Communication-efficient Federated Learning Design with Fronthaul Awareness in NG-RANs

Next Generation Radio Access Networks (NG-RANs) have become a promising paradigm to meet the strict demands of the 5G and beyond applications by distributively pushing the radio and computing functionalities in the close approximate to end-users. With the emergence of new technologies, network densification, and richer and more demanding applications, the limited capacity of the fronthaul links and privacy concerns poses a severe constraint on realizing NG-RAN systems in the real environment. To tackle these challenges, we propose a Federated Learning (FL)-based NG-RAN algorithm, named FedNG, in which the User Equipment (UEs), as well as NG-RAN infrastructures, collaborate throughout the learning process and the sharing prediction model to ensure privacy and relieve the burden on fronthaul interface. Specifically, our proposed scheme enables Distributed Units (DUs) to cooperatively learn a shared predictive model by taking the first-phase training models of the DUs as the initial input of the local training and then uploading sub-optimal DU models to the Central Unit (CU) to involve in the next phase of global training. Finally, numerical results are provided to evaluate our proposed scheme in terms of accuracy, service latency, and traffic size. The convergence of our proposed algorithm confirms that our approach significantly outperforms the existing state-of-the-art solution based on Federated Averaging (FedAvg).