Learned-Approximate Message Passing Under Karhunen–Loève Modeling for Fluid Antenna Systems

作者：Yuan-Hui Wu, Zhen-Tian Zhang, Hao Jiang, K. Wong, Chan-Byoung Chae · 发表于：IEEE Wireless Communications Letters · 年份：2026 · DOI：10.1109/lwc.2026.3682568 · 被引用次数：11 · 研究领域：Computer Science

Fluid antenna systems (FASs) can harvest substantial spatial diversity with a limited number of radio-frequency (RF) chains, yet reliable channel acquisition remains challenging due to the large number of candidate ports and the scarcity of pilot resources. This letter proposes a physics-guided compressive channel reconstruction framework for FAS channel estimation. We first construct a Karhunen-Loève (KL) basis from angle-of-arrival (AoA) statistics, yielding an information-theoretically optimal low-dimensional representation of the FAS channel. To enable pilot acquisition with limited RF resources, we further design a block-hopping sampling strategy that efficiently probes the spatial aperture. Leveraging the resulting KL-domain model, we develop a KL-domain learned approximate message passing (KL-LAMP) network, which unfolds AMP iterations and introduces learnable parameters to enhance robustness against noise and model mismatch. Simulation results show that the proposed method achieves competitive NMSE performance in the medium-to-high signal-to-noise ratio (SNR) regime and offers a favorable performance-complexity tradeoff. Ablation studies corroborate the contribution of each module to the overall performance gains.