Federated learning at the edge in Industrial Internet of Things: A review

Abstract

The convergence of Federated learning (FL) and Edge computing (EC) has emerged as an essential paradigm, particularly within the Industrial Internet of Things (IIoT) to enable the intelligent decision making. This work diligently examines the current state-of-the-art research at the intersection of FL, EC, and IIoT. An extensive review of the literature explores the diverse applications and challenges associated with this integration. The challenges range from privacy preservation and communication overhead to resource allocation. The incorporation of edge devices at which ensuring the federated learning in distributed manner helps to minimize energy consumption in IIoT, ultimately leads to a sustainable computing environment. By exploring the existing literature and research advancements, our goal is to highlight existing Edge-IoT software and hardware platforms and assess their usability in addressing challenges. In addition, we review existing recent frameworks, methodologies, and models employed to address these challenges, focusing on key performance matrices and its domain such as application, networking, and learning. We highlight the achievements and potential of FL and EC and underscore the need for tailored solutions to suit the unique demands of IIoT. Furthermore, we identify some of the major challenges as opportunities for future research, requires interdisciplinary collaboration and innovative algorithmic solutions. This work can help navigate through the challenges and unlock the full potential, contributing to the advancement of future IIoT applications.