[Paper Review] Edge Intelligence: Architectures, Challenges, and Applications

Dianlei Xu, Tong Li+6 more|arXiv (Cornell University)|Mar 26, 2020

Privacy-Preserving Technologies in Data336 references123 citations

TL;DR

A comprehensive survey that identifies four core edge intelligence components—edge caching, edge training, edge inference, and edge offloading—and provides a multi-dimensional taxonomy of architectures, techniques, and open challenges.

ABSTRACT

Edge intelligence refers to a set of connected systems and devices for data collection, caching, processing, and analysis in locations close to where data is captured based on artificial intelligence. The aim of edge intelligence is to enhance the quality and speed of data processing and protect the privacy and security of the data. Although recently emerged, spanning the period from 2011 to now, this field of research has shown explosive growth over the past five years. In this paper, we present a thorough and comprehensive survey on the literature surrounding edge intelligence. We first identify four fundamental components of edge intelligence, namely edge caching, edge training, edge inference, and edge offloading, based on theoretical and practical results pertaining to proposed and deployed systems. We then aim for a systematic classification of the state of the solutions by examining research results and observations for each of the four components and present a taxonomy that includes practical problems, adopted techniques, and application goals. For each category, we elaborate, compare and analyse the literature from the perspectives of adopted techniques, objectives, performance, advantages and drawbacks, etc. This survey article provides a comprehensive introduction to edge intelligence and its application areas. In addition, we summarise the development of the emerging research field and the current state-of-the-art and discuss the important open issues and possible theoretical and technical solutions.

Motivation & Objective

• Motivate edge-enabled AI to reduce latency, save bandwidth, and protect privacy.
• Systematically categorize the state-of-the-art across four edge intelligence components: caching, training, inference, and offloading.
• Compare edge approaches with traditional cloud-centric intelligence and discuss trade-offs.
• Summarize open issues and outline future research directions in data, models, privacy, and incentives.

Proposed method

• Propose a four-component framework: edge caching, edge training, edge inference, and edge offloading.
• Provide a multi-dimensional taxonomy for each component, including practical problems, techniques, objectives, performance, and drawbacks.
• Analyze literature from perspectives of application scenarios, methodology, and results.
• Discuss open issues and future research directions, including data scarcity, data consistency, model adaptability, privacy and security, and incentive mechanisms.

Experimental results

Research questions

• RQ1 What are the fundamental components and architectures of edge intelligence and how do they interact?
• RQ2 What practical problems, techniques, and goals characterize edge caching, edge training, edge inference, and edge offloading?
• RQ3 How does edge intelligence compare to traditional cloud-centric intelligence in terms of latency, privacy, and bandwidth?
• RQ4 What are the key open issues and future directions in data, models/algorithms, privacy, security, and incentives?

Key findings

• Identifies four key components of edge intelligence: edge caching, edge training, edge inference, and edge offloading.
• Provides a systematic, multi-dimensional classification and taxonomy for each component.
• Documents open issues and proposes five future research directions: data scarcity, data consistency, model/algorithm adaptability, privacy and security, and incentive mechanisms.
• Reviews growth trends and motivations, noting a rapid rise in edge training, inference, and offloading since 2014.
• Highlights that edge computing enables ultra-low latency, reduced end-device energy use, and scalable deployment through edge-cloud collaboration.