Beam Acquisition and Training in Millimeter Wave Networks Using Tones

This paper studies the initial access problem in millimeter wave networks consisting of multiple access points (AP) and user devices. A novel beam training protocol is presented with a generic frame structure. Each frame consists of an initial access sub-frame followed by data transmission sub-frames. During the initial subframe, APs and user devices sweep through a set of beams and determine the best transmit and receive beams via a handshake. Only narrowband tones are transmitted to mitigate mutual interference and training errors. Both non-coherent beam estimation using power detection and coherent estimation based on maximum likelihood (ML) are presented. To avoid exchanging information of beamforming vectors between APs and user devices, a locally maximum likelihood (LML) algorithm is presented. An efficient fast Fourier transform method is proposed for ML and LML to achieve highresolution beam estimation. A system-level optimization is performed to optimize the key parameters in the protocol, including the frame length, training time, and training bandwidth. The optimal training overhead is determined by those optimized parameters. Simulation results based on real-world topology are presented to compare the performance of different estimation methods and signaling schemes, and to demonstrate the effectiveness of the proposed protocol.