Physics-Aware Tensor Reconstruction for Radio Maps in Pixel-Based Fluid Antenna Systems

The deployment of pixel-based antennas and fluid antenna systems (FAS) is hindered by prohibitive channel state information (CSI) acquisition overhead. While radio maps enable proactive mode selection, reconstructing high-fidelity maps from sparse measurements is challenging. Existing physics-agnostic or data-driven methods often fail to recover fine-grained shadowing details under extreme sparsity. We propose a Physics-Regularized Low-Rank Tensor Completion (PR-LRTC) framework for radio map reconstruction. By modeling the signal field as a three-way tensor, we integrate environmental low-rankness with deterministic antenna physics. Specifically, we leverage Effective Aerial Degrees-of-Freedom (EADoF) theory to derive a differential gain topology map as a physical prior for regularization. The resulting optimization problem is solved via an efficient Alternating Direction Method of Multipliers (ADMM)-based algorithm. Simulations show that PR-LRTC achieves a 4 dB gain over baselines at a 10% sampling ratio. It effectively preserves sharp shadowing edges, providing a robust, physics-compliant solution for low-overhead beam management.

💡 Research Summary

The paper tackles the critical challenge of constructing high‑resolution radio maps for pixel‑based fluid antenna systems (FAS) while keeping channel state information (CSI) acquisition overhead minimal. Recognizing that conventional interpolation or purely data‑driven low‑rank tensor completion methods either oversmooth shadowing edges or require massive training data, the authors propose a physics‑aware reconstruction framework called PR‑LRTC (Physics‑Regularized Low‑Rank Tensor Completion).

The authors first model the spatial‑temporal signal field as a three‑way tensor X∈ℝ^{I×J×M}, where I and J are the horizontal and vertical grid indices of the service area and M denotes the number of antenna reconfiguration modes. Each entry X_{i,j,m} corresponds to the long‑term received signal strength (RSS) at location (i,j) when the base station operates in mode m. The antenna physics is captured through Effective Aerial Degrees‑of‑Freedom (EADoF) theory: the far‑field gain G_m(ϕ) of mode m is expressed as a truncated Fourier series of R orthogonal basis functions B_k(ϕ) weighted by mode‑specific complex coefficients w_{m,k}. The coefficients are designed to have a circular correlation structure, reflecting the smooth transition between adjacent modes.

A key insight is that the difference of RSS between any two modes at the same location eliminates the stochastic path‑loss and shadowing components, leaving only the deterministic gain difference G(r,m)−G(r,q). This leads to the definition of a differential gain topology map D∈ℝ^{I×J×M×M}, where D_{i,j}^{p,q}=G(r_{i,j},p)−G(r_{i,j},q). Because D depends solely on the antenna’s physical parameters and the geometric angle of departure, it is robust to environmental variations and provides a “physical skeleton” for reconstruction.

The reconstruction problem is formulated as a convex optimization that simultaneously enforces (1) data fidelity on the observed entries (Ω), (2) a low‑rank prior across the three tensor unfoldings via overlapped nuclear norms (∑{k=1}^{3}α_k‖X^{(k)}‖*), and (3) a physics‑aware regularizer R_{phys}(X,D)=∑{i,j}‖A x{i,j}−d_{i,j}‖2^2, where A is the cyclic difference operator (the incidence matrix of the mode‑cycle graph) and d{i,j} is the cycle‑consistent differential vector derived from D. The overall objective is

min_X ½‖P_Ω(X−Y)‖F^2 + λ_2 R{phys}(X,D) + λ_1∑{k=1}^{3}α_k‖X^{(k)}‖*.

To solve this problem efficiently, the authors employ the Alternating Direction Method of Multipliers (ADMM). They introduce auxiliary tensors M_k to decouple the nuclear‑norm terms, leading to sub‑problems that are either closed‑form (the X‑update, which reduces to solving independent quadratic problems per spatial location) or simple singular‑value thresholding (the M_k‑updates). Scaled dual variables U_k and a penalty parameter ρ are updated in the standard ADMM fashion, guaranteeing convergence under mild conditions. The X‑update naturally parallelizes over the I×J grid, making the algorithm suitable for real‑time or near‑real‑time deployment.

Simulation results are presented for a 2‑D urban scenario with I=64, J=64, and M=16. Sparse measurements are generated at sampling ratios of 5 % and 10 %. PR‑LRTC is benchmarked against conventional low‑rank tensor completion, a CNN‑based reconstruction, and simple linear interpolation. The proposed method consistently outperforms the baselines, achieving roughly a 4 dB improvement in mean‑square error at a 10 % sampling ratio and preserving sharp shadowing edges that other methods blur. Moreover, the ADMM implementation reduces computational load by about 30 % compared with a naïve proximal gradient approach, and the per‑iteration runtime is roughly halved thanks to the closed‑form X‑update.

In summary, the paper makes three major contributions: (1) it formalizes the deterministic differential gain between antenna modes as a physics‑based prior, (2) it integrates this prior with a low‑rank tensor model in a unified convex framework, and (3) it provides an efficient ADMM‑based solver that scales to realistic grid sizes. Limitations include the reliance on accurate EADoF‑based antenna modeling (which may be affected by manufacturing tolerances or temperature drift) and the computational cost of pre‑computing the differential map D for very large or three‑dimensional environments. Future work is suggested on adaptive learning of the physical prior from field measurements, extension to multi‑frequency and multi‑user settings, and development of lightweight online ADMM variants for dynamic environments.