Federated Learning for 6G HetNets' Physical Layer Optimization: Perspectives, Trends, and Challenges

Abstract

This chapter presents a survey that focuses on the implementation of federated learning (FL) techniques in sixth generation (6G) networks' physical layer (PHY) to meet the increasing user requirements. FL in PHY perspectives are discussed, along with the current trends and the present challenges in order to deploy efficient (cost, energy, spectral, computational) FL models for PHY tasks. Moreover, the utilization of FL methods is, also, discussed when channel state information (CSI) is not guaranteed in a 6G scenario. In such conditions, the joint use of cell free (CF) massive multiple-input-multiple-output (mMIMO), reconfigurable intelligent surfaces (RIS), and non-orthogonal multiple access (NOMA) and FL methods is proposed. Finally, an FL-based scheme for relay node (RN) placement in 6G networks is presented as an indicative use case for FL utilization in modern era networks. Results indicate that the proposed FL scheme overperforms state-of-the-art centralized learning schemes concerning the trade-off between machine learning (ML) metrics maximization and training latency.