Machine Learning Empowered Resource Allocation in IRS Aided MISO-NOMA\n Networks

Abstract

A novel framework of intelligent reflecting surface (IRS)-aided\nmultiple-input single-output (MISO) non-orthogonal multiple access (NOMA)\nnetwork is proposed, where a base station (BS) serves multiple clusters with\nunfixed number of users in each cluster. The goal is to maximize the sum rate\nof all users by jointly optimizing the passive beamforming vector at the IRS,\ndecoding order, power allocation coefficient vector and number of clusters,\nsubject to the rate requirements of users. In order to tackle the formulated\nproblem, a three-step approach is proposed. More particularly, a long\nshort-term memory (LSTM) based algorithm is first adopted for predicting the\nmobility of users. Secondly, a K-means based Gaussian mixture model (K-GMM)\nalgorithm is proposed for user clustering. Thirdly, a deep Q-network (DQN)\nbased algorithm is invoked for jointly determining the phase shift matrix and\npower allocation policy. Simulation results are provided for demonstrating that\nthe proposed algorithm outperforms the benchmarks, while the throughput gain of\n35% can be achieved by invoking NOMA technique instead of orthogonal multiple\naccess (OMA).\n