PaperlayerView exact revision on arXiv

Source-linked AI summary

MIMO-PASS: Uplink and Downlink Transmission via MIMO Pinching-Antenna Systems

Ali Bereyhi, Chongjun Ouyang, Saba Asaad, Zhiguo Ding, H. Vincent Poor

arXiv:2503.03117v1cs.ITeess.SP

TL;DR

The paper studies joint digital and analog design for uplink detection and downlink beamforming in multiuser MIMO-PASS systems. It develops iterative optimization algorithms and finds that PASS improves weighted sum-rate over fixed-location MIMO, massive MIMO, and hybrid analog-digital baselines.

  • Problem

    Uplink detection requires jointly designing the digital receiver matrix and analog receive beam, while the uplink weighted sum-rate depends on both components.

  • Method

    The paper jointly optimizes digital precoding or receiver matrices with pinching-element locations using fractional programming, block coordinate descent, Gauss-Seidel updates, and iterative zero-forcing or MMSE-based algorithms.

  • Results

    PASS achieves significantly higher weighted sum-rates than fixed-location baselines, exceeding fully digital massive MIMO by more than 30% and small-scale fully digital MIMO by over 200%.

  • Takeaways & Limitations

    The results highlight PASS as a promising reconfigurable antenna technology for next-generation wireless systems.

Abstract

from arXiv · show

Pinching-antenna systems (PASSs) are a recent flexible-antenna technology that is realized by attaching simple components, referred to as pinching elements, to dielectric waveguides. This work explores the potential of deploying PASS for uplink and downlink transmission in multiuser MIMO settings. For downlink PASS-aided communication, we formulate the optimal hybrid beamforming, in which the digital precoding matrix at the access point and the location of pinching elements on the waveguides are jointly optimized to maximize the achievable weighted sum-rate. Invoking fractional programming and Gauss-Seidel approach, we propose two low-complexity algorithms to iteratively update the precoding matrix and activated locations of the pinching elements. We further study uplink transmission aided by a PASS, where an iterative scheme is designed to address the underlying hybrid multiuser detection problem. We validate the proposed schemes through extensive numerical experiments. The results demonstrate that using a PASS, the throughput in both uplink and downlink is boosted significantly as compared with baseline MIMO architectures, such as massive MIMO~and classical hybrid analog-digital designs. This highlights the great potential of PASSs, making it a promising reconfigurable antenna technology for next-generation wireless systems.

I. INTRODUCTION

The paper positions PASS as a low-cost, flexible approach for reconfiguring wireless channels through movable pinching elements on dielectric waveguides. It extends PASS research to multiuser MIMO uplink and downlink transmission with jointly designed antenna locations and digital processing.

  • Motivation: Higher-frequency path loss motivates directional arrays and technologies that can actively reconfigure the wireless channel.Massive MIMO compensates attenuation through highly directional beams, while channel reconfiguration challenges fixed transmission models.
  • Related technologies: Earlier reconfigurable technologies include antenna selection, IRSs, fluid antennas, and mechanically movable antennas.These approaches modify antenna positions, phase shifts, liquid conductors, or mechanical element locations to alter propagation properties.
  • PASS motivation: PASS is presented as an alternative addressing real-time control, implementation cost, complexity, and large-scale path-loss challenges.The paper motivates PASS through its low-cost pinching elements and flexibility in forming line-of-sight links.
  • PASS motivation: PASS uses movable pinching elements attached to dielectric waveguides to flexibly establish or strengthen line-of-sight links.The elements are described as low-cost components whose locations can be tuned along the waveguides.
  • Paper scope: This study extends prior PASS work to multiuser MIMO beamforming and detection for an access point equipped with multiple waveguides and reconfigurable elements.The system uses an array of waveguides, each equipped with multiple pinching elements whose locations are design parameters.

A. Characterizing Downlink Channel

The PASS channel is determined by phase-shifted radiation from pinching elements whose locations control the effective waveguide-user channels. The resulting downlink broadcast and uplink multiple-access channels are reciprocal and location-reconfigurable.

  • Downlink channel: Each waveguide radiates phase-shifted versions of its fed signal through its pinching elements.The phase at each element is determined by its location along the waveguide, while the radiation vector includes attenuation and phase terms.
  • Propagation model: The model assumes line-of-sight propagation dominates non-line-of-sight paths in classical indoor scenarios.Users are assumed to be in the waveguides' LoS, with non-LoS signals negligible in amplitude and separable in delay.
  • Propagation model: Shadowing is assumed invariant across pinching elements on a waveguide, although varying shadowing would require coefficients indexed by element.The paper identifies this as a practical approximation for limited areas such as moderate-size indoor environments.
  • Downlink channel: The end-to-end channel depends on the location matrix L, making the vector Gaussian broadcast channel tunable through activated pinching-element positions.The effective channel vectors are assembled across waveguides into G(L).
  • Uplink channel: In the uplink, waveguides receive superimposed user signals, yielding a Gaussian multiple-access channel whose matrix is the transpose of the downlink channel under TDD reciprocity.The effective uplink channel remains reconfigurable through pinching-element locations.

III. HYBRID BEAMFORMING IN A DOWNLINK PASS

Downlink transmission uses a digital precoder together with tunable pinching-element locations, forming a hybrid beamforming design. The transmitted waveguide signals are linear combinations of user information signals under a power constraint.

  • Hybrid beamforming: The PASS downlink precoder and pinching-element locations are jointly optimized because digital and analog beam design are coupled.The digital precoder is W, while the location matrix L controls analog beam design.
  • Signal model: The AP forms each waveguide input as a linear superposition of the users' encoded information signals.The aggregate transmitted signal is represented as z = Ws, where W is the precoding matrix.
  • Signal model: The transmitted average power is constrained by a positive downlink power budget Pd.This constraint applies to the precoded waveguide-array signal.

B. Downlink Weighted Sum-Rate

The downlink design maximizes weighted sum-rate over digital beamforming and physically feasible pinching-element locations. Minimum spacing and waveguide boundaries constrain the locations, while permutation invariance simplifies the formulation under uniform shadowing.

  • Optimization objective: The weighted sum-rate is constructed from user rates determined by their SINRs under W and L.The digital beamforming matrix and location matrix jointly affect each user's achievable rate.
  • Optimization objective: The objective is to maximize achievable weighted sum-rate under a downlink power budget and physical PASS constraints.User weights λk are non-negative and proportional to expected quality of service.
  • Physical constraints: Pinching-element locations must remain within waveguide lengths and maintain a minimum spacing Δℓ between neighboring elements.The spacing prevents mutual coupling and reflects sliding-track implementation.
  • Problem simplification: Permuting pinching elements on a waveguide leaves the weighted sum-rate unchanged under the stated channel model.The invariance follows because locations affect the rate through channel sums, which are unchanged by permutation.
  • Problem simplification: If shadowing varies across pinching elements, the permutation invariance no longer applies and the original hybrid beamforming problem must be solved.The corresponding path-loss coefficients then depend on both user, waveguide, and element indices.

IV. EFFICIENT ALGORITHM FOR HYBRID BEAMFORMING

The hybrid beamforming problem is non-convex, so the paper reformulates it into an unconstrained variational problem whose solution preserves the original power-constrained optimum. This reformulation provides the basis for an efficient iterative solver.

  • The original weighted sum-rate optimization is non-convex, and its global solution is not computable in polynomial time.
  • The method first removes the explicit power constraint through an unconstrained variational formulation and then recovers the optimal precoder by scaling.
  • The optimal precoding solution satisfies the downlink power constraint with equality.
  • Scaling an underpowered feasible precoder to meet the power limit increases the weighted sum-rate, proving that inequality cannot be optimal.
  • The variational solution determines the original optimal precoder, with the power limit incorporated through the auxiliary objective and final scaling.

B. Solution via Fractional Programming

Fractional programming transforms the variational sum-of-log-ratios problem into smoother dual objectives that can be approximated through block-coordinate updates. The resulting procedure alternates updates of auxiliary variables and beamforming variables.

  • The feasible variational problem jointly optimizes the precoding matrix W and location variables L.
  • Fractional programming converts the sum-of-log-ratios objective into quadratic forms using Lagrange and quadratic dual transforms.
  • The Lagrange-dual formulation preserves the original optimum but remains non-convex, so block-coordinate descent alternates marginal updates.
  • Quadratic duality produces a smoother objective landscape, enabling approximate optimization despite remaining non-convexity.
  • The iterative updates alternately solve for auxiliary variables q and the joint design variables S.

THGT (L) W

The solver uses an outer block-coordinate loop for digital precoding and pinching-element locations, with a Gauss-Seidel inner loop for sequential location updates. Location subproblems are solved by grid search because oscillatory objectives create many stationary points.

  • The transformed marginal problem is non-convex and has many local minima caused by its oscillating objective.
  • For fixed locations, the digital precoder is updated through a standard regularized linear inverse problem.
  • The digital precoder is equivalent to regularized zero-forcing with an effective channel determined by G(L), T, and U.
  • For fixed precoding, the location update uses a Gauss-Seidel scheme that sequentially optimizes one pinching-element location while holding the others fixed.
  • Each scalar location problem is solved by grid search over a feasible interval while enforcing minimum-distance constraints.
  • Classical gradient methods are inefficient for the scalar location problem because cosine-induced oscillations create numerous stationary points.

E. Final Algorithm: Convergence and Complexity

The final FP-BCD algorithm alternates dual-variable, precoder, and location updates until the objective improvement is sufficiently small. It converges to a local maximizer, with complexity dominated by the classical RZF precoder computation.

  • Algorithm 1 updates dual variables, computes auxiliary matrices, updates W, and sequentially updates every location in L.
  • Convergence: The algorithm converges to a stationary point because each block update is non-decreasing while the power constraint bounds the weighted sum-rate.
  • Convergence: Because the stationary point combines marginal maximizers, it is also a local maximizer.
  • The iterations stop when the fractional objective increase falls below ε and return the precoding matrix W and location matrix L.
  • Complexity: The computational complexity is dominated by the RZF precoder update, which has the same order as classical linear precoding.

F. Alternative Algorithm via Zero-Forcing

The section develops a single-loop zero-forcing alternative for optimizing PASS element locations and effective precoding, reducing the computational burden of the earlier iterative design for large systems. The resulting algorithm uses coordinate-wise grid searches and converges to a local minimum.

  • Alternative Algorithm via Zero-Forcing: The alternative zero-forcing scheme optimizes the PASS location matrix for effective ZF without solving the dual outer-loop problem.It directly optimizes the ZF precoding matrix, eliminating the BCD loop used in Algorithm 1.
  • Alternative Algorithm via Zero-Forcing: Assuming M > K, the effective downlink channel can be zero-forced for a fixed location matrix L.
  • Alternative Algorithm via Zero-Forcing: Algorithm 2 initializes feasible precoding and location matrices, searches each location, and returns the ZF digital precoder with the optimized location matrix.Iterations stop when the fractional objective decrease falls below ε.
  • Alternative Algorithm via Zero-Forcing: The location optimization updates one element location at a time using Gauss-Seidel coordinate optimization and grid search.The Sherman-Morrison lemma provides a rank-1 decomposition that reduces the cost of repeatedly evaluating the matrix-inversion trace term.
  • Alternative Algorithm via Zero-Forcing: The algorithm converges to a local minimum because its objective is non-decreasing across iterations and bounded.

V. MULTIUSER DETECTION IN AN UPLINK PASS

The uplink PASS problem jointly designs digital multiuser detection and activated pinching-element locations to optimize throughput. Unlike the downlink case, the optimal digital detector for a fixed location configuration is directly characterized by the MMSE detector.

  • MULTIUSER DETECTION IN AN UPLINK PASS: Uplink PASS transmission is modeled as a Gaussian multiple access channel in which the receiver jointly designs its matrix and activated pinching-element locations.
  • MULTIUSER DETECTION IN AN UPLINK PASS: For uplink design, the digital detector can be characterized directly by the MMSE solution, enabling more efficient algorithmic development than the downlink case.
  • MULTIUSER DETECTION IN AN UPLINK PASS: The users transmit independent zero-mean unit-variance signals, each scaled by the uplink power P_u.
  • MULTIUSER DETECTION IN AN UPLINK PASS: The access point estimates each user signal with a digital linear filter combined with an analog receive beam determined by the location matrix L.
  • MULTIUSER DETECTION IN AN UPLINK PASS: The hybrid receiver design jointly optimizes the digital filters and analog pinching-element locations because the detected signals depend on both components.

B. Uplink Weighted Sum-Rate

The uplink weighted sum-rate formulation measures PASS receiver performance and reduces joint receiver-location optimization by substituting the optimal MMSE detector. The remaining location problem is solved through Gauss-Seidel updates and grid search in Algorithm 3.

  • Uplink Weighted Sum-Rate: The uplink objective is the achievable weighted sum-rate, formed by weighting each user’s achievable rate according to its uplink SINR.
  • Uplink Weighted Sum-Rate: The goal is to jointly design the digital receiver matrix M and location matrix L to maximize the uplink achievable weighted sum-rate.
  • Uplink Weighted Sum-Rate: Each per-user SINR is a generalized Rayleigh quotient maximized by setting the corresponding digital receiver to the MMSE detector.
  • Uplink Weighted Sum-Rate: After substituting the MMSE detector, the weighted sum-rate depends only on L, reducing the joint problem to throughput optimization over pinching-element locations.
  • Uplink Weighted Sum-Rate: Algorithm 3 applies Gauss-Seidel updates and grid search to the location variables, stopping when the fractional objective increase falls below ε before determining the MMSE detector.

VI. NUMERICAL INVESTIGATIONS

Numerical experiments evaluate convergence, grid-resolution effects, and uplink/downlink weighted sum-rate performance for MIMO-PASS against conventional MIMO, massive MIMO, and hybrid beamforming baselines.

  • Experimental setup: The experiments assess algorithm convergence and compare MIMO-PASS with classical MIMO baselines in a common simulated environment.The baselines include conventional MIMO, massive MIMO, and hybrid beamforming-based MIMO.
  • Convergence: The proposed downlink and uplink algorithms converge to stable solutions, with the considered algorithms converging within 5 iterations.The FP-BCD, ZF-based, and MMSE-based methods are evaluated through iterative weighted sum-rate optimization.
  • Grid Search Resolution: Weighted sum-rates for both uplink and downlink increase monotonically with grid search resolution, and L = 10^5 nearly attains the PASS performance upper bound.The grid resolution determines the number of discrete candidate locations along each waveguide.
  • Grid Search Resolution: MIMO-PASS gains more than 30% over fully digital massive MIMO with the same number of antennas and over 200% over small-scale fully digital MIMO.The reported gains are attributed to flexibly tuned pinching locations that establish strong and stable line-of-sight paths.
  • Downlink Weighted Sum-Rate: In downlink, PASS throughput exceeds conventional fixed-location systems particularly at medium and high power, while FP-BCD is advantageous over ZF-based beamforming at low power.At medium and high power, ZF-based and FP-BCD methods achieve nearly the same performance.
  • Downlink Weighted Sum-Rate: Increasing the number of pinching elements improves weighted sum-rate, with PASS gains of 27% over mMIMO and 76% over hMIMO.Adding elements increases spatial degrees of freedom and array gains, while location tuning provides the reported improvements.
  • Uplink Weighted Sum-Rate: For uplink, PASS outperforms fixed-location systems across the considered transmit powers, but when K > M its sum-rate decays rapidly because of insufficient hybrid-system degrees of freedom.When K ≤ M = 5, PASS achieves higher sum-rate than the baseline schemes; fully digital mMIMO can outperform PASS when M = 8.
Loading 2503.03117v1…