Two-Stage Precoding For 3D Multi-User Massive MIMO Systems

Massive multiple-input multiple-output(MIMO)becomes a promising technology for fifthgeneration wireless communications because the use of a large number of antennas at the base station(BS)provides substantial gains for spectral efficiency and energy efficiency.Compared with linear antenna array,planar antenna array configures massive antennas within acceptable antenna size at the BS,which helps to enhance the system performance by achieving beamforming gains in both azimuth and elevation dimensions.Focusing on the large feedback overhead and computational complexity of three-dimensional(3D)massive MIMO system,we adopt reduced-dimensional channel state information at the transmitter(CSIT)with the aid of channel sparsity and investigate the two-stage precoding for downlink massive multiuser MIMO(MU-MIMO)system.The major contributions of this work are summarized as follows.Considering the sparsity of the 3D massive MIMO channel,we propose a two-stage precoding scheme based on beam selection in scenarios where the number of users is small and the channel orthogonality among users is sufficient.A pre-precoding matrix which depends only on the channel second-order statistics is obtained by a proposed beam selection method,which aims to decrease the channel dimension and the cost of CSIT feedback under the premise of guaranteeing primary channel energy.User Scheduling and a classical multiuser precoding are presented in the second stage to align scheduled users based on the instantaneous knowledge of the resulting effective channel matrix.This scheme can dramatically reduce feedback and complexity on the basis of enough multiuser transmission performance.When the number of users is not much smaller than the number of BS antennas,the orthogonality among users will be destroyed,resulting in serious interference.In this case,a major drawback of the linear precoding is its low performance,and the nonlinear precoding leads to a substantially higher performance with unbearable complexity.Utilizing joint spatial division and multiplexing(JSDM)approach that exploits the structure of the correlation of the channel vectors,we further propose a hybrid linear/non-linear precoding scheme,which partitions the users into groups with approximately similar horizontal or vertical channel covariance eigenvectors and serves these groups by concatenating a linear pre-precoding matrix with a post-nonlinear precoder.The linear pre-precoding matrix is chosen in order to minimize the inter-group interference and reduce the dimension with respect to the full channel matrix by approximate block diagonalization(BD)algorithm.For each group,user scheduling and a nonlinear precoding successively mitigate the intra-group interference,which require the instantaneous effective channel realization.The validity of the proposed hybrid precoding scheme is demonstrated through simulations.These results reveal that this scheme can achieve better performance than the existing multi-user transmission schemes while significantly saving the feedback overhead and computational complexity.