Edge computing is a distributed Internet of Things (IoT) network architecture that enables computational power and data storage in close proximity to the data sources. Edge computing brings the computing on local servers rather than sending the data to cloud data centers to ensure real-time computing with low latency and high network speed.

Initially, edge computing is conducted using deep learning and machine learning which train their models by sharing the data with third parties, edge, or cloud servers. Still, these learning techniques have several constraints, such as communication cost, reliability, data privacy, security, and legalization. Federated learning is evolved as a powerful tool in addressing such constraints by implementing collaborative training of models at mobile edge networks via centralized servers while maintaining the data localization for the concernment of privacy and security.

Application areas of federated learning in edge computing are healthcare systems, vehicular networks, intelligent recommendations, and unmanned aerial vehicles, and some specific applications are computation offloading and content caching, malware and anomaly detection, task scheduling, and resource allocation. Challenges in federated learning for edge computing that need to be scrutinized are communication and computation efficiency, heterogeneity handling, participant selection, resource allocation, privacy-preserving, and service pricing.