Online Beam Learning for Interference Nulling in Hardware-Constrained mm Wave MIMO Systems

Abstract

Employing large antenna arrays is a key characteristic of millimeter wave (mmWave) and terahertz communication systems. Due to the hardware constraints and the lack of channel knowledge, codebook based beamforming/combining is normally adopted to achieve the desired array gain. Existing codebooks, however, are typically pre-defined and focus only on improving the beamforming gain of their target user, without taking interference into account, which incurs critical performance degradation. In this paper, we propose an efficient deep reinforcement learning approach that learns how to iteratively optimize the beam pattern to null the interference. The proposed solution achieves that while not requiring any explicit channel knowledge of the desired or interfering users and without requiring any coordination with the interferers. Simulation results show that the developed solution is capable of finding a well-shaped beam pattern that significantly suppresses the interference while sacrificing negligible beamforming/combing gain, highlighting a promising solution for dense mmWave/terahertz networks.