Study Of Beamforming In Millimeter Wave Massive MIMO Communications

This paper studies the beamforming for millimeter wave(mm Wave)massive MIMO,mainly including the codebook design and beam training,for both point to point systems and multi-user systems.A two-step codeword design method is proposed.In the first step,codewords are designed without considering the the hardware constraint of hybrid structure,which is termed as ideal codeword design.For the ideal codeword design,additional phase is introduced to the beam gain to provide extra degree of freedom and a phase-shifted ideal codeword design(PS-ICD)method is proposed.To determine the additional phase,an alternative minimization method is used,where each iteration of the method is based on a closed-form solution we derive.Given the ideal codewords designed in the first step,the practical codeword design in the second step aims to approach the ideal codewords by considering the hardware constraints of the hybrid beamforming structure in terms of phase shifter resolution and the number of radio frequency(RF)chains.A fast search based alternative minimization(FS-Alt Min)algorithm that alternatively designs the analog beamformer and digital beamformer is proposed.Then the beam training for point-to-point mm Wave massive MIMO is studied.A beam training method based on dynamic hierarchical codebook is proposed.A generalized hierarchical codebook design method based on the summation of beams is first designed.Then based on this method,a beam training method which dynamically updates the hierarchical codebook by removing the contribution of the estimated multipath components(MPCs)from the codebook is proposed.The beam training for multiuser mm Wave massive MIMO is also studied.For the user equipments(UEs)served by the base station(BS),an alternative minimization method with a closed-form expression(AMCF)method is proposed to design the hierarchical codebook under the constant modulus constraint.To speed up the convergence of the AMCF,two initialization methods are proposed,namely AMCF with Zadoff-Chu sequence initialization(AMCF-ZCI)and AMCF with eigenvalue initialization(AMCF-EVI).For the BS,the hierarchical codebook is adaptively designed,where the codewords in the current layer of the hierarchical codebook are determined according to the beam training results of the previous layer.In particular,multimainlobe codewords are designed for the BS to simultaneously perform the beam training with all UEs,where each mainlobe covers a spatial region for one or more UEs.Excluding the bottom layer of the hierarchical codebook,there are only two codewords at each layer in the designed adaptive hierarchical codebook,which only requires two rounds of simultaneous beam training for all the UEs no matter how many UEs the BS serves.