| 摘要: |
| 随着6G技术对数据传输速率的高要求，超大规模多输入多输出(Extremely Large-scale Multiple-Input Multiple-Output,XL-MIMO)系统的应用前景广泛。然而，XL-MIMO系统中的近场球面波前特性使得传统基于平面波前的信道估计方法不再适用，从而影响信道状态信息(Channel State Information,CSI)的获取。为了解决这一问题，提出了一种新的信道估计方案，分析了XL-MIMO的近场球面波特性，并将其与适用于高速移动场景的正交时频空间(Orthogonal Time Frequency Space,OTFS)调制技术结合，构建了一个考虑球面波特性的XL-MIMO OTFS系统框架。此外，利用模型驱动的思想设计了一种可学习的去噪近似消息传递(Learned Denoising Based Approximate Message Passing,LDAMP)算法，用于有效估计信道状态信息。仿真结果表明，所提出的XL-MIMO OTFS信道估计框架在近场高速移动场景下具有显著的性能提升和良好的鲁棒性。在信噪比为20 dB时，LDAMP算法的归一化均方误差达到10－2，相比传统算法提高了多个数量级。 |
| 关键词: 近场球面波 超大规模多输入多输出 正交时频空间 信道估计 深度学习 模型驱动 |
| DOI：10.20079/j.issn.1001-893x.240910001 |
|
| 基金项目:国家自然科学基金资助项目（61701062） |
|
| Channel Estimation for Extremely Large-scale MIMO-OTFS System with Near-field Spherical Wave |
| PU Xumin,XU Peng,WANG Kehao |
| (School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China) |
| Abstract: |
| With the increasing demands of 6G technology for high data transmission rates,extremely large-scale multiple-input multiple-output(XL-MIMO) systems hold significant potential for future applications.However,the near-field spherical wave properties in XL-MIMO systems make traditional channel estimation methods based on plane waves no longer applicable,thereby affecting the acquisition of channel state information(CSI).To address this issue,a novel channel estimation framework is proposed.The near-field spherical wave properties of XL-MIMO are analyzed and combined with orthogonal time frequency space(OTFS) modulation technology,which is suitable for high-speed mobility scenarios,to construct an XL-MIMO OTFS system that considers spherical wave properties.Additionally,a model-driven approach is used to design a learned denoising based approximate message passing(LDAMP) algorithm to effectively estimate the CSI.Simulation results demonstrate that the proposed XL-MIMO OTFS channel estimation framework offers significant performance improvements and good robustness in near-field high-speed mobility scenarios.The LDAMP algorithm achieves a normalized mean squared error of 10－2 at an signal-to-noise ratio of 20 dB,several orders of magnitude better than that of traditional algorithms. |
| Key words: near-field spherical wave XL-MIMO OTFS channel estimation deep learning model-driven |
