Sampling Limits and Linear Reconstruction for Port Selection in 6G Fluid Antenna Systems
Year: 2026
Type: article
Abstract: Fluid antennas are a promising technology for sixth-generation (6G) networks. A fluid antenna moves a single radiating element among a dense grid of ports and activates the port with the best signal quality. Before it can choose a port, however, the receiver must measure the ports, and in dense-port designs this measurement is the main overhead. This paper uses sampling theory to determine how few ports need to be measured, and it reaches two main results. First, the receiver does not observe the channel itself but the signal-to-interference-plus-noise ratio, which is a nonlinear function of several intensity fields. We derive the exact spatial spectrum of the port field and of its intensity, and we show that the covariance of the observed quantity is a series in the powers of the intensity kernel. Its leading component occupies a spectral disc of twice the radius of the channel spectrum, so the observed field has considerably more degrees of freedom than the channel, and we give a closed-form expression for this number. Sampling the aperture at the Nyquist rate of the channel is therefore not enough, which explains why reconstruction on the channel eigenbasis performs worse than the same reconstruction on the intensity eigenbasis. Second, when the observations are expressed in decibels, the reconstruction problem is close to linear. A matrix fitted by ridge regression matches or outperforms a tuned two-hidden-layer neural network at every observation budget with far fewer multiplications, and it comes within a fraction of a decibel of exhaustive search while measuring fewer than one port in ten. The comparison is reported with confidence intervals and is repeated under unequal user powers, specular and anisotropic scattering, correlated interferers, hardware impairments, and model mismatch. We also give a simple design rule that sets the number of ports to be measured from the aperture size alone. The rule holds for moderate measurement errors, and under large errors the linear predictor even outperforms a receiver that measures every port.
Source: IEEE Access
Language: English
Topic: Antenna Design and Analysis
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W7225259245
