A Two-stage Beam Selection Method using Machine Learning and a Peak Finding Algorithm

The multiple-input and multiple-output (MIMO) techniques have evolved significantly in recent decades. We expect to reshape the 6G cellular networks by adopting new innovative techniques. In particular, beamforming is used to allow for directional multiple beams by weighting the magnitude or phase of signals in multiple antennas. It can improve spectral efficiency, reduce interferences, and expand coverage. It will be helpful for maintaining faster and more reliable connectivity in 6G networks. However, there are still many research challenges to design and implement beamforming systems. We need to optimize beam management schemes in terms of beam selection, connection set-up time, energy efficiency, throughput, network outage, QoS/QoE, and so on. It is not easy to find optimal network parameters and keep good link quality. The 3rd generation partnership project (3GPP) 5G new radio (NR) beam selection is based on an exhaustive search method, grid-of-beams (GoB), and precoding matrix indicator (PMI). The exhaustive search method results in a long beam search time, a high overhead cost, and a high energy consumption. In this chapter, we propose a two-stage beam selection method consisting of machine learning and peak finding algorithm to reduce beam sweeping overhead by predicting the beam selection subset and finding the optimal beam pair efficiently.