Compressed Beam Alignment with Out-of-Band Assistance in Millimeter Wave Cellular Networks

Abstract

Network transmission over millimeter-wave (mmW) bands has a big potential to provide orders of higher bandwidth. However, beamforming is generally needed to compensate for the high path loss. As mmW antennas have a potentially large number of candidate beamforming directions, to achieve high network throughput, the finding of a high gain direction between a base station and each mobile in the mmW network may involve a large overhead if training signals are directly sent along all possible directions or according to a large volume of codebook. Taking advantage of the block sparse characteristics of the mmW channel and coexistence of legacy antennas, we propose a comprehensive design for more efficient beam direction finding. Different from existing compressive-sensing-based schemes which just take a random subset of directions to measure, taking advantage of the path clustering feature of the mmW channel, we develop a self-adaptive block sparse algorithm which can benefit from preliminary channel estimation during each iteration of the problem solving to significantly improve the overall channel estimation accuracy thus the beam alignment gain. We also explore two methods to exploit co-located legacy antennas to provide further guidance for transmission direction finding. Simulation results indicate that our proposed beam alignment scheme outperforms the baseline and peer schemes in terms of the beamforming gain and training cost. By taking advantage of the block sparse properties of mmW channel, our proposed design is able to achieve the transmission throughput comparable with the exhaustive direction search at much lower overhead.