AI-Aided 3-D Beamforming for Millimeter Wave Communications

Abstract

Millimeter wave (mmWave) communications are a mainstream technology of the fifth generation (5G) systems. Beamforming plays an important role in mmWave communications thanks to its ability to provide beamforming gain to compensate the high propagation and penetration loss in the mmWave bands. To reduce the overhead (time and power consumption) in mmWave beamforming, AI techniques can be employed to enhance beam selection efficiency. We here propose a method that uses a machine learning algorithm to learn the features of mmWave channels by off-line training to assist online beam selection so that the overhead is significantly reduced without degrading the beamforming performance. The proposed framework involves codebook-based beam selection and local learning-based clustering algorithm with feature selection (LLC-fs). Simulation results are conducted to confirm the performance of the proposed method in three aspects, scalability, robustness and compatibility.