Edge Graph Intelligence: Reciprocally Empowering Edge Networks with Graph Intelligence

TL;DR

This work defines Edge Graph Intelligence (EGI) as the mutual reinforcement between Graph Intelligence and edge networks, outlining how GI enhances edge services while edge infrastructures enable GI deployment. It offers a primer on GI and edge computing, presents a four-part framework of EGI enablers, and surveys applications, edge computation, GI-based edge optimization, and EGI ecosystems. The review introduces a six-level EGI rating, highlights representative edge applications (smart cities, robotics, health, POI, and mobile vision), and discusses operational architectures (federated, distributed, and on-device GI) and ecosystem components (hardware, software, datasets). It also identifies open challenges—ranging from new applications and large-scale GI to native edge support, explainability, security, and privacy—and outlines future research directions to advance EGI toward Level 5 integration. The survey aims to catalyze cross-disciplinary collaboration and accelerate practical deployments of GI at the network edge.

Abstract

Recent years have witnessed a thriving growth of computing facilities connected at the network edge, cultivating edge networks as a fundamental infrastructure for supporting miscellaneous intelligent services.Meanwhile, Artificial Intelligence (AI) frontiers have extrapolated to the graph domain and promoted Graph Intelligence (GI). Given the inherent relation between graphs and networks, the interdiscipline of graph learning and edge networks, i.e., Edge GI or EGI, has revealed a novel interplay between them -- GI aids in optimizing edge networks, while edge networks facilitate GI model deployment. Driven by this delicate closed-loop, EGI is recognized as a promising solution to fully unleash the potential of edge computing power and is garnering growing attention. Nevertheless, research on EGI remains nascent, and there is a soaring demand within both the communications and AI communities for a dedicated venue to share recent advancements. To this end, this paper promotes the concept of EGI, explores its scope and core principles, and conducts a comprehensive survey concerning recent research efforts on this emerging field. Specifically, this paper introduces and discusses: 1) fundamentals of edge computing and graph learning,2) emerging techniques centering on the closed loop between graph intelligence and edge networks, and 3) open challenges and research opportunities of future EGI. By bridging the gap across communication, networking, and graph learning areas, we believe that this survey can garner increased attention, foster meaningful discussions, and inspire further research ideas in EGI.

Figures (17)

• Figure 1: Illustration of the interplay between GI and edge networks, where GI can be applied as a data-driven tool to optimize edge networks, and conversely, edge networks perform as digital infrastructure to support GI deployment.
• Figure 2: Outline and conceptual relationships of EGI aspects discussed in this survey: Based on fundamental elements of GI (Sec. \ref{['sec:fundamental-gnn']}) and edge networks (Sec. \ref{['sec:fundamental-edge']}), EGI ecosystems (Sec. \ref{['sec:edge-infra-gnn']}) sustain all stakeholders in the closed loop of edge networks and graph intelligence. Edge Network for GI (Sec. \ref{['sec:edge_computation_gnn']}) reviews techniques for supporting edge computation of GI models and GI for Edge Network (Sec. \ref{['sec:gnn-optimization-edge']}) discusses GI-based optimizations on edge networks. Both of them serve as support for a rich set of EGI applications (Sec. \ref{['sec:gnn-application-edge']}).
• Figure 3: General workflow of GI models. Given an input graph with feature vectors, a GI model iteratively performs sampling, aggregation, update, and pooling through consecutive model layers. The obtained embeddings will be finally converted to results in an expected form through a readout function.
• Figure 4: Architecture overview of cloud-edge-end hierarchy, where distributed edge devices within edge networks serve as infrastructure for graph-intelligent applications and their networked data can be analyzed by graph representation learning models.
• Figure 5: The spectrum of edge networks can be classified into five categories: sensors and micro control units, embedded and mobile devices, robotics and Vehicles, edge servers, and edge cloud.