Port Selection in Fluid Antennas: GP Interpolation with Spatial Jitter Modeling

Fluid antenna system (FAS) achieves spatial diversity by selecting the best radiation port among many closely spaced candidates. However, measuring all ports leads to excessive pilot overhead, switching latency, and implementation complexity. This paper proposes a Gaussian processes (GP)-based spatial interpolation framework that estimates the unmeasured port responses from a small subset of observations, enabling near- optimal port selection with low spatial over-sampling. The model also accounts for antenna position jitter by treating positional deviations as uncertainty in the sampling locations. Simulation results demonstrate that the proposed method achieves performance close to the ideal case with perfect channel knowledge, while requiring significantly fewer measurements. Moreover, it consistently outperforms direct port selection and even exhaustive scanning schemes, highlighting the efficiency and robustness of the proposed technique.