Low-Complexity CSI Acquisition Exploiting Geographical Diversity in Fluid Antenna System

The fluid antenna system (FAS) employs reconfigurable antennas for high spatial gains in compact spaces, enhancing physical layer flexibility. Channel state information (CSI) acquisition is vital for port selection and FAS optimization. Greedy algorithms rely on signal assumptions, and model-free methods face high complexity. A flexible, low-complexity solution is needed for massive connectivity in FAS. Based on expectation maximization-approximate message passing (EM-AMP) framework, efficient matrix computations and adaptive learning without prior model knowledge naturally suit CSI acquisition for FAS. We propose a EM-AMP variant exploiting FAS geographical priors, improving estimation precision, accelerating convergence, and reducing complexity in large-scale deployment. Simulations validate the efficacy of the proposed algorithm.