Reconfigurable Pinching Antenna

• Reconfigurable Pinching Antennas are flexible antenna systems formed on dielectric waveguides, where radiating points can be dynamically controlled through mechanical or electronic actuation.
• They employ diverse reconfiguration mechanisms—such as continuous spatial placement, discrete activation, and electronic leakage control—to optimize line-of-sight availability, phase coherence, and amplitude tuning.
• Advanced optimization and modeling frameworks, including coupled-mode theory and robust design methods, enable significant enhancements in communication, energy efficiency, and sensing while mitigating practical challenges like waveguide attenuation.

A reconfigurable pinching antenna (PA) is a radiating element formed on a dielectric waveguide whose effective radiation behavior is controlled by changing where, when, or how the waveguide leaks energy into free space. In the pinching-antenna-system (PASS) literature, this reconfigurability is used to alter large-scale geometry, line-of-sight (LoS) availability, phase accumulation, and sometimes radiation amplitude and phase directly, rather than relying only on conventional array processing. PASS has accordingly been positioned as a channel-altering architecture in which base-station RF chains feed long dielectric waveguides and radiating points are “pinched” at custom positions of interest (Illi et al., 6 Apr 2026, Ding et al., 2024).

1. Definition, emergence, and conceptual position

A PA is described in closely related but consistent ways across the literature: as a flexible antenna composed of a waveguide and multiple dielectric particles, as a small dielectric particle or “plastic pinch” applied to a dielectric waveguide, and as a controllable radiation point created on a low-loss guiding structure (Hou et al., 17 Feb 2025, Ding et al., 2024). A PA system therefore differs from a rigid array in that the radiating location is not fixed by array manufacture; the radiating point is part of the deployment and control space.

The indoor immersive-communications literature states that PAs were first demonstrated by NTT DOCOMO in 2021 and emphasizes their suitability for metaverse, holographic displays, virtual tourism, gaming, and intelligent manufacturing (Wang et al., 9 Jun 2025). The survey literature places PASS alongside reconfigurable intelligent surfaces and movable antennas as a wireless channel-altering scheme, but distinguishes it by extending the reach of a base station through long waveguides on which one or many radiating antennas are pinched at chosen positions (Illi et al., 6 Apr 2026).

Relative to adjacent reconfigurable-antenna paradigms, the pinching mechanism occupies a specific niche. The hybrid pinching-fluid literature states that a PA is strong when path loss dominates, because it changes the transmit-side geometry and directly strengthens the channel, whereas a fluid antenna is strong when fading dominates, because it selects among receive ports to exploit spatial diversity (Lin et al., 7 May 2026). The PASS survey similarly argues that conventional MIMO and even massive MIMO improve array gain and multiplexing but do not readily convert a blocked non-LoS geometry into a strong direct link, whereas PASS can do so by placing the effective radiator close to the intended service region (Illi et al., 6 Apr 2026).

2. Physical architecture and propagation model

At the physical level, PASS transmissions are usually modeled as a two-stage process: in-waveguide propagation from the feed point to the PA, followed by free-space propagation from the PA to the user or target (Illi et al., 6 Apr 2026). In the survey formulation, the in-waveguide signal is written as

$s_L(t)=\sqrt{P}\exp(-\delta L)s_0(t),$

with $\delta=\alpha+j\beta$, where $\alpha$ captures amplitude attenuation and $\beta=2\pi n_W/\lambda$ captures the waveguide phase shift (Illi et al., 6 Apr 2026).

In the uplink PASS model for indoor LoS reconfiguration, the large-scale channel gain between PA $i$ and user $i$ is

$h(l_i,u_i)=\frac{c}{4\pi f_c |U_i-L_i|},$

with distance

$$|U_i-L_i| = \sqrt{(x_{li}-x_{ui})^2 + y_{ui}^2 + h^2},$$

and the small-scale term is represented only as a deterministic phase shift,

$$\exp\!\left(-j\frac{2\pi (|U_i-L_i|+x_i)}{\lambda}\right),$$

so that reconfiguration acts primarily through geometry and phase coherence rather than stochastic fading (Hou et al., 17 Feb 2025).

A more hardware-explicit treatment models the PA as an open-ended directional coupler and uses coupled-mode theory for the interaction between the main waveguide mode and the pinching-antenna mode: $$\frac{dA(x)}{dx} = -j\kappa B(x)e^{-j\Delta\beta x}, \qquad \frac{dB(x)}{dx} = -j\kappa A(x)e^{j\Delta\beta x}.$$ This yields the power exchange laws

$\delta=\alpha+j\beta$0

and motivates both equal-power and proportional-power models for multiple PAs on the same waveguide (Wang et al., 9 Feb 2025).

The multiport-network treatment adds a physically consistent circuit-level constraint. In that model, a PA is a three-port scattering device, the waveguide segment is a two-port network, and ideal PA reconfigurability is bounded by passivity,

$\delta=\alpha+j\beta$1

which is the fundamental constraint allowing full amplitude and phase control in the ideal case (Wang et al., 6 Sep 2025).

3. Modes of reconfigurability

The literature uses “reconfigurable PA” to denote several distinct control mechanisms, all implemented on or around a dielectric waveguide.

Reconfigurability mode	Mechanism	Representative papers
Continuous spatial placement	Move the PA along the waveguide to reduce distance or align phase	(Jiang et al., 3 Sep 2025, Hou et al., 17 Feb 2025)
Discrete activation	Select one fixed PA location from a finite candidate set	(Tyrovolas et al., 3 Nov 2025, Tyrovolas et al., 21 Dec 2025)
Electronic leakage control	Varactor-loaded modules switch leakage OFF/ON or tune it continuously	(Plata-Orozco et al., 25 Jun 2026)
Amplitude/phase control	Scattering coefficients or directional coupler parameters tune radiation weights	(Wang et al., 6 Sep 2025)
Feed-point selection	Choose which end of a dual-fed waveguide injects the signal	(Xie et al., 5 Mar 2026)

In the mechanically reconfigurable formulation used for SWIPT, the PA position $\delta=\alpha+j\beta$2 on a waveguide is a direct design variable, and optimal placement is the point on the feasible interval closest to the user, subject to the energy-harvesting constraint and probabilistic LoS blockage (Jiang et al., 3 Sep 2025). In the uplink PASS formulation, multiple PAs can be positioned so that guided-wave and free-space phases differ by integer multiples of $\delta=\alpha+j\beta$3, producing constructive addition at the access point (Hou et al., 17 Feb 2025).

A distinct practical line of work removes continuous motion and studies two-state PASs, where the PA locations are fixed and only the activation states are controlled. In that model, the waveguide contains a finite set of radiating points with spacing

$\delta=\alpha+j\beta$4

and one location is activated per channel use; the resulting pinching discretization efficiency quantifies the gap to the continuous benchmark (Tyrovolas et al., 3 Nov 2025).

Electronic reconfiguration is exemplified by the E-pinching antenna, which replaces mechanical pinching with varactor-loaded modular perturbations on a rectangular dielectric waveguide. There, capacitance tuning controls whether a local module is near-OFF or ON, so the radiation point is electrically rather than mechanically realized (Plata-Orozco et al., 25 Jun 2026). The multiport and directional-coupler literature further generalizes reconfigurability to amplitude-only control and amplitude-constrained phase control, with practical directional couplers characterized by a coupling coefficient $\delta=\alpha+j\beta$5 and an intrinsic phase range $\delta=\alpha+j\beta$6 (Wang et al., 6 Sep 2025).

This suggests that “reconfigurable PA” is not a single actuation paradigm but a layered family of mechanisms: meter-scale spatial relocation, binary spatial selection, electronic leakage modulation, amplitude/phase tuning, and feed-path selection are all treated as legitimate forms of PA reconfiguration in the current literature.

4. Modeling and optimization frameworks

Because PA position enters both path-loss and phase terms, PASS design problems are typically highly coupled and nonconvex. Consequently, the literature is dominated by alternating, block-coordinate, swarm-based, and robust-convexification methods rather than by a single universal synthesis procedure.

Under user-location uncertainty, the robust PA literature formulates joint power allocation and antenna placement with bounded error disks. For the single-antenna case, a worst-case robust design is transformed via the S-procedure into a convex semidefinite program with an affine LMI, yielding a globally solvable formulation. For the multi-antenna case, the worst-case gain

$\delta=\alpha+j\beta$7

is evaluated numerically, the optimal power has the closed form

$\delta=\alpha+j\beta$8

and antenna placement is optimized by multi-start block coordinate descent (Feng et al., 10 Apr 2026).

For multi-user interference exploitation, symbol-level precoding (SLP) is combined with PAS and the joint beamforming/placement problem is decoupled by alternating optimization. The PA-position subproblem is smoothed through a log-sum-exp approximation and solved by projected gradient descent over feasible movable regions that enforce minimum inter-PA spacing (Pang et al., 14 Mar 2026). In PASS-aided over-the-air computation, the objective becomes mean-squared-error minimization over PA positions, user powers, and decoding vectors; the solution combines a closed-form receive combiner, a KKT-based power update, and Gauss-Seidel one-dimensional PA searches on feasible grids (Lyu et al., 12 May 2025).

Two spatial generalizations further expand the optimization landscape. The sensing architecture with PASS transmission and LCX reception minimizes the trace of the target-location Cramér–Rao bound through a two-stage procedure: particle swarm optimization for PA geometry and a convex SDP in CVX for waveform covariance (Wang et al., 21 May 2025). The 2D-PASS architecture extends the feasible region from a line to a waveguide plane,

$\delta=\alpha+j\beta$9

and studies both a continuous max–min SNR design via PSO and a discrete, hardware-aware variant via MILP with McCormick linearization (Zhong et al., 12 Nov 2025).

Large-system formulations preserve the same pattern. Multi-cell weighted-sum-rate maximization in multi-waveguide PASS combines fractional programming, block coordinate descent, and PSO for joint precoding, waveguide power allocation, and antenna placement (Chen et al., 11 Jun 2026). Blockage-aware PASS uses the Hungarian algorithm and surrogate-assisted block-coordinate search in a discrete one-user-per-PA case, and WMMSE-DDPG in the general continuous-placement case where non-smooth LoS transitions make conventional smooth optimization unreliable (Xie et al., 3 Jan 2026). Covert PASS likewise separates a closed-form single-PA placement rule from TwinPSO for multi-waveguide multi-PA pinching beamforming (Jiang et al., 14 Apr 2025).

5. Communication, computation, energy, and sensing performance

The most explicit analytical rate characterization for reconfigurable uplink PASS appears in the three-scenario formulation of multiple PAs for a single user (MPSU), a single PA for a single user (SPSU), and a single PA for multiple users (SPMU). In MPSU, the optimized locations are

$\alpha$0

with $\alpha$1. The paper proves that the near-zone optimized positions are asymmetric and non-uniform, while in the far zone they simplify to the symmetric uniform spacing

$\alpha$2

It further states that PASS significantly outperforms conventional MISO networks, that the ergodic sum rate of MPSU is strictly greater than SPSU, and that optimizing PA positions significantly enhances the ergodic sum rate (Hou et al., 17 Feb 2025).

Discrete reconfigurability does not eliminate most of the gain. The two-state PAS literature defines pinching discretization efficiency as the ratio of discrete-system ergodic rate to the continuous benchmark and reports that, for $\alpha$3 m, the PDE exceeds 95% with only two PAs; for larger spaces, about 3 PAs are needed for $\alpha$4 m and about 4 PAs for $\alpha$5 m to achieve similarly high efficiency (Tyrovolas et al., 21 Dec 2025). The closely related ergodic-rate analysis likewise reports that over 95% of continuous-PAS performance is achieved with only two antennas in the compact geometry $\alpha$6 m (Tyrovolas et al., 3 Nov 2025).

Power-efficiency results are similarly strong but conditional on geometry optimization. The hardware-modeling and beamforming paper reports that PASS reduces transmit power by over 95% compared to conventional and massive MIMO, that discrete activation causes minimal performance loss but requires a dense antenna set to match continuous activation, and that the proportional-power model yields performance comparable to the equal-power model (Wang et al., 9 Feb 2025). Under user-location uncertainty, a multi-antenna robust benchmark shows that at $\alpha$7, the fixed-antenna deployment needs more than $\alpha$8 mW, while the optimized PA scheme needs only $\alpha$9 mW (Feng et al., 10 Apr 2026).

The current literature also corrects two common oversimplifications. First, adding more PAs is not by itself sufficient: the SLP-PAS study reports that random or fixed placements gain little from increasing PA count, indicating that placement optimization is crucial, and the robust-design study notes that with fixed positions, adding more antennas can actually increase worst-case power because of stronger unfavorable superposition (Pang et al., 14 Mar 2026, Feng et al., 10 Apr 2026). Second, the relative value of phase control depends on the geometry. The multiport-network study finds that, in single-user scenarios with optimized PA positions, performance gains arise primarily from amplitude reconfigurability and DC-based PAs approach ideal performance; with fixed PA positions, both amplitude and phase reconfigurability are critical and DC-based PAs incur non-negligible loss (Wang et al., 6 Sep 2025).

Beyond rate and power, reconfigurable PAs have been specialized to several noncanonical tasks. PASS-aided AirComp improves aggregation accuracy through joint PA placement, transmit-power design, and receive combining (Lyu et al., 12 May 2025). PASS-assisted SWIPT under probabilistic LoS blockage yields a closed-form joint design of PA position and power-splitting ratio, with the optimal PA deployed as close to the user as possible within the energy-feasible interval (Jiang et al., 3 Sep 2025). PASS-enhanced covert communications uses PA placement to strengthen Bob’s channel while respecting Willie’s covertness constraint; in the single-waveguide single-PA case, the optimal PA is the feasible point closest to Bob’s $\beta=2\pi n_W/\lambda$0-coordinate (Jiang et al., 14 Apr 2025).

6. Practical constraints, blockage mitigation, and implementation outlook

The reported gains are obtained under assumptions that are explicit in the literature. The uplink PASS study assumes predominantly LoS propagation, an ideal waveguide with negligible path loss, uniformly distributed users on the floor, TDMA uplink access, and far-zone or $\beta=2\pi n_W/\lambda$1 approximations for several closed forms (Hou et al., 17 Feb 2025). The survey identifies additional practical bottlenecks: in-waveguide attenuation, mechanical movement cost, antenna coupling, CSI acquisition, and integration with current standards (Illi et al., 6 Apr 2026).

In-waveguide attenuation is especially consequential. The dual-fed PASS work models power decay as

$\beta=2\pi n_W/\lambda$2

and gives the dielectric-waveguide approximation

$\beta=2\pi n_W/\lambda$3

For PTFE at 28 GHz it reports $\beta=2\pi n_W/\lambda$4 dB/m and notes that power can drop from 30 dBm to 15 dBm after 10 m. Its remedy is dual-feed selection, in which the effective in-waveguide distance is

$\beta=2\pi n_W/\lambda$5

and the resulting high-SNR ergodic-rate gain over the single-fed architecture is

$\beta=2\pi n_W/\lambda$6

which halves the average attenuation penalty (Xie et al., 5 Mar 2026).

Blockage is not treated only as a liability. The obstacle-aware PASS model develops an exact cylinder-obstacle LoS test and reports that PAs can effectively leverage obstacles to suppress co-channel interference, converting potential blockages into performance gains (Xie et al., 3 Jan 2026). A complementary hardware demonstration of electronically reconfigurable mmWave pinching antennas shows that a low-loss rectangular dielectric waveguide with varactor-loaded radiation modules can serve both LoS and NLoS users separated by a metallic partition. In that letter, the waveguide loss is about 3.2 dB/m versus about 68 dB/m free-space transmission loss, a single module reaches 87.1% radiation efficiency at maximum capacitance, and the two-module blocked scenario retains $\beta=2\pi n_W/\lambda$7 dB, far better than the 57.545 dB free-space path loss over the same span (Plata-Orozco et al., 25 Jun 2026).

The sensing literature broadens the implementation picture further by pairing waveguide-fed PA transmission with LCX-based echo collection. There, PASS improves the average position error bound and robustness relative to conventional sensing baselines by combining reconfigurable transmit illumination with wide-area receive collection (Wang et al., 21 May 2025). The hybrid pinching-fluid literature indicates another architectural direction, namely combining transmitter-side LoS enhancement from PAs with receiver-side diversity from fluid antennas to cover both path-loss-limited and fading-limited regimes (Lin et al., 7 May 2026).

The survey’s forward directions are correspondingly hardware- and control-oriented rather than purely analytical: hybrid PASS-RIS architectures, extension to higher frequencies and optics, low-complexity real-time control, robust CSI acquisition under channel aging and rank-deficient estimation, and hybrid deployments where conventional arrays and PASS coexist (Illi et al., 6 Apr 2026). Taken together, the present literature portrays the reconfigurable PA not as a single device but as a controllable waveguide-radiation interface whose most important research questions now concern physical consistency, actuation granularity, attenuation management, scalable control, and deployment realism.