Deep Learning-Based Beamforming for Intelligent Reflecting Surface-Assisted mmWave Systems

Abstract

As a means to realize a smart and programmable wireless propagation environment, intelligent reflecting surface (IRS) has received a great deal of attention in recent years. By properly adjusting the phase shifts, the reflected signals can be added coherently at the receiver side to improve the signal power. In order to do so, the active beamforming at the transmitter and the passive beamforming at the IRS should be jointly optimized. In this paper, we put forth a deep learning (DL)-based beamforming for the IRS-assisted millimeter wave (mmWave) downlink systems. By applying the training data in the properly designed deep neural network (DNN), the proposed DL-based approach can optimize the whole transceiver processes in an end-to-end manner. Numerical results demonstrate that the proposed DL framework achieves nearly-optimal beamforming in terms of data rate performance.