Hybrid Precoding for Multiuser Millimeter Wave Massive MIMO Systems: A Deep Learning Approach

Abstract

In multi-user millimeter wave (mmWave) multiple-input-multiple-output (MIMO)\nsystems, hybrid precoding is a crucial task to lower the complexity and cost\nwhile achieving a sufficient sum-rate. Previous works on hybrid precoding were\nusually based on optimization or greedy approaches. These methods either\nprovide higher complexity or have sub-optimum performance. Moreover, the\nperformance of these methods mostly relies on the quality of the channel data.\nIn this work, we propose a deep learning (DL) framework to improve the\nperformance and provide less computation time as compared to conventional\ntechniques. In fact, we design a convolutional neural network for MIMO\n(CNN-MIMO) that accepts as input an imperfect channel matrix and gives the\nanalog precoder and combiners at the output. The procedure includes two main\nstages. First, we develop an exhaustive search algorithm to select the analog\nprecoder and combiners from a predefined codebook maximizing the achievable\nsum-rate. Then, the selected precoder and combiners are used as output labels\nin the training stage of CNN-MIMO where the input-output pairs are obtained. We\nevaluate the performance of the proposed method through numerous and extensive\nsimulations and show that the proposed DL framework outperforms conventional\ntechniques. Overall, CNN-MIMO provides a robust hybrid precoding scheme in the\npresence of imperfections regarding the channel matrix. On top of this, the\nproposed approach exhibits less computation time with comparison to the\noptimization and codebook based approaches.\n