为了满足5G移动通信系统对高速时变大规模MIMO信道传输的可靠性提出的更高要求，本项目提出将时变信道建模于微分流形上，通过深入研究基于微分流形的大规模MIMO发射技术，以解决大规模MIMO有限反馈系统中量化误差、反馈延迟、反馈开销、用户干扰等问题。在深入研究微分流形的理论，分析Grassmannian流形的几何特性，研究Grassmannian流形上最优化算法的基础上，本项目提出利用信道的时域相关性设计高分辨率的动态聚焦码本以降低量化误差、减少反馈量；提出将Grassmannian流形上的最优化算法与预测预编码相结合以减轻反馈延迟对大规模MIMO时域相关性信道下发射预编码性能的影响；提出基于Grassmannian流形上最优化算法的干扰对齐发射预编码以解决分布式大规模MIMO时变干扰信道下的用户干扰问题。本项目的研究工作对于5G大规模MIMO有限反馈系统的设计具有实用价值和发展前景。

In order to meet the higher requirements of 5G mobile communication system for reliability of high-speed time-varying large-scale MIMO channel transmission, this project proposes the transmission technology based on differential manifolds to solve quantization errors, feedback delays, feedback overhead and multiuser interference in large-scale limited feedback MIMO systems. Based on the theory of differential manifold, the geometric characteristics of Grassmannian manifolds and the optimization algorithms on Grassmannian manifolds are studied. By using the time domain correlation of the channel, the high-resolution dynamic focus codebook is designed to reduce the quantization errors and the feedback overhead. In order to reduce the feedback delay for the large-scale MIMO time-domain correlation channel, it is proposed to combine the optimization algorithms on Grassmannian manifolds with the predictive precoding. In order to solve the problem of multiuser interference in distributed large-scale MIMO time-varying interference channels, the interference alignment precoding methods based on the optimization algorithms on Grassmannian manifolds are proposed. The research work of this project is of practical value and development prospect for the design of 5G large-scale MIMO limited feedback system.