Fast Millimeter-Wave Base Station Discovery via Data-driven Beam Training Optimization

Millimeter-Wave (mm-wave) communications is an important element of 5G. Due to the high propagation loss of mm-wave signals, directional transmissions are required even in the initial access (IA), where the base station (BS) needs to broadcast the reference signals with beamforming to reach sufficient coverage ranges. Sequential scanning with narrow beams at the BS, without considering the non-uniform distribution of user equipment (UE) in the angular space, may lead to long IA delay at UEs. To reduce the IA delay, we propose a data-driven approach that learns the spatial distribution of UEs from the historical channels of UEs served by the BS and a beam identification method based on density-based spatial clustering of applications with noise (DBSCAN) to find the optimized set of beams to match to the distribution of the UEs. Two time resource allocation strategies are then investigated to evaluate the performance of IA based on the optimized beam set identified according to the UE distribution. Numerical results via realistic ray-tracing experiments demonstrate the performance improvement of the proposed approach over sequential beam training and omnidirectional training.