Novel Scalable DNN-based Relay Selection Scheme for Wireless Powered Communication Networks

Abstract

In this paper, we propose an optimal relay selection scheme based on a deep neural network (DNN) to evaluate and minimize the outage probability (OP) in cooperative wireless powered communication networks (WPCNs). We explicitly split the relay selection problem into two parts: OP estimation and relay selection itself. We first consider the task from the perspective of the supervised learning regression problem. We train offline the DNN-based model on synthetic data. The proposed model is easy to scale for a network of any size since it predicts OP for each relay separately. Then, based on the results obtained in the previous step, the node with a minimum OP value is selected. Simulation results show that the proposed DNN-based relay selection scheme achieves minimum OP and exhibits the lowest mean square error than the models based on the state-of-the-art machine learning approaches.