Location and Attitude Information Aided Codeword Selection in Millimeter Wave MIMO System

Abstract

Recently, the location and attitude information (LAI) from sensors have been utilized to assist beamforming in millimeter wave (mmWave) system for the potential of reducing training overhead. In this paper, we propose a machine learning based LAI assisted codeword selection algorithm. Specifically, based on the spatial consistency of mmWave channel, we transform the codeword selection problem into a nonlinear classification problem with the LAI of the user equipment (UE). Furthermore, we derive the relationship between the LAI of UE and the arrival angles of the line-of-sight (LOS) path. To solve this nonlinear classification problem, a custom kernel function based on the definition of linearly separable sample space is proposed for the support vectors machine (SVM) method. Finally, simulation results are presented to show the effectiveness of the proposed algorithms.