Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Context-Aware Deep Learning for Robust Channel Extrapolation in Fluid Antenna Systems

Yanliang Jin, Runze Yu, Yuan Gao, Shengli Liu, Xiaoli Chu, Kai-Kit Wong, Chan-Byoung Chae

TL;DR

Fluid antenna systems (FAS) enable dense spatial diversity but incur substantial CSI overhead. CANet combines a context-adaptive block and cross-scale contextual attention on a ConvNeXt v2 backbone, augmented by a Fourier-domain loss and spatial amplitude perturbation to robustly extrapolate high-resolution CSI from sparse observations. The architecture integrates a local fusion module (CAB) and a global completion module (CSCA), with dropout to enhance generalization. Across simulations, CANet consistently outperforms baselines in NMSE and outage probability, under multiple frequencies and observation ratios, demonstrating strong robustness to noise and practical potential for reducing pilot overhead in FAS deployments.

Abstract

Fluid antenna systems (FAS) offer remarkable spatial flexibility but face significant challenges in acquiring high-resolution channel state information (CSI), leading to considerable overhead. To address this issue, we propose CANet, a robust deep learning model for channel extrapolation in FAS. CANet combines context-adaptive modeling with a cross-scale attention mechanism and is built on a ConvNeXt v2 backbone to improve extrapolation accuracy for unobserved antenna ports. To further enhance robustness, we introduce a novel spatial amplitude perturbation strategy, inspired by frequency-domain augmentation techniques in image processing. This motivates the incorporation of a Fourier-domain loss function, capturing frequency-domain consistency, alongside a spectral structure consistency loss that reinforces learning stability under perturbations. Our simulation results demonstrate that CANet outperforms benchmark models across a wide range of signal-to-noise ratio (SNR) levels.

Context-Aware Deep Learning for Robust Channel Extrapolation in Fluid Antenna Systems

TL;DR

Fluid antenna systems (FAS) enable dense spatial diversity but incur substantial CSI overhead. CANet combines a context-adaptive block and cross-scale contextual attention on a ConvNeXt v2 backbone, augmented by a Fourier-domain loss and spatial amplitude perturbation to robustly extrapolate high-resolution CSI from sparse observations. The architecture integrates a local fusion module (CAB) and a global completion module (CSCA), with dropout to enhance generalization. Across simulations, CANet consistently outperforms baselines in NMSE and outage probability, under multiple frequencies and observation ratios, demonstrating strong robustness to noise and practical potential for reducing pilot overhead in FAS deployments.

Abstract

Fluid antenna systems (FAS) offer remarkable spatial flexibility but face significant challenges in acquiring high-resolution channel state information (CSI), leading to considerable overhead. To address this issue, we propose CANet, a robust deep learning model for channel extrapolation in FAS. CANet combines context-adaptive modeling with a cross-scale attention mechanism and is built on a ConvNeXt v2 backbone to improve extrapolation accuracy for unobserved antenna ports. To further enhance robustness, we introduce a novel spatial amplitude perturbation strategy, inspired by frequency-domain augmentation techniques in image processing. This motivates the incorporation of a Fourier-domain loss function, capturing frequency-domain consistency, alongside a spectral structure consistency loss that reinforces learning stability under perturbations. Our simulation results demonstrate that CANet outperforms benchmark models across a wide range of signal-to-noise ratio (SNR) levels.

Paper Structure

This paper contains 15 sections, 16 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Channel Extrapolation for FAS
  3. Proposed Method
  4. Masking Strategy
  5. Model Architecture
  6. CAB
  7. CSCA
  8. ConvNeXt v2 with dropout regularization
  9. Spatial amplitude perturbation
  10. Loss Function
  11. Numerical Results
  12. Simulation settings
  13. Performance analysis
  14. Ablation experiments
  15. Conclusions

Figures (3)

  • Figure 1: The proposed CANet for CSI extrapolation in FAS channels, where $\oplus$, $\otimes$ and $\odot$ denote fuse, dot product and element-wise product, respectively.
  • Figure 2: NMSE versus the number of observed ports with different operating frequencies. CANet consistently outperforms benchmark models across all observation ratios, demonstrating superior robustness in reconstructing high-resolution CSI from sparse observations.
  • Figure 3: Outage probability versus SINR threshold at a fixed observation ratio of $10\%$ under 0 dB and 20 dB observation SNR. CANet achieves the lowest outage probability among extrapolation schemes, effectively mitigating noise to approach the performance of direct transmission.