**1. Introduction**

The International Data Corporation (IDC) forecasts that, by 2025, over 150 billion devices will be connected across the globe [1], with the majority of them generating data in real-time. In the same year, the source forecasts also that Internet of Things (IoT) devices located at the network edge will generate over 90 Zettabytes of data, namely more than the half of the Global Datasphere (i.e., the amount of data created, captured, and replicated in any given year across the world).

This shift in the digital landscape, from centralized data centers to a network of dispersed ubiquitous devices, requires a reassessment of the current methods of data analysis and processing to keep pace with the burgeoning volume and velocity of data. Currently, cloud-based data analysis and learning systems enable organizations to store and process vast amounts of data, making them readily accessible from any location at any time [2]. This enables organizations to make rapid and informed decisions, optimize operations, and reduce costs. However, with the increasing volume of data being generated and the high-expectations of IoT services' QoS/QoE [3], there is a growing need for more efficient methods for data processing/analysis and for system management, to be implemented closer to the source of data collection. This is where Edge Intelligence (EI) comes in, as it allows for the processing and analysis of data as much as possible at the network's edge, instead of solely relying on centralized data centers. This truly distributed and pervasive computing approach not only reduces latency and data traffic, but it also enables real-time decision-making, improves scalability, privacy, and reliability, and ultimately, leads to more efficient and effective data analysis.

Being recognized as an extremely promising enabler for many next-generation IoT services in different smart-\* domains, in a few years, the EI has become the centerpiece of a wide literature, spanning from very narrow technical contributions to comprehensive

**Citation:** Barbuto, V.; Savaglio, C.; Chen, M.; Fortino, G. Disclosing Edge Intelligence: A Systematic Meta-Survey. *Big Data Cogn. Comput.* **2023**, *7*, 44. https://doi.org/10.3390/ bdcc7010044

Academic Editor: Moulay A. Akhloufi

Received: 31 January 2023 Revised: 21 February 2023 Accepted: 27 February 2023 Published: 2 March 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

151

studies and informative analysis. As a result, currently, EI looks like a container of so many entangled concepts (astride IoT, AI, edge and cloud computing, data science) that are complex to approach and even more challenge to productively apply. Therefore, in order to offer an extensive and in-depth understanding of the theoretical basis, architectures, technologies, and application scenarios of the novel and multidisciplinary field of EI, this survey provides an overview of the research efforts made so far, by exploring the literature in accordance with two key principles:


The joint exploitation of these two distinct, but highly compatible, approaches to the research synthesis is a novelty in the EI literature (precisely, only [7] provided a systematic review) but as demonstrated in many other fields [8], it allows summarizing wide bodies of knowledge, quantifying the size, strength, and trend of research directions, and generating new valuable insights. Ultimately, this survey sets out to serve as a valid resource for anyone looking to stay current in this rapidly evolving field and to gain insights into the potential future developments in EI.

This manuscript is organized as follows. In Section 2, we introduce the main concepts to approach the EI paradigm, its diffusion in the IoT scenario, and its relationships with other mainstream paradigms such as edge and cloud computing, AI, etc. In Section 3, we provide a detailed report of the research objectives we pursued and of the search methodology we adopted, thus reviewing the obtained literature in Section 4. Final remarks conclude the manuscript in Section 5.
