Failure Detection for Pinching-Antenna Systems

A signal processing-based framework is proposed for detecting random segment failures in segmented waveguide-enabled pinching-antenna systems. To decouple the passively combined uplink signal and to provide per-segment observability, tagged pilots are employed. A simple tag is attached to each segment and is used to apply a known low-rate modulation at the segment feed, which assigns a unique signature to each segment. Based on the tagged-pilot model, a low-complexity per-segment maximum-likelihood (ML) detector is developed for the case in which the pilot length is no smaller than the number of segments. For the case in which the pilot length is smaller than the number of segments, sparsity in the failure-indicator vector is exploited and a compressive sensing-based detector is adopted. Numerical results show that the per-segment detector approaches joint ML performance, while the compressive sensing-based detector achieves reliable detection with a short pilot and can outperform baselines that require much longer pilots.

💡 Research Summary

The paper addresses the critical problem of detecting random segment failures in segmented waveguide‑enabled pinching‑antenna systems (SWAN), a promising architecture for next‑generation (6G) networks. Conventional SWAN designs aggregate the signals from all waveguide segments into a single RF chain, which makes the effective measurement matrix rank‑deficient and prevents per‑segment observability. To overcome this, the authors introduce a “tagged‑pilot” framework: each segment’s feed is modulated by a low‑rate binary sequence (the tag) that multiplies the transmitted pilot signal. This tag imprints a unique signature on the contribution of each segment, turning the aggregated observation into a linear model

  y = √P · B · diag(h) · s + n,

where B ∈ {±1}^{T×M} is the tag matrix, T is the pilot length, M is the number of segments, h contains the known channel gains, s ∈ {0,1}^M denotes the operational state of each segment, and n is AWGN.

Two regimes are considered.

1. Over‑determined regime (T ≥ M). In this case B can be made full‑column‑rank. By choosing orthogonal columns (e.g., selecting rows from a Walsh‑Hadamard matrix), B^HB = T·I_M, which yields a closed‑form least‑squares estimate

  â = (1/(T√P)) · B^H y.

The estimation error covariance becomes σ²/(PT)·I_M, i.e., errors are independent and identically distributed across segments. Consequently, a per‑segment maximum‑likelihood (ML) decision rule

  Re{h_m^* â_m} ≷ |h_m|²/2

achieves the same performance as a joint ML detector but with linear O(M) complexity, avoiding exponential tree search. The authors also discuss practical tag design when exact orthogonality is impossible, recommending random sub‑selection from a larger Hadamard matrix to keep inter‑column correlation low.

2. Under‑determined regime (T < M). Here B is rank‑deficient, so direct LS recovery of s is ill‑posed. The authors exploit the fact that failures are sparse: only a small fraction of segments change state between monitoring intervals. Defining a reference state s₀ (known from commissioning) and a sparse failure indicator f = s₀ − s, the residual after removing the known contribution becomes

  r = y + A_f s₀ = A_f f + n,

with A_f = −diag(x) B diag(h). By converting the complex model to a real‑valued one, they formulate an ℓ₁‑regularized least‑squares problem (LASSO):

  min_f ½‖r − A_f f‖₂² + λ‖f‖₁.

The regularization parameter λ is selected via a discrepancy principle that matches the residual energy to the expected noise power. After solving (using MATLAB’s built‑in LASSO solver), a simple threshold τ (typically 0.5) yields binary decisions on each segment’s failure status. This compressive‑sensing‑based detector works with very short pilots (e.g., T = 64 for M = 128) and still attains error probabilities below 10⁻³, outperforming baseline methods that require pilots an order of magnitude longer.

Simulation results use realistic parameters: carrier frequency 28 GHz, effective dielectric index n_eff = 1.4, segment length 1 m, noise variance –90 dB, and per‑segment failure probability 0.02. In the over‑determined case, the per‑segment ML detector’s error curves closely follow the joint ML bound across SNRs from –15 dB to –11 dB. In the under‑determined case, the LASSO detector’s performance improves as the pilot length grows but remains robust even when T ≪ M, confirming the sparsity assumption’s validity.

Overall, the paper makes three key contributions: (i) a novel tagging mechanism that restores full rank to the measurement matrix without requiring additional RF chains; (ii) a low‑complexity per‑segment ML detector that matches joint ML performance when enough pilots are available; (iii) a compressive‑sensing‑based detector that exploits sparsity to enable reliable failure detection with very short pilots. The proposed framework is hardware‑friendly (requiring only simple PIN‑diode switches or 0/π phase shifters) and scalable to massive segmented arrays, making it a practical solution for maintaining large‑scale pinching‑antenna infrastructures in future high‑frequency networks.