大规模3D MIMO系统基于量子神经网络的信道建模及其稀疏估计方法研究

Massive MIMO technique, combined with three-dimensional beamforming technology can further increase cell coverage and reduce inter-cell interference effectively. It can lead to an increase in the capacity and spectral efficiency of the system. 3D Massive MIMO technology is a promising solution for the ultra-high demands of traffic, data rate and link reliability in future mobile communications. This project focuses on several key technologies in 3D massive MIMO. Channel modeling, pilot design, and channel estimation in 3D Massive MIMO system will be intensively investigated. Specific study contents include: ①We will focus our researches on 3D MIMO channel modeling, and try to establish statistical models for 3D massive MIMO based on multivariate statistical methods. We will try to propose novel synthesis methods for the fast calculation of coefficients in the channel fading correlation matrix based on function approximation theory. Try to set up 3D massive MIMO array domain fading correlation models based on spherical wave theory. ②We will develop a method to find the optimal pilot length using quantum bacterial foraging optimization algorithm. In order to decrease pilot pollution caused by multi-cell pilot multiplexing, an intelligent pilot allocation scheme based on quantum transition neural network will be constructed according to the criteria of maximum mutual correlation minimization. ③The block sparse characteristics of 3D massive MIMO in the spatial -time-frequency domain will be analyzed. Due to highly parallel computing characteristic and deep learning capability of Quantum Neural Networks (QNN), we will develop a new sparse channel estimation method for 3D massive MIMO system based on QNN and the designed optimal pilot. The new method will be designed based on a feedback quantum neuron (FQN) model proposed by our team. The performance of the new sparse channel estimation method based on FQN will be discussed to solve the computational complexity of 3D massive MIMO channels estimation.

大规模MIMO技术结合3D波束赋形技术能进一步增大小区覆盖范围，能更有效避免小区间干扰、增加系统容量、提高频谱效率，以满足未来移动通信超大流量、超高速率、更高链路可靠性持续发展的需求。本项目拟深入研究大规模3D MIMO信道建模、导频设计和信道估计等关键技术，具体研究内容有：①基于多元统计方法建立大规模3D MIMO信道统计表征模型；基于函数逼近论提出信道衰落相关性矩阵系数的综合快速计算方法；基于球面波理论建立大规模3D MIMO阵列域衰落相关性模型；②采用量子菌群优化算法设计最优导频长度，针对多小区导频复用造成的导频污染，根据最大互相关最小化等准则构建基于量子跃迁神经网络的智能导频分配方案；③从空-时-频角度联合分析大规模3D MIMO信道的块稀疏特性，基于设计出的最优导频，利用量子神经网络高度并行计算和深度学习能力，提出基于反馈型量子神经元模型的大规模3D MIMO信道稀疏估计新方法。

大规模3D MIMO技术能更进一步增加系统容量、提高频谱效率，更好地满足B5G乃至未来6G移动通信巨大流量、极高速率、极高可靠性的需求。本项目按照项目计划对3D MIMO系统的信道建模、导频优化和信道估计展开了深入研究，并延展对发射预编码和接收信号检测展开了研究，形成了一套完整的3D MIMO收发系统的物理层信号处理方案，在有效降低导频污染、增加系统吞吐量、提高频谱效率、降低运算复杂度等方面提出了新的理论模型和新的方法。主要解决的关键问题与创新点包括：.（1）提出了一种新型3D MIMO信道建模方法，推导出3D MIMO信道发射端和接收端相关矩阵的闭合表达式，建立不同场景下大规模3D MIMO多径衰落信道模型和基于量子菌群算法的系统和速率优化理论。.（2）提出了3D MIMO导频系列优化方法：①去蜂窝大规模MIMO基于量子菌群优化算法的导频分配方案；②基于K-means聚类的软导频复用方案；③基于小区间多用户干扰和的低复杂度导频分配方案；④基于QBFO算法的导频分配方案。这些方法有效降低了3D MIMO系统导频污染问题。.（3）提出了基于3D MIMO信道角域稀疏结构的信道估计方法，包括①基于量子菌群优化算法和自适应相位旋转的三维MIMO系统稀疏信道估计方案；②用于MIMO检测的模型驱动深度学习网络BGS-Net；③时变3D MIMO系统中基于深度学习的信道估计与反馈方案。这些方法从不同角度解决了3D MIMO信道估计问题，并有效提高了信道估计精度。.（4）提出了三种新型3D MIMO混合预编码方法：①大规模3D MIMO系统中基于Kronecker积的有限反馈混合预编码；②基于广义线性实值信号模型的混合预编码方案；③毫米波大规模MIMO下的免导频设计和信道估计的低复杂度混合预编码方案。.（5）提出了①大规模3D MIMO系统中基于矩阵近似求逆的信号检测；②多模式信道大规模MIMO系统中基于持续学习的信号检测方案；③IRS辅助大规模MIMO系统的模型和数据混合驱动检测方案，有效解决了3D MIMO接收信号检测难题。.本项目超额完成了预期研究计划，取得了一系列创新成果，为5G发展和广泛应用提供了理论依据和技术支持。