A Two-stage Prediction-based Beam Selection Algorithm in MmWave Massive MIMO Systems

Millimeter Wave (mmWave) is usually combined with large-scale multiple antenna technology to produce a large number of narrow beams for better coverage. In order to improve the network throughput,an optimal beam or beam pair is selected to achieve high data rate. However, it is challenging to perform beam selection in vehicular systems with dynamic blocking scenario due to its complex geometric environment and variable number of multipaths. In this paper, a two-stage prediction-based beam selection (TS-PBBS) algorithm is proposed to find the optimal beam accurately and quickly in blocking scenarios. A deep neural network (DNN) is constructed to perform the mapping between the performance of predesigned probing beams and narrow beams. Furthermore, a local search method is introduced to improve the beam selection accuracy. Simulation results reveal that the proposed algorithm is able to recognize the blockage in the environment and achieve high accuracy for beam selection with relatively low searching complexity. In addition, it also appears to have the capability of resisting channel estimation error because the deep learning model weakens the effects of noise, which provides an effective solution for practical deployment of mmWave massive MIMO systems.