基于高维随机矩阵的大规模MIMO系统理论及算法研究

Massive MIMO is considered as the core technology of 5G for its better facilitation on space domain and higher spectrum efficiency. But the existing MIMO technology is based on low-dimensional classical statistical signal processing techniques under the assumption of a Gaussian channel, which is difficult to apply to high-dimensional, large-scale structural MIMO system; and the modem of existing MIMO systems is resource-intensive and with high computational complexity, which is difficult to implemented in large-scale high-dimensional MIMO systems. finally, the existing technology is based on a framework which processes large-scale data in one central processor, which is difficult to break the transmission and storage bottlenecks in massive MIMO. This application is intended to use high-dimensional random matrix theory as the theoretical basis, and solve above problem by 3 parts of research: Study the mathematical model of random matrices to model the channel with structural characteristics in real-world, and calculate its performance metrics; Build modulation and demodulation algorithms using modern signal processing methods which is based on high-dimensional statistical theory, and the object is less resource-intensive and real-time processing; Research the distributed system architecture to compatible with data compression and data scketching, which is developed for real-time signal processing. The project is committed to building a system to model the actual channel environment and meet the real-time and efficiency requirements in massive MIMO system.

Massive MIMO能更好利用空域资源，增加系统频谱效率，因此成为5G的核心技术。但现有MIMO技术的理论基础是高斯信道假设下的经典统计和低维信号处理技术，难以应用于以高维、结构化为特征的大规模MIMO系统中；并且现有MIMO系统调制解调需要占用大量资源、计算复杂度高，难以适应大规模MIMO系统对高维数据高效、实时处理的需求；最后，现有MIMO技术基于大规模数据集中处理的技术框架，难以突破大规模MIMO带来的信号传输和存储的瓶颈。本申请拟以高维随机矩阵理论为理论基础，从以下三方面尝试解决问题: 发掘现实信道的结构化特性，研究基于随机矩阵的数学模型来进行信道建模，并计算其性能界；使用基于高维统计理论的现代信号处理方法，构造占用资源少、可实时处理的调制解调算法；使用分布式数据压缩的系统架构，保障信号实时处理。本项目致力于构建更贴近实际信道环境的、满足实时性和效率需求的大规模MIMO系统。

大规模多输入多输出(Multiple Input Multiple Output, MIMO) 是未来移动通信系统的.关键技术。它可以促进无线通信，智能电网和智能健康管理等领域的快速发展。具体而.言，MIMO 系统可以为上述领域提供更多更高维度的测量数据。海量高维的数据对于深.刻地理解，认识和分析这些领域地问题提供了新的机遇。本项目的研究，以高维数据建模和分析关键方法——随机矩阵理论为基础去.解决对应的问题。取得的具体工作和成果包括：..1）以随机矩阵理论为理论基础，分别提出了基于特征值推理的信道相关的未知参数.估计方法和基于旋转不变估计方法的信道重建方法。在上述方法的基础上进一步.提出了基于Bussgang 定理的线性量化预编码方案，解决了非理性信道条件下的.基于低比特数模转换器的大规模MIMO 系统的预编码问题。仿真数据验证了提.出方案的鲁棒性。..2）以非凸优化技术为基础提出了低计算复杂度的非线性预编码技术，证明了所提出.的算法在弱假设条件下具有收敛性，解决了经典的线性量化预编码方法.在中高信噪比条件下存在的性能饱和问题。通过仿真数据验证了所提出方法的计.算效率和可靠性。..3）以随机矩阵理论为理论依据，利用同步向量测量装置测量到的海量数据流，提出.了基于多个高维协方差矩阵检测的异常数据检测方法。通过主元计算和冗余计算.消除等技术进一步提升了所提出方法的计算效率。此外，通过仿真数据和实测数.据验证了上述方法的正确性。