A New RF Beam Training Method and Asymptotic Performance Analysis for Multi-User Millimeter Wave Systems

In this paper, we consider a multi-user (MU) radio-frequency (RF) beam training scenario with user selection. We propose a new beam training scheme that alleviates the latency issue in the conventional IEEE 802.11ad by adopting a sequential downlink-downlink transmit sector sweep combination. Then, we analyze the average rate performance of several RF beam training schemes in two different asymptotic scenarios. In addition, we characterize the performance gain of the proposed method over other schemes. Our analytic results confirm that the proposed method achieves the average rate performance of the optimal fullsearch method in the asymptotic region with much reduced training overhead. It is shown from the simulation results that the proposed scheme outperforms the conventional beam training schemes and achieves a 35% performance gain over the optimal full search scheme when considering the beam training overhead in practical MU millimeter-wave channel environments. We also confirm that our analytical results match well with the numerical results.