产生了海量的数据,使得一种全新的计算范式——边缘计算兴起。同时,得益于深度学习算法和摩尔定律的突破,使得人工智能的发展再一次迎来了高潮。在这一趋势下,将边缘计算与人工智能相结合是必然的,由此产生的新的交叉研究——边缘智能引起了许多学者的广泛关注。在该综述中,边缘智能被分为基于边缘计算的人工智能和基于人工智能的边缘计算(即AI on edge和AI for edge)两部分。AI on edge侧重于研究如何在边缘计算平台上进行人工智能模型的构建,主要包括模型训练和模型推理两部分;AI for edge侧重于借助先进的人工智能技术,为边缘计算中的关键问题提供更优的解决方案,主要包括任务卸载和边缘缓存两部分。该综述从一个广阔的视角对边缘智能的研究进行了归纳总结,为涉足该领域的相关学者提供了一个详细的背景知识。

In recent years,the popularity of the internet of things has enabled hundreds of millions of mobile devices to connect to the Internet and generated massive data at the edge of network,leading to the rise of a new computing paradigm—edge computing.At the same time,thanks to the breakthrough of deep learning algorithm and Moore’s Law,the development of artificial intelligence once again ushered in a climax.In this trend,it is inevitable to combine edge computing with artificial intelligence,and a new cross research called edge intelligence has attracted extensive attention of many scholars.In this paper,edge intelligence is divided into two parts:artificial intelligence based on edge computing and edge computing based on artificial intelligence(AI on edge and AI for Edge).AI on Edge focuses on how to construct artificial intelligence model on edge computing platform,including model training and model reasoning.AI for Edge focuses on providing better solutions to key problems in edge computing with the help of advanced artificial intelligence technology,mainly including task offloading and edge caching.This paper summarizes the research on edge intelligence from a broad perspective,which provides a detailed background knowledge for related scholars involved in this field.