2026 Fotovat Fluid Antenna Port Selection Pilot Allocation

Fluid antenna systems (FAS) provide spatial degrees of freedom through the selective activation of radiating ports from a dense candidate grid, although traditional port selection approaches require complete channel state information (CSI) at all ports. This study innovatively formulates port selection and pilot allocation as a joint optimization problem using active inference. A spatio-temporal generative model integrates sub-wavelength spatial correlation with an autoregressive temporal prior, generating a Gaussian belief over N port channels. The transmitter strategically selects both the operational ports and the pilots, optimizing for expected free energy that balances achievable rate, information gain, and port-switching costs, achieving a greedy and submodular solution with a computational complexity of O(NM) per time slot. In a simulation involving a 441-port grid with a hybrid front end, the proposed method approaches 91% of the achievable sum rate as compared to a full-CSI benchmark while only requiring measurements from 2.3% of the candidate ports. Moreover, it allows for pilot withdrawals from active ports, demonstrating efficiency in rate maintenance with minimal pilot overhead. This approach negates the need for prior training.