Over-the-Air Federated Learning: Rethinking Edge AI Through Signal Processing
Seyed Mohammad Azimi-Abarghouyi, Carlo Fischione, Kaibin Huang
TL;DR
This work surveys Over-the-Air Federated Learning (AirFL), a framework that leverages wireless signal superposition to perform model aggregation directly in the air, dramatically reducing communication latency, bandwidth, and energy in edge learning. It classifies AirFL into CSIT-aware, blind, and weighted designs, and provides a unified analytical framework spanning signal models, synchronization, nomographic function computation, and convergence guarantees. Key contributions include concrete CSIT-based schemes (Local and Global CSIT), fully and partially blind approaches, and the WAFeL weighted-aggregation method with tractable weight optimization and heterogeneity considerations, supported by convergence analyses and experimental comparisons. The work highlights practical trade-offs, proposes scalable design directions, and outlines future avenues such as multi-antenna and decentralized variants, security and personalization, and integration with advanced wireless technologies.
Abstract
Over-the-Air Federated Learning (AirFL) is an emerging paradigm that tightly integrates wireless signal processing and distributed machine learning to enable scalable AI at the network edge. By leveraging the superposition property of wireless signals, AirFL performs communication and model aggregation of the learning process simultaneously, significantly reducing latency, bandwidth, and energy consumption. This article offers a tutorial treatment of AirFL, presenting a novel classification into three design approaches: CSIT-aware, blind, and weighted AirFL. We provide a comprehensive guide to theoretical foundations, performance analysis, complexity considerations, practical limitations, and prospective research directions.