fluid antenna reconfiguration

To address the challenges of tightly coupled user activity detection and channel estimation, strong reliance on high-SNR assumptions, and high computational complexity in fluid antenna systems (FAS) under high-dimensional sparse channels, this paper proposes an adaptive EM-AMP framework integrating geographic and angular priors. Unlike conventional model-driven approaches, the method avoids pre-specified channel models; instead, it jointly exploits angular-domain parameterization and geometric modeling to characterize spatial sparsity, while adaptively learning hyperparameters within the AMP iterations. Theoretically, we establish that angular information fundamentally determines the Cramér–Rao lower bound of estimation performance. Experiments demonstrate that, particularly under large active regions, the proposed method achieves significantly improved estimation accuracy, converges in fewer than ten iterations, and reduces computational complexity by an order of magnitude—outperforming both state-of-the-art model-based and model-free methods across all metrics.

This study addresses the challenge that conventional channel models fail to accurately capture the dynamic port configuration and spatially non-uniform near-field characteristics of Fluid Antenna Systems (FAS) in unmanned aerial vehicle (UAV) near-field communications. To this end, it presents the first unified model that jointly characterizes the dynamic activation of FAS ports and the spatial non-uniformity of the near-field channel, incorporating both line-of-sight and non-line-of-sight components along with UAV mobility dynamics. Building upon this model, a low-complexity greedy subarray partitioning scheme coupled with a high-gain port selection mechanism is proposed to enable efficient port grouping and real-time updates. The proposed approach significantly enhances channel modeling accuracy while reducing computational complexity, thereby demonstrating the performance advantages and practical viability of FAS in dynamic UAV communication scenarios.

This study investigates the outage performance and diversity gain of fluid antenna systems (FAS) over spatially correlated Rayleigh fading channels, focusing on multiple-input single-output FAS (MISO-FAS) and dual-end single-antenna (Dual-FAS) configurations. By leveraging stochastic process modeling, probabilistic analysis, and diversity theory, the work derives—for the first time—exact closed-form expressions for the outage probability of both system types under channel correlation, along with high-SNR approximations. The analysis further reveals their respective diversity orders. Results demonstrate that increasing the number of ports significantly enhances performance, that lower spatial correlation yields greater gains, and that Dual-FAS outperforms MISO-FAS in the high-SNR regime. This work provides a theoretical foundation and practical design guidance for evaluating FAS performance in realistic correlated channel environments.

To address the high computational complexity of acquiring high-dimensional channel state information (CSI) in fluid antenna systems (FAS), which hinders massive connectivity, this paper proposes a low-complexity CSI estimation algorithm that incorporates geographic distribution priors. Within the expectation-maximization approximate message passing (EM-AMP) framework, we embed an adaptive geographic prior modeling mechanism—eliminating the need for predefined channel models or sparsity assumptions—and achieve end-to-end, model-free sparse signal recovery. By leveraging efficient matrix operations and iterative convergence acceleration, the method simultaneously reduces computational complexity and improves estimation accuracy. Simulation results under large-scale FAS deployments demonstrate that the proposed algorithm achieves over 35% lower CSI mean-square error and approximately 60% shorter runtime compared to conventional EM-AMP and orthogonal matching pursuit (OMP), significantly enhancing port selection and system optimization efficiency.

This study addresses the impact of geometric diversity in planar fluid antenna arrays under finite apertures on channel estimation performance. The authors develop an analytical framework to investigate the statistical properties of minimum inter-port spacing in random layouts and its inherent trade-off with spatial ambiguity. Leveraging stochastic geometry, Cramér–Rao bound analysis, and eigenstructure characterization of associated matrices, they demonstrate that the expected minimum spacing scales as 𝒪(M⁻¹). A geometric inertia matrix is introduced to quantify joint elevation–azimuth estimation accuracy, whose trace and determinant are theoretically shown to be independent of azimuth angle. Furthermore, placing ports near aperture boundaries enhances estimation precision but exacerbates sidelobe-induced ambiguity, thereby establishing—for the first time—a quantitative relationship among array geometry, estimation performance, and ambiguity effects.

This work addresses the high hardware complexity and cost associated with conventional multi-port receivers in fluid antenna multiple access systems, which stem from the need for numerous radio frequency (RF) chains. To overcome this challenge, a hybrid multi-port receiver architecture is proposed, leveraging a low-complexity analog beamforming network to decouple port selection from signal combining. By integrating a generalized eigenvector-based port selection algorithm with a tailored stopping criterion, the architecture achieves performance approaching that of fully digital schemes while utilizing only two RF chains. The proposed approach reduces computational overhead by more than 60%, effectively balancing hardware efficiency and receiver performance.

This work addresses the port selection problem for multi-port receivers in Fluid Antenna Multiple Access (FAMA) systems, where existing approaches struggle to balance spectral efficiency and computational complexity. To tackle this challenge, two complementary methods are proposed: first, a greedy forward selection algorithm enhanced with swap-based refinement (GFwd+S) that significantly improves spectral efficiency; second, a novel application of the Transformer architecture to this task, which leverages imitation learning for pretraining and Reinforce-based policy gradient fine-tuning to achieve performance close to GFwd+S while drastically reducing computational overhead. Together, these approaches establish a new trade-off between performance and complexity, substantially enhancing the practicality of FAMA systems.

This study addresses the challenge of suppressing peak sidelobe level (PSLL) in sparse planar fluid antenna arrays under stringent sparsity constraints. To this end, an improved genetic algorithm (IGA) is proposed that optimizes the port activation pattern to significantly reduce PSLL while preserving the mainlobe width. The IGA integrates tournament selection, adaptive operator probabilities, hybrid crossover, multi-point mutation, and an elitist pool retention strategy to enhance both convergence speed and solution quality. Experimental results demonstrate that, compared to conventional genetic algorithms, the proposed method achieves a 4.45 dB reduction in PSLL with nearly identical mainlobe width, thereby confirming its superior performance in pattern synthesis for sparse arrays.

This work addresses the challenges in compact-aperture fluid antenna arrays, where channel-driven port placement often leads to port clustering, exacerbated mutual coupling, and uneven current loading. To overcome these issues, the authors propose an electromagnetically guided graph neural network framework that, for the first time, integrates electromagnetic constraints—such as mutual impedance and geometric layout—into graph-based learning. This approach jointly optimizes port configuration and current-domain beamforming strategies under a unified electromagnetic feasibility criterion. By doing so, it enables a controllable trade-off among communication rate, current distribution uniformity, and configuration latency, thereby significantly enhancing the overall performance of multiuser downlink transmissions.

This work addresses the limited angular sensing performance of conventional arrays and reconfigurable surface-assisted integrated sensing and communication systems, which struggle to exploit the advantages of large-scale electromagnetic apertures. Focusing on the emerging Enormous Fluid Antenna System (E-FAS), the study develops a bidirectional sensing channel model encompassing surface wave routing, distributed reradiation, target scattering, and echo propagation, and establishes a parametric observation framework. The Fisher information matrix and corresponding Cramér–Rao bound for angle estimation are derived, revealing a fundamental trade-off between surface wave routing gain and sensing diversity in programmable environments. The results demonstrate that E-FAS significantly enhances angular estimation accuracy under identical transmit power, validating the efficacy of jointly optimizing propagation paths and sensing functionality, and thereby positioning E-FAS as a novel paradigm for integrated sensing and communications.