Optimization For Sub Chanel and Power Allocation 5 G / 6G Wireless Communication Systems using Machine Learning

Abstract

For several decades, optimization has been a key component in solving issues with wireless resource management, including beamformer design and power control. But these algorithms frequently need a large number of iterations to converge, which presents problems for real-time processing in systems like 5G/6G. We present a novel learning-based strategy for wireless resource management in this work. The main concept is to approximate the input and output of a resource allocation algorithm unknown non-linear mapping using a deep neural network (DNN). In order to create DNNs that can approximate certain algorithms of interest in wireless communications, we first characterize a class of `learnable algorithms' in this study. We show through extensive numerical simulations that DNNs can better approximate two extremely complicated algorithms used in wireless transmit signal design for power allocation while providing orders of magnitude computational time savings.