Beam Allocation based on Deep Learning for Wideband mmWave Massive MIMO

Abstract

Beam allocation is considered for wideband multiuser mmWave massive MIMO systems. By introducing the interference-free achievable rate, the analog precoder and the digital precoder is decoupled for the beam allocation problem. Then the beam allocation is treated as a multi-label classification problem and a deep learning-based beam allocation (DLBA) scheme is proposed, where a convolutional neural network is trained offline using the simulated environments to predict the beam allocation for all the users. In order to avoid the beam conflict and maximize the sum-rate, a rule to avoid the beam conflict is also proposed. Simulation results demonstrate that the DLBA scheme can substantially reduce the computational complexity with a marginal sacrifice of sum-rate performance, compared to the existing schemes.