Reconfigurable Pinching-Antenna System (PASS)

• PASS is a flexible antenna system where movable pinching antennas on low-loss dielectric waveguides dynamically adjust radiation points to shape both large-scale and small-scale channel characteristics.
• It employs various architectures including segmented, center-fed, and multi-mode variants to address challenges in MIMO transmission, indoor positioning, and integrated sensing and communications.
• Performance analyses reveal optimal antenna spacing and count configurations that balance power extraction, path loss, and mutual coupling to enhance beamforming gain and system robustness.

Reconfigurable Pinching-Antenna System (PASS) is a flexible-antenna paradigm in which electromagnetic energy is conveyed by low-attenuation dielectric waveguides and radiated into free space by movable pinching antennas (PAs) attached at chosen locations along the waveguides. In the survey literature, PASS is presented as physically reconfigurable at the meter scale: radiating points can be attached, removed, and repositioned, so the system reshapes both large-scale and small-scale channel characteristics, establishes stable line-of-sight (LoS) links, mitigates blockage, and exploits very large apertures and near-field operation (Liu et al., 26 Jan 2026). Subsequent work extends this basic idea from geometry-only reconfiguration toward richer forms of control, including multi-waveguide MIMO transmission, integrated sensing and communications (ISAC), indoor positioning, AirComp, symbiotic radio, and phase-mismatch-based radiation-weight tuning (Bereyhi et al., 5 Mar 2025).

1. Architectural principle and physical realization

At the hardware level, PASS consists of one or more dielectric waveguides, feed points that inject signals into the waveguides, and one or more PAs attached to the waveguide surfaces. The waveguide serves as a low-loss transmission medium; a signal is launched at a feed point, propagates inside the waveguide with attenuation and phase rotation, and reaches a selected PA, which extracts part of the guided energy and radiates it into free space toward users. The control mechanism is mechanical or electromechanical in the sense that “pinching” means placing a secondary dielectric or coupled radiating structure into the evanescent-field region surrounding the waveguide. When the secondary element is removed, the waveguide returns to pure transmission-line operation; this reversibility is central to PASS (Liu et al., 26 Jan 2026).

The survey distinguishes three implementation abstractions. In the small dielectric scatterer model, a small dielectric object perturbs the evanescent field and acts as a localized scatterer. In the directional coupler waveguide model, the secondary element is an elongated strip or short auxiliary waveguide running parallel to the main waveguide; the survey separates this into leaky mode and feeder mode. In the multiport network abstraction, the PA is represented as a lumped three-port network with $S_{21}$ capturing the remaining guided signal, $S_{31}$ the radiation/coupling efficiency, and $S_{11}$ reflections caused by mismatch at the pinching point (Liu et al., 26 Jan 2026).

The distinctive architectural claim of PASS is that antenna geometry becomes a control variable. Conventional fixed arrays optimize digital or analog weights over immutable element positions. PASS, by contrast, optimizes where radiation occurs. This physical relocation is broader than wavelength-scale movable or fluid antenna adaptation, which the survey characterizes as mainly effective against small-scale fading; PASS is instead described as manipulating both small-scale and large-scale channel characteristics through long-range repositioning along the guiding structure (Liu et al., 26 Jan 2026).

2. Electromagnetic and signal models

The fundamental PASS signal model is a cascade of in-waveguide propagation and free-space propagation. For a narrowband single-waveguide, single-PA, single-user downlink, the survey writes the signal at the PA as

$$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$

where $\gamma_g=\alpha_g+j\beta_g$ is the complex waveguide propagation constant, with $\alpha_g>0$ the attenuation constant and $\beta_g>0$ the phase constant, and $\beta_g=\frac{2\pi}{\lambda_g}$ for guided wavelength $\lambda_g$. The corresponding end-to-end channel coefficient in the lossless-waveguide reference model is

$$h=\frac{\eta \exp\left(-j \tfrac{2\pi}{\lambda} R\right)}{R}\times \rho(\theta,\phi)\,\kappa \times \exp\left(-j \frac{2\pi}{\lambda_g} L\right),$$

where $S_{31}$0 is the PA-to-user LoS distance, $S_{31}$1, $S_{31}$2 is the PA radiation pattern, and $S_{31}$3 is the power-coupling factor (Liu et al., 26 Jan 2026).

With multiple PAs on a shared waveguide, the survey uses a cascaded coupling model in which earlier pinches deplete the guided signal available to later ones. If $S_{31}$4 is the local coupling factor of PA $S_{31}$5, then the effective cascaded coupling factor is

$S_{31}$6

and the user receives the coherent superposition

$S_{31}$7

This shared-waveguide power extraction is one reason PASS differs analytically from conventional arrays with independently driven elements (Liu et al., 26 Jan 2026).

The literature is not uniform on waveguide loss modeling. Several works neglect in-waveguide attenuation and retain only guided phase, especially when active antennas are expected to be relatively close. Other works explicitly incorporate attenuation. In the indoor positioning model, the waveguide propagation factor from PA position $S_{31}$8 back to the AP is

$S_{31}$9

and this term is essential because received signal strength is inverted into range estimates; the paper therefore argues that waveguide loss matters for RSSI-based ranging (Zhang et al., 11 Aug 2025).

MIMO-oriented PASS models make the same structural point in matrix form. In MIMO-PASS, the effective downlink channel $S_{11}$0 is itself a function of the location matrix $S_{11}$1, because each effective channel entry contains both distance-dependent attenuation and the phase accumulated inside the dielectric waveguide before radiation. The result is a hybrid architecture in which digital transceiver variables and physical pinching-element locations are jointly optimized (Bereyhi et al., 5 Mar 2025).

3. Reconfigurability mechanisms and architectural variants

The survey presents several major PASS variants developed to overcome limitations of the original end-fed single-waveguide formulation. Segmented PASS, described through segmented waveguide-enabled pinching-antenna system (SWAN), replaces one long waveguide with multiple short segments arranged end-to-end but not physically connected. Each segment has its own feed point. The stated motivations are to prevent uplink inter-antenna radiation within a segment, reduce in-waveguide propagation loss, and simplify maintenance. SWAN is further organized into segment selection, segment aggregation, and segment multiplexing protocols, which trade RF-chain count against performance (Liu et al., 26 Jan 2026).

Center-fed PASS (C-PASS) introduces bidirectional propagation on one waveguide. In the survey’s summary, conventional end-fed PASS has $S_{11}$2, whereas center-fed PASS has $S_{11}$3. The survey reports the asymptotic comparison

$S_{11}$4

with nonzero multiplexing gain for the center-fed case and zero multiplexing gain for the end-fed case. The associated control protocols are power splitting, direction switching, and time switching (Liu et al., 26 Jan 2026).

Multi-mode PASS (M-PASS) addresses the rank-one limitation of single-mode waveguides by allowing one waveguide to carry multiple guided modes. The survey distinguishes mode-selective and mode-combining structures. In the former, each PA is engineered to couple predominantly to a single mode, using a phase-matching condition of the form

$S_{11}$5

In the latter, each PA couples to multiple guided modes simultaneously, enabling electrically tunable modal mixing at the expense of higher hardware complexity (Liu et al., 26 Jan 2026).

Two later extensions enlarge the notion of PASS reconfigurability. First, two-dimensional PASS extends the conventional line-shaped structure into a continuous dielectric waveguide plane, thereby forming a reconfigurable radiating plane with PA coordinates

$S_{11}$6

subject to minimum spacing constraints in the plane. The stated objective is to expand spatial DoFs for indoor and multiuser settings where a one-dimensional waveguide becomes restrictive (Zhong et al., 12 Nov 2025). Second, amplitude-tunable PASS introduces direct, physics-grounded control of complex radiation weights through single-mode phase-mismatch radiation. In that model, the $S_{11}$7-th antenna’s radiated amplitude and phase are functions of $S_{11}$8, with

$S_{11}$9

and

$$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$0

so PASS becomes a weight-adaptive analog beamforming architecture rather than a purely geometry-controlled one (Altinoklu et al., 26 May 2026).

4. Beamforming and optimization methodologies

The optimization literature treats PASS as a tri-hybrid or hybrid architecture in which digital beamforming is coupled to electromagnetic-level or geometry-level control. The survey groups the main methods into analytical optimization, branch-and-bound, penalty-based alternating optimization, ADMM, penalty dual decomposition, majorization–minimization, successive convex approximation (SCA), fractional programming (FP), particle swarm optimization (PSO), and machine learning. Its general conclusion is that PASS optimization is harder than conventional beamforming because channel phase and path loss depend nonlinearly on physical PA positions, while multiple PAs on one waveguide are coupled through shared power extraction (Liu et al., 26 Jan 2026).

Representative communication designs follow this pattern. In MIMO-PASS, the downlink weighted sum-rate problem is solved by FP combined with Gauss–Seidel location updates; the digital precoder update has an RZF-like closed form, while each pinching-element position is updated by scalar grid search. The uplink counterpart exploits the MMSE detector for fixed locations and then iteratively refines the pinching-element positions via another Gauss–Seidel procedure (Bereyhi et al., 5 Mar 2025). In multi-user downlink PASS, a two-timescale design separates fast digital precoding from slow PA-position adaptation: primal-dual decomposition yields a short-term subproblem handled by a KKT-guided dual learning approach and a long-term subproblem handled by stochastic successive convex approximation (SSCA) (Zhang et al., 13 Apr 2025).

Several application-specific formulations use the same architectural logic. PASS-enabled multicast communication optimizes only the PA positions on a single waveguide to maximize the worst-user SNR, and solves the resulting non-convex max–min problem by PSO (Mu et al., 23 Feb 2025). PASS-assisted symbiotic radio formulates a joint transmit and pinching beamforming problem under power, spacing, deployment-region, and detection-error constraints, and proposes two solution strategies: a learning-aided gradient descent method based on reparameterized PA positions and projected beamforming variables, and a two-stage SCA-PSO approach (Wang et al., 9 Aug 2025). PASS-aided AirComp minimizes MSE jointly over PA positions, user powers, and the decoding vector using alternating optimization with Gauss–Seidel PA updates (Lyu et al., 12 May 2025).

Sensing-oriented designs show a similar decomposition. Wireless sensing via PASS employs a two-stage PSO-based algorithm: stage 1 optimizes PA positions under an unbiased waveform covariance $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$1, and stage 2 refines the waveform covariance by solving a convex semidefinite program with CVX (Wang et al., 21 May 2025). PASS-assisted ISAC uses alternating optimization in which the digital subproblem is solved by semidefinite relaxation and the pinching-beamforming subproblem is solved by SCA, a penalty method, and element-wise one-dimensional search over PA positions (Li et al., 27 Aug 2025). In the information-theoretic ISAC formulation, the beamformer is explicitly the vector of activated pinching locations $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$2, and the Pareto CR–SR frontier is obtained through rate-profile optimization, inner bounds from element-wise alternating optimization, and outer bounds based on Cauchy–Schwarz and Karamata’s inequalities (Ouyang et al., 15 May 2025).

5. Applications and reported performance characteristics

Performance analysis across the literature repeatedly shows that PASS does not obey naive “more antennas and denser packing are always better” rules. For array gain under equal-power radiation and fixed half-wavelength spacing, the array-gain paper proves

$$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$3

so an optimal finite number of pinching antennas exists. In the reported simulation with $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$4 GHz, $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$5 m, $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$6, and $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$7 m, the optimum is $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$8, corresponding to a waveguide length $$s_{\mathrm{pa}}(t)=\exp(-\gamma_g L)\, s_{\mathrm{in}}(t),$$9. For the two-antenna spacing problem with mutual coupling, the same paper reports an optimal spacing $\gamma_g=\alpha_g+j\beta_g$0 for $\gamma_g=\alpha_g+j\beta_g$1, again showing that denser placement is not always better (Ouyang et al., 10 Jan 2025).

In uplink and downlink MIMO transmission, MIMO-PASS reports that throughput is boosted significantly as compared with baseline MIMO architectures. The paper states gains of more than $\gamma_g=\alpha_g+j\beta_g$2 over fully digital massive MIMO with the same number of antennas as PASS, and over $\gamma_g=\alpha_g+j\beta_g$3 over small-scale fully digital conventional MIMO. For downlink versus number of pinching elements $\gamma_g=\alpha_g+j\beta_g$4, it further reports $\gamma_g=\alpha_g+j\beta_g$5 gain over massive MIMO and $\gamma_g=\alpha_g+j\beta_g$6 gain over classical hybrid MIMO. All proposed algorithms are reported to converge within about $\gamma_g=\alpha_g+j\beta_g$7 iterations under the tested settings (Bereyhi et al., 5 Mar 2025).

Indoor positioning reveals a different aspect of PASS. In the single-waveguide uplink model, RSSI-derived ranges are combined by a weighted least-squares estimator to obtain two-dimensional user coordinates. The paper’s main observations are that increasing the number of PAs improves positioning accuracy and robustness, that the benefit becomes marginal once the number of PAs exceeds about $\gamma_g=\alpha_g+j\beta_g$8, and that users located between and near PAs yield superior accuracy. Under $\gamma_g=\alpha_g+j\beta_g$9 Monte Carlo simulations at $\alpha_g>0$0 dBm, the reported mean error and variance are $\alpha_g>0$1 m and $\alpha_g>0$2 for $\alpha_g>0$3, and $\alpha_g>0$4 m and $\alpha_g>0$5 for $\alpha_g>0$6 (Zhang et al., 11 Aug 2025).

PASS has also been specialized to over-the-air computation, multicast, symbiotic radio, sensing, and ISAC. In PASS-aided AirComp, the proposed alternating optimization with Gauss–Seidel PA updates converges within about $\alpha_g>0$7 iterations and consistently outperforms fixed PASS, conventional MIMO, discrete PASS, and a projected-gradient baseline in MSE (Lyu et al., 12 May 2025). In PASS-enabled multicast, PSO-optimized PA placement significantly outperforms conventional multiple-antenna transmission, with particularly large advantage when the number of PAs is small (Mu et al., 23 Feb 2025). In wireless sensing with LCX reception, the paper reports that over $\alpha_g>0$8 random samples the proposed PASS architecture significantly improves the average PEB and robustness relative to conventional MIMO, and that optimized PA positions outperform fixed PA placements (Wang et al., 21 May 2025). In PASS-assisted ISAC, the CR–SR rate-region paper proves that the conventional fixed-antenna region is contained in the single-pinch PASS region, and the multiple-pinch analysis derives inner and outer bounds that closely approximate the true CR–SR region (Ouyang et al., 15 May 2025). The full-duplex ISAC beamforming paper further reports that PASS is less affected by stringent communication constraints than conventional MIMO-ISAC and benefits from increasing the number of waveguides and PAs per waveguide (Li et al., 27 Aug 2025).

6. Limitations, clarifications, and open problems

Despite the breadth of models and applications, the literature remains highly idealized. The survey identifies narrowband assumptions, dominant LoS propagation, physically position-dependent path loss and phase, and simplified coupling models as recurring premises. It also lists hardware mobility, wideband support, insertion loss and in-waveguide attenuation, reflection and mismatch, uplink inter-antenna radiation, calibration and position control, control signaling, scalability and maintenance, and channel estimation as open implementation challenges (Liu et al., 26 Jan 2026).

Several specific cautions recur across papers. First, reconfigurability is not always used in the same sense. Some formulations optimize PA positions directly; others fix a set of PA locations and only select which one is active. The indoor positioning paper repeatedly emphasizes that its formulation assumes fixed and known PA positions along a waveguide and activates only one PA in each time slot, so it does not develop a dynamic PA-placement optimization algorithm for localization (Zhang et al., 11 Aug 2025). Second, many analytical results assume negligible waveguide attenuation, equal-power radiation, or perfect CSI; these assumptions simplify tractability but limit direct transfer to wideband, lossy, or fast-varying deployments (Liu et al., 26 Jan 2026).

A second recurrent misconception is that PASS benefits should scale monotonically with antenna count or decreasing inter-element spacing. The array-gain analysis explicitly rejects both claims: there exists an optimal number of pinching antennas and an optimal inter-antenna spacing under the paper’s models, and overly many antennas or excessively small spacing can degrade performance through equal-power splitting, path loss, and mutual coupling (Ouyang et al., 10 Jan 2025).

The principal open problems identified across the literature concern reciprocity and uplink modeling, robust channel estimation under one-or-few RF chains per waveguide, dynamic PA control with realistic actuation latency, richer multipath and blockage-aware models, wideband and dispersive dielectric-waveguide effects, scalable multi-cell optimization, and hardware realization of advanced variants such as amplitude-tunable PASS and two-dimensional PASS. A plausible implication is that PASS will remain a joint electromagnetic, geometric, and signal-processing design problem rather than a direct extension of fixed-array beamforming. That implication is consistent with the survey’s broader view of PASS as an “outer layer” of beamforming, complementary to digital and analog processing but governed by distinct propagation and hardware constraints (Liu et al., 26 Jan 2026).