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.
@article{arxiv.2507.04435,
title = {Context-Aware Deep Learning for Robust Channel Extrapolation in Fluid Antenna Systems},
author = {Yanliang Jin and Runze Yu and Yuan Gao and Shengli Liu and Xiaoli Chu and Kai-Kit Wong and Chan-Byoung Chae},
journal= {arXiv preprint arXiv:2507.04435},
year = {2026}
}