Deep learning based low complexity relay selection for wireless powered cooperative communication networks

Abstract

Energy harvesting relays significantly improve network performance in Wireless Powered Cooperative Communication Networks (WPCCNs). The relay selection problem in WPCCNs is commonly solved by iterative algorithms with high runtimes, which is unpractical for real-life applications. This paper proposes a low complexity solution based on deep learning to solve the relay selection problem with the objective of minimum schedule length in multi-source-multi-relay WPCCNs. We formulate the relay selection problem as a novel multi-class classification problem whose classes represent all possible relay selection combinations for all sources. To solve this classification problem, a feed-forward deep neural network (DNN) architecture is designed. The inputs are the channel gains and parameters derived from these gains based on the optimality conditions of the problem. The output is the relay selection for all sources represented by a class. Conventional supervised Machine Learning (ML) algorithms, including Decision Tree, Random Forest, Support Vector Machine, and K-Nearest Neighbour, are also implemented for benchmark comparisons. The proposed network outperforms the benchmark ML algorithms and previous iterative heuristic algorithms regarding precision, recall, f1-score, accuracy, and optimality gap in schedule length with lower runtime.