基于数据驱动稀疏信号处理的波束成形理论与关键技术研究

Beamforming can be used in antenna array to from high-gain narrow beams by weighted combinations of signals in such a way that signals at particular angles experience constructive interference while others experience destructive interference, which can be applied to millimetre-wave communication systems to compensate for excessive path loss, and divide space more finely to serve more users with the same time-frequency resources, so as to improve the reliability and effectiveness of communication systems. However, due to the high carrier frequency of millimetre-wave communication system and the movement of users, the time-varying characteristics of the channel are more obvious, which brings great challenges to millimetre-wave mobile communication. Therefore, this project aims to study the theory and key technologies of beamforming based on data-driven sparse signal processing. Through the research of data-driven signal processing theory and methods, the cognitive and learning ability of wireless communication network is enhanced, so as to break through the limitations of traditional model-based methods, and solve the key technical issues in beamforming. By employing the sparse property of millimetre-wave channels and on-line data analysis and learning, we aim to carry out research on high-resolution channel estimation based on multiple measurement vectors, model-free adaptive beam tracking, and low-complexity precoder design based on low-rank matrix approximation to improve the effectiveness and robustness of beamforming for high quality millimeter-wave mobile transmission.

波束成形通过对多天线阵元接收到的各路信号的加权合成，可以形成高增益的窄波束，从而可应用到毫米波通信系统以弥补过高的路径损耗，并对空间进行更精细的划分以用相同的时频资源服务更多用户，提高通信系统的可靠性和有效性。然而由于毫米波通信系统的高载波频率以及用户的移动，使得信道的时变特性更加明显，这给毫米波移动通信带来了极大挑战。为此，本项目拟开展基于数据驱动稀疏信号处理的波束成形理论与关键技术研究，通过数据驱动的信号处理理论与方法的研究，增强无线通信网络的认知和学习能力，突破传统基于模型方法的局限性，解决波束成形中的关键问题。基于毫米波信道的稀疏特性，本项目拟通过在线数据的分析和学习挖掘数据中的有效信息，从而实现多重测量向量的高分辨率信道估计、无模型自适应控制的动态波束跟踪以及基于低秩矩阵近似的低复杂度预编码器设计，提升波束成形的有效性与鲁棒性，为毫米波移动传输提供更高的通信质量。

由于毫米波通信系统的高载波频率以及用户的移动，使得信道的时变特性更加明显，给毫米波移动通信带来了极大挑战。对于低时延高速率的通信系统而言，如何快速跟踪信道变化是毫米波通信的一个重要难题。为此，本项目开展了基于数据驱动稀疏信号处理的波束成形理论与关键技术研究。通过基于无模型动态梯度理论的数据驱动算法与稀疏信号处理的低秩矩阵恢复理论与方法研究，增强了无线通信网络的认知和学习能力。基于毫米波信道的稀疏特性，通过在线数据的分析和学习挖掘数据中的有效信息，从而实现了基于联合稀疏的高分辨率信道估计，无模型自适应控制的动态波束跟踪以及低复杂度预编码器设计，提升了波束成形的有效性与鲁棒性，为毫米波移动传输提供更高的通信质量。相关研究成果在国际期刊和会议发表学术论文6篇；申请发明专利5项，其中授权发明专利3项；培养硕士研究生5人。