在无线通信领域，毫米波大规模MIMO技术首次同时将大型的天线阵列与极大的传输带宽相结合。此类系统中，电磁波跨越天线阵列两端的时间可以达到符号周期的量级，在部署大规模阵列的接收端可以在不同天线对同一信道径观测到不同的符号。该效应为宽带毫米波大规模MIMO 系统所特有，被称为“空间宽带效应”或“孔径效应”。本文围绕空间宽带效应，针对毫米波大规模MIMO通信系统中一些特有的关键问题进行研究，具体贡献如下：1. 首先，基于实际的物理路径，通过阵列信号处理理论，对传统基于角度域物理路径的大规模MIMO信道模型进行了修正，提出了可以精确描述空间宽带效应的毫米波大规模MIMO信道模型，研究并证明了由空间宽带效应导致的一系列特殊的信道性质。进一步地，讨论了空间宽带效应在不同域（空间域、频率域、角度域、时延域）下信道的不同表现形式，阐明了空间宽带效应与其导致的在角度-频率域波束偏移（beam squint）现象的关系。2. 针对传统基于物理角度域或虚拟角度域的信道估计算法未考虑空间宽带效应引起的性能损失问题，基于前面推导的信道模型，提出了充分考虑空间宽带效应的两种信道估计算法，分别为基于快速傅里叶变换（FFT）的低复杂度信道估计算法和基于压缩感知的高精度信道估计算法。与传统两种典型估计算法相比，所提算法在天线数少和带宽较小的场景下与传统算法表现一致，在天线数多和带宽较大的场景下的性能较传统算法提升了两个数量级。3. 针对空间宽带效应在二维虚拟角度-时延域造成的功率泄露问题，提出了一种基于窗函数的解决方案。基于最大化给定区域信道能量的目标法则，设计了能够最小化功率泄露的二维空-频联合窗函数，缓解了功率泄露问题，并保留了毫米波信道在二维虚拟角度-时延域的稀疏性。4. 因毫米波大规模MIMO系统的高硬件成本和高功耗，实际系统普遍采用混合模拟/数字预编码（简称混合预编码）等架构，前述信道估计算法和二维窗函数不能直接应用于该架构。针对该问题，对前述信道估计算法进行推广，使其适用于混合预编码系统；并对前述二维窗函数进行一维降维近似，提出了一种窗函数在混合预编码架构下基于硬件的实现方案。本文相关研究成果可望在毫米波大规模MIMO系统实用化的道路上提供必要的理论指导。
The millimeter-wave (mmWave) massive multiple-input multiple-output (MIMO) technology first employs both the large antenna array and the large bandwidth in wireless communications, where the physical propagation delays of electromagnetic waves travelling across the whole array will become large and comparable to the time-domain sample period. This phenomenon is called spatial-wideband effect and is a fundamental property of mmWave massive MIMO systems. This thesis focuses on the spatial-wideband effect and investigates certain specific issues in mmWave massive MIMO. The contributions of this thesis are summarized as follows.1. From the array signal processing point of view, we propose a revised mmWave massive MIMO channel model based on the physical angles and delays, and investigate certain particular channel characteristics resulting from the spatial-wideband effect. The different manifestations of the spatial-wideband effect in different domains, e.g., the spatial domain, the frequency domain, the angle domain, and the delay domain, are also discussed, including the relationship between the spatial-wideband effect and beam squint.2. To address the performance degradation issue in the channel estimation algorithms that use the conventional angle domain-based channel models and ignore the spatial-wideband effect, we develop two channel estimation algorithms, one of which is based on fast Fourier transform (FFT) and has low computational complexity while the other one is based on compressive sensing (CS) and achieves the high estimation accuracy. Compared with the conventional algorithms, the proposed ones have more than two orders-of-magnitude improvement in certain scenarios with large antenna arrays and large bandwidths.3. To eliminate the power leakage in angle-delay domain caused by the spatial-wideband effect, an optimal energy-focusing window is designed to shape the channel in spatial-frequency domain to maximally concentrate the energy of each multipath component inside the corresponding square area in angle-delay domain. With the help of this window, the power leakage can be significantly reduced. As a result, the sparsity of mmWave channels can be primely preserved and the inter-user interference can be significantly alleviated for angle- and delay-based channel estimation and precoding.4. We further generalize the proposed CS-based channel estimation algorithm and the energy-focusing window approach to the hybrid analog/digital precoding architecture, which is popular in mmWave massive MIMO systems to reduce the high hardware cost and high energy consumption.
