Constrained Deep Neural Network Based Hybrid Beamforming for Millimeter Wave Massive MIMO Systems

Abstract

Hybrid beamforming is a promising technology to reduce power consumption and provide high spectrum efficiency for millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) system. However, it is intractable to obtain global optima for similar constrained joint optimization problems by limitation of hardware architecture. In this work, we proposed a constrained deep neural network (constrained-DNN) based hybrid beamforming for mmWave massive MIMO system, which employs neural networks to replace the beamforming matrices in traditional hybrid beamforming to achieve end-to-end autonomous hybrid beamforming. Traditional hybrid beamforming optimization problem is transformed into a neural network optimization problem, which break the limitation of non-convex optimization. We also present numerical results on the performance of the proposed algorithms, which exhibits significant improvement on bit error rate (BER) performance compared with existing hybrid beamforming schemes.