Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lightweight Deep Learning Based Channel Estimation for Extremely Large-Scale Massive MIMO Systems

Shen Gao, Peihao Dong, Zhiwen Pan, Xiaohu You

TL;DR

This work tackles XL-MIMO channel estimation under severe dimensionality and near-field propagation by introducing XLCNet, a universal DL-based estimator trained on a hybrid-field model so it applies to both near-field and far-field users without explicit prior statistics. The method preprocesses received pilots into a coarse LS-like estimate and refines it via a two-branch 2D CNN residual denoiser, trained with data generated from a mixed near-field/far-field channel model. A lightweight variant, C-XLCNet, uses magnitude-based weight pruning and post-training quantization to reduce model size by up to ~36x and compute by ~10x with only minor performance loss, supported by a complexity analysis showing clear reductions in FLOPs proportional to pruning. Simulation results demonstrate XLCNet’s superiority over conventional estimators and the robustness of the approach to propagation regime mismatches, indicating strong practical potential for fast, scalable CSI acquisition in 6G XL-MIMO deployments.

Abstract

Extremely large-scale massive multiple-input multiple-output (XL-MIMO) systems introduce the much higher channel dimensionality and incur the additional near-field propagation effect, aggravating the computation load and the difficulty to acquire the prior knowledge for channel estimation. In this article, an XL-MIMO channel network (XLCNet) is developed to estimate the high-dimensional channel, which is a universal solution for both the near-field users and far-field users with different channel statistics. Furthermore, a compressed XLCNet (C-XLCNet) is designed via weight pruning and quantization to accelerate the model inference as well as to facilitate the model storage and transmission. Simulation results show the performance superiority and universality of XLCNet. Compared to XLCNet, C-XLCNet incurs the limited performance loss while reducing the computational complexity and model size by about $10 \times$ and $36 \times$, respectively.

Lightweight Deep Learning Based Channel Estimation for Extremely Large-Scale Massive MIMO Systems

TL;DR

This work tackles XL-MIMO channel estimation under severe dimensionality and near-field propagation by introducing XLCNet, a universal DL-based estimator trained on a hybrid-field model so it applies to both near-field and far-field users without explicit prior statistics. The method preprocesses received pilots into a coarse LS-like estimate and refines it via a two-branch 2D CNN residual denoiser, trained with data generated from a mixed near-field/far-field channel model. A lightweight variant, C-XLCNet, uses magnitude-based weight pruning and post-training quantization to reduce model size by up to ~36x and compute by ~10x with only minor performance loss, supported by a complexity analysis showing clear reductions in FLOPs proportional to pruning. Simulation results demonstrate XLCNet’s superiority over conventional estimators and the robustness of the approach to propagation regime mismatches, indicating strong practical potential for fast, scalable CSI acquisition in 6G XL-MIMO deployments.

Abstract

Extremely large-scale massive multiple-input multiple-output (XL-MIMO) systems introduce the much higher channel dimensionality and incur the additional near-field propagation effect, aggravating the computation load and the difficulty to acquire the prior knowledge for channel estimation. In this article, an XL-MIMO channel network (XLCNet) is developed to estimate the high-dimensional channel, which is a universal solution for both the near-field users and far-field users with different channel statistics. Furthermore, a compressed XLCNet (C-XLCNet) is designed via weight pruning and quantization to accelerate the model inference as well as to facilitate the model storage and transmission. Simulation results show the performance superiority and universality of XLCNet. Compared to XLCNet, C-XLCNet incurs the limited performance loss while reducing the computational complexity and model size by about and , respectively.
Paper Structure (14 sections, 19 equations, 5 figures, 1 table)

This paper contains 14 sections, 19 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. System Model
  3. Signal Transmission Model
  4. Channel Model
  5. XLCNet Design
  6. Preprocessing
  7. XLCNet Architecture
  8. XLCNet Training
  9. Lightweight Design of XLCNet
  10. Weight Pruning
  11. Weight Quantization
  12. Complexity Analysis
  13. Simulation Results
  14. Conclusion

Figures (5)

  • Figure 1: A multi-user XL-MIMO system.
  • Figure 2: Workflow of the lightweight DL based channel estimation framework.
  • Figure 3: NMSE versus SNR for the near-field user.
  • Figure 4: NMSE versus SNR for the far-field user.
  • Figure 5: NMSE versus bit-width $b$ for the near-field user with $L=3$.