cGAN-Based Slow Fluid Antenna Multiple Access

Abstract

The emerging fluid antenna system (FAS) technology enables multiple access utilizing deep fades in the spatial domain. This paradigm is known as fluid antenna multiple access (FAMA). Despite conceptual simplicity, the challenge of finding the position (a.k.a. port) that maximizes the signal-to-interference plus noise ratio (SINR) at the FAS receiver output, cannot be overstated. This letter proposes to take only a few SINR observations in the FAS space and infer the SINRs for the missing ports by employing a conditional generative adversarial network (cGAN). With this approach, port selection for FAMA can be performed based on a few SINR observations. Our simulation results illustrate great reductions in the outage probability (OP) with only few observed ports, showcasing the efficacy of our proposed scheme.