Pinching Antenna System (PASS): Dynamic Beamforming

• PASS is a reconfigurable wireless architecture that employs low-loss dielectric waveguides and attachable pinching antennas to deliver dynamically tailored beamforming over the last meter.
• Its flexible placement and user-centric design allow rapid configuration of service areas, optimizing performance for multicast, localization, and multiuser applications.
• Advanced electromagnetic models and optimization methods demonstrate PASS's capability to reduce transmit power and enhance near-field beamforming while addressing blockage and mobility challenges.

Pinching Antenna System (PASS) is a wireless architecture in which a signal is conveyed through a low-loss dielectric waveguide and radiated only at selected locations where small passive radiators, called pinching antennas (PAs), are attached. In this architecture, the radiating interface is physically moved close to users or targets, so the design emphasis shifts from overcoming long free-space propagation in the “last mile” to exploiting short, often line-of-sight, links in the “last meter.” PASS therefore differs fundamentally from conventional fixed-location antenna arrays: its beamforming dimension is not only electrical but also spatial, because the actual radiation points can be moved, activated, deactivated, added, or removed along extended waveguides (Liu et al., 30 Jan 2025).

1. Definition, emergence, and architectural identity

A PASS consists of a dielectric waveguide and multiple attachable pinching points formed by bringing a separate dielectric into contact or close proximity with the waveguide. The separate dielectrics are described as often mounted on plastic clips or clothespins; when clipped onto the waveguide, they become PAs. Their positions can be changed along the waveguide, and they can be added or removed with considerable flexibility. The waveguide itself can extend over walls, ceilings, rooftops, facades, or guardrails, so the physical aperture is not restricted to the compact footprint of a conventional base-station array (Liu et al., 30 Jan 2025).

This deployment model distinguishes PASS from several adjacent concepts. In conventional multi-antenna systems, antennas are fixed, radiation occurs directly from active electronic antennas, and post-deployment reconfiguration is costly. In distributed antenna systems and fluid or movable antenna concepts, reconfiguration typically occurs over limited apertures of a few wavelengths, whereas PASS can extend the waveguide arbitrarily and place radiators physically close to users over meter- or tens-of-meters scales. In the modeling literature, PASS is also contrasted with RIS: RIS can create alternative reflected paths, but suffers from the double-fading effect, whereas PASS carries the signal through the guide and radiates only on the final short hop (Wang et al., 9 Feb 2025).

The practical feasibility of the concept is supported by a prototype developed by NTT DOCOMO in 2021, cited as the first PASS prototype. That prototype demonstrated that clipping onto the waveguide at arbitrary points creates communication spots on demand, and removing the clip stops radiation there immediately. The same demonstration is used to motivate low implementation cost, scalable deployment, rapid creation and deletion of service areas, and the possibility of establishing near-line-of-sight or line-of-sight local links with minimal wireless exposure beyond the target region (Liu et al., 30 Jan 2025).

2. Electromagnetic operation and signal models

The physical principle of PASS is electromagnetic coupling between the main dielectric waveguide and the separate dielectric used as the PA. In the overview formulation, the PA is modeled as an open-ended directional coupler, and coupled-mode theory gives the normalized power exchange

$$P_{\mathrm{guide}} = 1 - F \sin^2(\kappa L), \qquad P_{\mathrm{pinch}} = F \sin^2(\kappa L),$$

where $L$ is the coupling length, $\kappa$ is the coupling coefficient, and $F \le 1$ is the maximum coupling efficiency. In the matched-index case, $F=1$, and full power transfer from a single PA is possible when $L=\pi/(2\kappa)$ (Liu et al., 30 Jan 2025).

At the communication level, the end-to-end PASS channel is cascaded: guided propagation inside the waveguide followed by free-space propagation from each PA to the user. For the $n$-th PA, a line-of-sight contribution is written as

$$y_n = \frac{\beta_n \sqrt{P_n}}{r_n} e^{-j \frac{2 \pi}{\lambda} \left( r_n + n_{\mathrm{eff}} d_n \right)} x,$$

where $x$ is the signal injected into the waveguide, $\beta_n$ is the free-space channel coefficient at 1 m, $L$0 is the radiated power of the $L$1-th PA, $L$2 is the PA-to-user free-space distance, $L$3 is the guided distance to that PA, and $L$4 is the effective refractive index. The effective channel is the coherent sum of these terms,

$L$5

so moving a PA changes both amplitude and phase through $L$6 and $L$7 simultaneously (Liu et al., 30 Jan 2025).

A distinctive feature of PASS is that the radiated powers are coupled. In the equal power model, coupling lengths are designed so that each PA radiates the same amount of power. In the proportional power model, all PAs have the same coupling length and thus radiate the same fraction of the remaining guided power, producing decreasing emitted powers along the guide. A more detailed adjustable radiation model expresses the power radiation ratio on waveguide $L$8 as

$L$9

with coupling coefficient

$\kappa$0

where $\kappa$1 is the spacing between waveguide and PA. On this basis, a closed-form spacing arrangement was derived to realize equal-power radiation for arbitrary activated sets (Wang et al., 9 Feb 2025, Xu et al., 30 Apr 2025).

Because PASS uses long apertures and can place radiators close to intended points, the relevant propagation regime is often near-field rather than only far-field plane-wave steering. The overview literature therefore emphasizes spatial focusing, not merely directional beam steering, as a native capability of PASS (Liu et al., 30 Jan 2025).

3. Pinching beamforming and transmission architectures

The central systems concept introduced with PASS is pinching beamforming: beamforming by dynamically adjusting where along a waveguide passive radiators are located. In this formulation, positioning serves two simultaneous purposes. First, it reduces path loss by placing PAs near users. Second, it reshapes phases through both guided and free-space propagation, enabling constructive combining at intended users and less constructive combining elsewhere. The literature explicitly states that this is not just antenna selection and not merely analog phase control (Liu et al., 30 Jan 2025).

PASS operation has been framed in both BS-centric and user-centric forms. In BS-centric PASS, the base station determines PA locations to optimize metrics such as spectral efficiency or fairness; positions can be continuously varied or selected from a discrete candidate set. In user-centric PASS, users themselves pinch or release portable PAs onto the waveguide, lowering infrastructure complexity but making each PA effectively more user-dedicated and shifting adaptation burdens to baseband processing (Liu et al., 30 Jan 2025).

At the architectural level, two broad transmission classes have been proposed. The non-multiplexing architecture feeds each waveguide with a single data stream and relies primarily on pinching beamforming; it uses simple baseband processing and is suited to geographically isolated service regions. The multiplexing architecture is the PASS counterpart of MIMO transmission and combines joint transmit beamforming with pinching beamforming. Within multiplexing PASS, the literature distinguishes a sub-connected scheme, in which each RF chain feeds one waveguide; a fully-connected scheme, in which each RF chain feeds all waveguides through power splitters; and a phase-shifter-based fully-connected scheme, which adds phase shifters and is described as tri-hybrid beamforming because it combines digital beamforming, analog beamforming via phase shifters, and pinching beamforming via PA placement (Liu et al., 30 Jan 2025).

This architectural taxonomy was extended in multiuser MIMO-PASS, where an access point uses $\kappa$2 dielectric waveguides with $\kappa$3 movable pinching elements per waveguide for both downlink and uplink multiuser MIMO. In that framework, the effective channel matrix $\kappa$4 is explicitly a function of element locations, and hybrid design proceeds by jointly optimizing digital precoding or combining together with the element positions. Downlink weighted sum-rate maximization was treated by fractional programming and Gauss-Seidel updates, while the uplink was handled by location optimization followed by the closed-form MMSE combiner (Bereyhi et al., 5 Mar 2025).

4. Performance characterization and optimization methodologies

Several numerical and analytical studies show that PASS performance is strongly geometry-dependent and is not monotonic in the most obvious design variables. In a single-user array-gain analysis with fixed half-wavelength spacing, PASS was shown to have an optimal finite number of activated PAs rather than benefitting indefinitely from larger $\kappa$5. Under the setup $\kappa$6 GHz, $\kappa$7 m, and $\kappa$8, the reported optimum was $\kappa$9, corresponding to a waveguide length of approximately $F \le 1$0 m, and when mutual coupling was included in the two-PA case with $F \le 1$1, the optimal spacing was reported as $F \le 1$2 (Ouyang et al., 10 Jan 2025).

In the overview paper’s multiuser transmit-power study, a 15 GHz system with five parallel 50-meter waveguides spaced 5 meters apart, four users located 15–20 meters from the BS, six PAs per waveguide, a sub-connected multiplexing architecture, and 200 candidate PA positions required significantly less transmit power than both a fully digital MIMO baseline with 5 antennas and a massive MIMO baseline with 30 antennas and 5 RF chains. In that setting, the proportional power model performed nearly the same as the equal power model (Liu et al., 30 Jan 2025).

A more detailed beamforming-optimization study then reported that PASS reduces transmit power by over 95% compared to conventional and massive MIMO, that a ZF-based low-complexity algorithm performs similarly to a penalty-based alternating optimization algorithm, 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). A related discrete-activation framework introduced globally optimal branch-and-bound designs for single-user and multi-user settings together with a many-to-many matching algorithm; in the reported multi-user comparisons, PASS reduced power by over 22 dBm compared to conventional MIMO and about 7.5 dBm compared to massive MIMO, while the matching method incurred only slight performance loss with much lower computational overhead (Xu et al., 30 Apr 2025).

Optimization methods developed for PASS reflect the multimodal coupling between spatial placement and signal processing. Reported tools include penalty dual decomposition for joint transmit and pinching beamforming, PSO for multicast position design and sensing CRB minimization, penalty-based alternating optimization, ZF-based low-complexity beamforming, branch-and-bound for discrete activation, many-to-many matching with KKT-based beamforming, fractional programming with Gauss-Seidel updates for MIMO-PASS, MM with candidate-search and bisection-search coordinate solvers for blockage-aware multicasting, and AO combined with SDR, SOCP, SCA, and penalty methods for ISAC (Liu et al., 30 Jan 2025, Mu et al., 23 Feb 2025, Hanif et al., 7 Feb 2026).

Blockage-aware multicast results are especially important because they address a common objection that PASS requires universal line of sight. In a probabilistic blockage model with $F \le 1$3, PASS still outperformed conventional antenna systems. The MM procedure converged in about 3 to 4 iterations in the reported experiments, and when the system had 8 PAs and 25 users, the execution time with the candidate search method was approximately 2.5 times that with the bisection search method (Hanif et al., 7 Feb 2026).

5. Communication, positioning, sensing, and mobility-oriented applications

Multicast is a natural PASS use case because all PAs on a waveguide can carry the same signal set. In one single-waveguide multicast formulation, PA positions were optimized by PSO and PASS significantly outperformed conventional multiple-antenna transmission, especially when the number of PAs was small (Mu et al., 23 Feb 2025). A later multicast study showed that for a single PA and linearly distributed users, the optimal activated location is the midpoint between the leftmost and rightmost users,

$F \le 1$4

and proved that the corresponding multicast rate is larger than that of a conventional fixed-location antenna system; for multiple PAs and multiple waveguides, the same work developed element-wise AO for pinching beamforming and an MM-plus-SOCP framework for joint transmit and pinching beamforming (Shan et al., 31 May 2025).

Uplink PASS has also been studied in dedicated indoor line-of-sight settings. Three scenarios were analyzed: 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 PA placement law

$F \le 1$5

implies an asymmetric non-uniform PA distribution in the near zone, while the far-zone limit approaches approximately wavelength-spaced placement. The paper reports that optimizing PA positions significantly enhances ergodic sum rate and that PASS significantly outperforms conventional Multiple-input Single-output networks (Hou et al., 17 Feb 2025). Multiuser uplink and downlink were further generalized in MIMO-PASS, where throughput in both directions was reported to increase substantially relative to conventional MIMO, hybrid analog-digital designs, and massive MIMO, although the uplink performance drops rapidly when $F \le 1$6, that is, when the number of users exceeds the number of waveguides or RF chains (Bereyhi et al., 5 Mar 2025).

PASS has also been proposed as a localization infrastructure. The first indoor positioning framework based on PASS used uplink RSSI-based ranging and weighted least squares on a single ceiling-edge waveguide. In that model, one PA is activated per time slot, waveguide attenuation is explicitly included in the ranging equation, and positioning is performed in 2D. The reported observations were that more PAs improve positioning accuracy and robustness, that major improvement occurs from $F \le 1$7 to $F \le 1$8 while performance gain becomes marginal when the number of PAs exceeds about 7, and that user locations between and near PAs yield superior positioning accuracy (Zhang et al., 11 Aug 2025).

Sensing and ISAC have become one of the most active PASS extensions. The overview paper introduced PASS for near-field sensing and integrated sensing and communications, using the terms NISE and ISAC, and described both ISAC-PASS and NISE-PASS architectures (Liu et al., 30 Jan 2025). A dedicated sensing paper then proposed PASS for transmit probing and LCX cables for receive echo collection, derived a multi-target CRB, and minimized it through joint waveform and PA-position optimization; the reported results showed significant gains in sensing accuracy and robustness over conventional sensing systems (Wang et al., 21 May 2025). A separated two-waveguide PASS-ISAC design used one waveguide for information-bearing transmission and the other for receiving echoes, with target illumination power as the sensing objective under communication QoS constraints (Zhang et al., 10 Apr 2025). A later multiple-waveguide PASS-assisted ISAC framework used transmitting PAs on waveguides and a receiving ULA at a full-duplex BS, formulated target sensing CRB minimization under QoS and power constraints, and reported that PASS was less affected by stringent communication constraints than conventional MIMO-ISAC while benefitting further from more waveguides and more PAs per waveguide (Li et al., 27 Aug 2025).

A mobility-oriented extension treated PASS-enabled UAV delivery. In that model, the BS uses a single dielectric waveguide with $F \le 1$9 predefined PAs, the outer layer optimizes the UAV delivery sequence through a hierarchical alternating optimization scheme combining a genetic algorithm and dynamic programming, and the inner layer optimizes the PA activation vector using either branch-and-bound or incremental search and local refinement. The reported conclusion was that PASS outperforms conventional multi-antenna systems, especially with higher communication rate requirements (Lv et al., 30 Sep 2025).

6. Limitations, recurring misconceptions, and open research directions

A recurring misconception in the PASS literature is to view it as ordinary beamforming with unusual hardware. The foundational papers explicitly reject that description: pinching beamforming is not just antenna selection, not merely analog phase control, and not reducible to standard fixed-array reconfigurable transceiver design, because moving a PA changes both the free-space path length and the guided-wave path length, thereby altering amplitude, phase, and often the propagation regime itself (Liu et al., 30 Jan 2025).

The same literature also makes clear that PASS does not make all standard tradeoffs disappear. More PAs are not always better; both array-gain analysis and numerical case studies report optimal finite PA counts or threshold behavior beyond which gains saturate or reverse (Ouyang et al., 10 Jan 2025, Zhang et al., 11 Aug 2025). Radiated powers are passively coupled, not independently programmable as in conventional arrays, so early PAs affect the residual power available to later PAs. Coupling lengths are not as easily tunable as electronic weights, and real-time movement or activation of PAs requires advanced hardware (Liu et al., 30 Jan 2025).

Many reported gains are established under deliberately structured assumptions: LoS or LoS-dominated propagation, narrowband models, point-target sensing, or idealized in-waveguide attenuation. The indoor positioning framework assumes LoS, single-waveguide single-edge deployment, 2D localization, one PA active per slot, and reports simulation-based rather than experimental validation (Zhang et al., 11 Aug 2025). The LCX-based sensing architecture is narrowband, restricts targets to the ground plane, and evaluates performance through CRB rather than a full estimator implementation (Wang et al., 21 May 2025). The separated PASS-ISAC model assumes a static or low-speed point target, neglects Doppler, and optimizes illumination power rather than detection probability or estimation error directly (Zhang et al., 10 Apr 2025). The multiple-waveguide PASS-assisted ISAC paper neglects self-interference after qualitative justification and assumes equal power splitting among PAs on each waveguide (Li et al., 27 Aug 2025).

Open research directions identified in the overview include stochastic-geometry models for networks with line-of-sight-dominated, near-field, placement-flexible links; machine-learning methods for channel estimation and user-centric pinching over large, spatially non-stationary apertures; scalable optimization under coupled power leakage; mobility tracking; protocol design for BS-centric and user-centric control; and advanced hardware design for real-time movement or activation, adjustable leakage, loss mitigation, and suitable dielectric materials (Liu et al., 30 Jan 2025). A plausible implication is that PASS will remain a multidisciplinary topic, requiring joint progress in electromagnetics, hardware engineering, signal processing, and network optimization before its full architectural potential can be evaluated under deployment-grade conditions.