Sparse Fluid Antenna Arrays: Continuous Position Design for DOA Estimation
Abstract
This paper studies continuous antenna-position design for direction-of-arrival (DOA) estimation with sparse fluid antenna systems (FAS). We first show that an <inline-formula><tex-math notation="LaTeX">$N$</tex-math></inline-formula>-position array, whether grid-constrained or continuous, has at most <inline-formula><tex-math notation="LaTeX">$N(N-1)+1$</tex-math></inline-formula> distinct signed pairwise differences. Thus, continuous positioning does not bypass the universal <inline-formula><tex-math notation="LaTeX">$O(N^{2})$</tex-math></inline-formula> bound on exact contiguous difference-coarray degrees of freedom (DOFs); the quantity <inline-formula><tex-math notation="LaTeX">$2\left\lfloor D/d_{0}\right\rfloor + 1$</tex-math></inline-formula>, where <inline-formula><tex-math notation="LaTeX">$d_{0}=\lambda/2$</tex-math></inline-formula>, is a geometric ceiling imposed by the deployment interval <inline-formula><tex-math notation="LaTeX">$[0,D]$</tex-math></inline-formula>. The benefit of FAS is instead runtime geometry reconfiguration within a prescribed aperture and hardware constraints. The position-design analysis is not mathematically exclusive to FAS and also applies to calibrated movable and synthetic-aperture implementations satisfying the same coherent observation model. Relative to a particular canonical fixed layout, continuous optimization may improve distinct-lag utilization, Fisher information, or estimation accuracy, but any such improvement is construction-, objective-, and source-dependent rather than a new universal DOF limit. We derive the universal coarray bound, formulate Fisher-information-based position design, and obtain a closed-form single-source design and a Frank–Wolfe procedure for multiple sources. For the multi-source conditional model, an exact normalized-geometry factorization shows that, under uniform identifiability, <inline-formula><tex-math notation="LaTeX">$\det\mathbf{F}=\Theta(D^{2L})$</tex-math></inline-formula> and <inline-formula><tex-math notation="LaTeX">$\det(\mathbf{F}^{-1})=\Theta(D^{-2L})$</tex-math></inline-formula>, whereas each individual DOA variance scales as <inline-formula><tex-math notation="LaTeX">$\Theta(D^{-2})$</tex-math></inline-formula>. We specialize the classical MUSIC-initialized local maximum-likelihood strategy to an ambiguity-aware coarse/fine reconfigurable geometry. The numerical evaluation comprises six unequal-aperture studies, a nonideal-acquisition sensitivity study, and a matched-resource comparison using the same <inline-formula><tex-math notation="LaTeX">$N$</tex-math></inline-formula>, realized aperture <inline-formula><tex-math notation="LaTeX">$D$</tex-math></inline-formula>, minimum spacing, coarse initializer, and local ML refinement. Under this matched setting, the optimized continuous and half-wavelength-grid layouts perform nearly identically for the tested source pair, while both improve upon the matched canonical geometries. This result separates the modest off-grid geometry gain from the larger aperture and estimator effects present in unequal-aperture comparisons.