A Review on Hierarchical Federated Learning: Architectures and Learning Algorithms

The rapid advancement of Hierarchical Federated Learning (HFL) has motivated researchers to explore novel algorithms and design strategies to address core challenges such as statistical heterogeneity, communication efficiency, and scalability. This study presents a comprehensive review of recent contributions over the past five years, focusing on HFL algorithms, including their architectural design and classification according to the key challenges they address. We define two principal architectures three-tier and clustered highlighting their advantages and limitations. Furthermore, we provide a taxonomy of algorithms based on design objectives and system constraints. Based on our analysis, we identify key open issues and outline promising future research directions to advance the development of robust and efficient HFL systems.