Federated Learning-based Power Control and Computing for Mobile Edge Computing System

This work considers a federated learning (FL) framework for enabling the energy consumption optimization in a mobile edge computing (MEC) system by optimizing the local computing and the transmit power jointly. Firstly, we prove that in the maximum power transmission case the optimal solution is one of the two modes: the local execution mode and the full offloading mode. Consequently, the corresponding analytical energy consumption expressions are provided. Secondly, in the general case, the objective function has the non-convex rate expression and it is difficult to obtain the optimal solution. However, it can be transformed by the alternative optimization method to simplify the original problem. Finally, we propose a FL-based power control and local computing scheme by analyzing the characteristics of the energy expression to reduce the energy consumption. Numerical results show that the proposed scheme outperforms the benchmark schemes with more than 35% performance gain.