Scholarlink

<p>Terahertz (THz) cell-free ultra-massive multiple-input multiple-output (UM-MIMO) systems are a promising technology for future 6G wireless networks due to their ability to support high data rates and low latency. However, accurate channel state information (CSI) estimation remains challenging because of large antenna dimensions, severe propagation losses, and hardware limitations. In this work, a convolutional neural network (CNN)-based approach is proposed to estimate CSI from pilot observations in THz environments. The model leverages spatial and frequency correlations in the channel to improve estimation accuracy compared to conventional techniques. Simulation results show that the proposed method achieves an NMSE improvement of approximately 12.7&#xa0;dB over the least squares (LS) estimator, along with lower bit error rates. The model also supports low-latency operation with an average inference time of 3.4&#xa0;ms and achieves a throughput of up to 22.7 Gbps under the considered conditions. Overall, the results indicate that the proposed method is effective for CSI estimation and suitable for real-time implementation in THz cell-free UM-MIMO systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CNN-Aided CSI Estimation for Real-Time Machine Learning Precoding in Terahertz Cell-Free UM-MIMO Systems

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Ch. JyotsnaRani,
  • V. Malleswara Rao

摘要

Terahertz (THz) cell-free ultra-massive multiple-input multiple-output (UM-MIMO) systems are a promising technology for future 6G wireless networks due to their ability to support high data rates and low latency. However, accurate channel state information (CSI) estimation remains challenging because of large antenna dimensions, severe propagation losses, and hardware limitations. In this work, a convolutional neural network (CNN)-based approach is proposed to estimate CSI from pilot observations in THz environments. The model leverages spatial and frequency correlations in the channel to improve estimation accuracy compared to conventional techniques. Simulation results show that the proposed method achieves an NMSE improvement of approximately 12.7 dB over the least squares (LS) estimator, along with lower bit error rates. The model also supports low-latency operation with an average inference time of 3.4 ms and achieves a throughput of up to 22.7 Gbps under the considered conditions. Overall, the results indicate that the proposed method is effective for CSI estimation and suitable for real-time implementation in THz cell-free UM-MIMO systems.