Convolutional Neural Network Based Transmit Power Control for D2D Communication in Unlicensed Spectrum

Abstract

In this paper, we propose a means of Device-to-Device communication extended to unlicensed spectrum (D2D-U) to alleviate the dense deployment of smart devices in licensed spectrum with consideration of fairly coexisting with Wi-Fi. To achieve high system performance in D2D-U, a method of managing D2D mutual interference is needed. For this issue, we propose a convolutional neural network (CNN)-based transmit power control scheme which experiences a low computational complexity compared with conventional transmit power control scheme. Simulation results indicate that the CNN-based power control scheme can achieve superior performance with a low computational complexity.