【作者】 李静；

【导师】 岳殿武；

【作者基本信息】 大连海事大学 ， 信息与通信工程， 2021， 博士

【摘要】 毫米波通信是当今5G蜂窝网络的关键技术之一。虽然毫米波有巨大的带宽优势,但是高路径损耗、易遮挡的传播特性以及高频器件的硬件设计和功耗成本制约了它的发展。学术界和产业界对毫米波技术的研究以及大规模MIMO和混合预编码技术的应用使毫米波MIMO通信系统的商用成为了可能。为了减小天线之间的空间相关性对系统性能的影响,论文引入了分布式结构的思想,提出了分布式毫米波大规模MIMO(Distributed Millimeter Wave Massive MIMO,D-MIMO)系统。主要研究内容总结如下:(1)研究了单用户D-MIMO系统的频谱效率。首先,考虑了三种典型的无线接入单元(Radio Access Unit,RAU)分布式结构:无中心RAU圆形分布、有中心RAU圆形分布和随机RAU分布,总结了不同RAU分布时小区内用户频谱效率的变化规律。然后,利用矩阵因式分解算法得到了最优的混合预编码矩阵,提出了基于网格搜索算法的混合预编码器设计方案,其中模拟预编码器具有有限分辨率,数字预编码器满足功率约束条件。仿真结果表明,采用有中心RAU圆形分布的D-MIMO系统性能最佳,而且网格搜索算法与矩阵因式分解算法的性能非常接近。(2)研究了RAU数目已知的多用户D-MIMO系统的频谱效率。该系统采用无中心RAU圆形分布和部分连接的混合预编码结构。首先,提出了基于距离和基于信干噪比的RAU分配算法。然后,使用基于距离的RAU分配算法,推导了用户渐近频谱效率和小区渐近平均频谱效率的上限。为了比较,同时推导了集中式毫米波大规模MIMO系统中用户渐近频谱效率以及小区渐近平均频谱效率的表达式。仿真结果表明,从小区的角度来看,使用基于距离RAU分配算法的D-MIMO系统的性能优于集中式毫米波大规模MIMO系统,而且与使用基于信干噪比RAU分配算法的D-MIMO系统相比,实现相似性能的同时算法的计算复杂度更低。(3)研究了RAU数目未知的多用户D-MIMO系统的频谱效率。首先,为了体现实际场景中RAU部署的空间随机性和排斥性,将RAU的分布建模为泊松点过程(Poisson Point Process,PPP)和Matérn硬核点过程(Matérn Hard-Core Point Process,MHCPP)。然后,当系统采用混合连接的混合预编码结构时,讨论了用户数和RAU数对系统性能的影响。使用基于距离的RAU分配算法,推导了PPP D-MIMO系统和MHCPP D-MIMO系统中用户平均频谱效率的上限。仿真结果表明,在人口密集的场景中,PPP分布的RAU结构比无中心RAU圆形分布更适合D-MIMO系统。同时,当利用PPP D-MIMO系统中使用的PPP生成MHCPP时,具有较小硬核距离的MHCPP D-MIMO系统与PPP D-MIMO系统的性能非常接近。(4)研究了5G NR中分布式混合波束成形的设计问题。首先,总结了不同标准组织提出的5G NR协议栈的物理层内功能划分方案,分析了各种划分方案的特点。然后,根据是否包含数字波束成形功能,将无线单元(Radio Unit,RU)分为两类:A类和B类。基于这两类RU,提出了两种混合波束成形器的实现方式,其中使用A类RU时,数字波束成形在分布式单元(Distributed Unit,DU)中执行,而使用B类RU时,数字波束成形在RU中执行。针对每种实现方式,提出了混合波束成形器、模拟合并器和前传压缩策略的联合设计方案。仿真结果表明,在大功率情况下,数字波束成形在DU中实现时系统的性能更好,而在大功率且对时延敏感的流量密集型网络中,数字波束成形在RU中实现有利于系统在前传带宽、传输时延和性能之间达到更好的平衡。更多还原

【Abstract】 Millimeter wave(mm Wave)communications are considered to be key technologies in current 5G cellular networks.Although there is huge bandwidth available in the mm Wave band,its application is restricted by high path-loss and easily obstructed propagation characteristics as well as the hardware design and power consumption cost of high-frequency devices.The research on mm Wave technology in both academia and industry as well as the application of massive MIMO and hybrid precoding makes the commercial use of mm Wave communications be possible.The idea of distributed architecture is introduced and a distributed mm Wave massive MIMO(D-MIMO)system is proposed to reduce the impact of antenna correlation on system performance.The main research contents are summarized as follows:(1)The spectral efficiency of a single-user D-MIMO system is studied.First,the D-MIMO systems with three typical distributed radio access unit(RAU)layouts,including circular layout without a central RAU,circular layout with a central RAU and random layout,are considered.The changes in spectral efficiencies of users in the cell with different RAU layouts are summarized.Then,an optimal hybrid precoding matrix is obtained through the matrix factorization algorithm,and a trellis exploration algorithm is proposed to get the hybrid precoder,in which the analog precoder has finite resolution and the digital precoder satisfies the power constraint.Simulation results show that,the D-MIMO system using a circular layout with a central RAU performs best.Furthermore,the trellis exploration algorithm achieves a very close performance to the matrix factorization algorithm.(2)The spectral efficiency of a multi-user D-MIMO system is introduced when the number of RAUs is known.The circular layout without a central RAU and partially-connected hybrid precoding structure are adopted in the system.First,two RAU allocation algorithms based on distance and signal-to-interference-noise-ratio(SINR)respectively are proposed.Then,by using the distance-based RAU allocation algorithm,the upper bounds of the user’s asymptotic spectral efficiency and the cell’s asymptotic average spectral efficiency are derived.For comparison,the expressions of the user’s asymptotic spectral efficiency and the cell’s asymptotic average spectral efficiency for a centralized mm Wave massive MIMO(C-MIMO)system are also derived.Numerical results show that,from the perspective of the cell,the D-MIMO system with distance-based scheme outperforms the C-MIMO system and achieves almost alike performance compared with the SINR-based solution while requiring lower computational complexity.(3)The spectral efficiency of a multi-user D-MIMO system is considered when the number of RAUs is unknown.First,taking into account spatial randomness and repulsiveness of RAUs in practical scenarios,the distribution of RAUs is modeled as a Poisson point process(PPP)or a Matérn hard-core point process(MHCPP).Then,when the hybridly-connected hybrid precoding structure is adopted,the impacts of the number of users and the number of RAUs on system performance are discussed.The upper bounds on average spectral efficiencies of the users in the PPP D-MIMO and MHCPP D-MIMO systems are derived by using the distance-based RAU allocation algorithm.Simulation results show that,for highly-populated scenarios,the PPP-distributed RAU layout is more suitable for the D-MIMO system than the circular layout without a central RAU.Moreover,when the MHCPP is generated from the PPP used in the PPP D-MIMO system,the MHCPP D-MIMO system with a low hard-core distance achieves a quite close performance to the PPP D-MIMO system.(4)The design of distributed hybrid beamforming in 5G NR is studied.First,the functional splits in the physical layer proposed by different standard organizations are summarized,and the characteristics of different split options are analyzed.Then,according to whether the digital beamforming function block is included,the remote radio unit(RU)is divided into two categories: Category A and Category B.Based on the two types of RUs,two implementation modes of hybrid beamforming are proposed,where digital beamforming is performed either at the distributed unit(DU),as in the Category A RU,or at the RUs,as in the Category B RU.The joint design schemes of hybrid beamformer,analog combiner and fronthaul compression strategies are respectively developed for each implementation mode.Numerical results show that,in the large power regime,the implementation of digital beamforming at the DU contributes to a better system performance.However,in the large power and delay-sensitive traffic-intensive networks,the implementation of digital beamforming at the RUs helps to balance the tradeoff among fronthaul bandwidth,transmit latency and system performance.更多还原