Location information aided beam allocation algorithm in mmWave massive MIMO systems

Millimeter wave (mmWave) band exhibits large bandwidth availability and meets the demands for high data rate of the fifth generation cellular networks (5G). However, mmWave signals suffer from severe path loss. Beamforming technology provides high array gain to overcome this limitation. Before data connection procedure, mmWave base stations (BSs) need to search for the suitable serving beams aiming at user equipments (UEs). This searching process is typically completed in the initial access procedure and causes fatal delays. Therefore, a reasonable beam allocation algorithm (BAA) is essential to speed up initial access procedure. In this paper, a support vector machine based beam allocation algorithm (SVM-BAA) is proposed, which utilizes location information and absorbs the thought of supervised learning. It constructs a SVM multi-classifier by using location information as features and serving beams as classes. When new UEs attempt to get access to mmWave BSs, their serving beams are allocated according to the decision results of SVM multi-classifier. Simulation results show that the proposed SVM-BAA can reduce the initial access delay compared with existing methods when there's enough previous location information.