Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Performance Analysis of XL-MIMO with Rotary and Movable Antennas for High-speed Railway

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read For high-speed rail links with parallel, unrotated square antenna panels, the paper derives a closed-form spacing that makes the line-of-sight MIMO channel orthogonal, and shows that allowing panels to rotate adds up to 1.4x capacity.

desk verdict Useful ROMA/HSR framework, but the central spacing formula in Corollary 1 is not proven and rests on a false approximation. read the letter →

arxiv 2412.03940 v1 pith:FCG3JML2 submitted 2024-12-05 eess.SP cs.ITmath.IT

classification eess.SPcs.ITmath.IT
keywords XL-MIMOhigh-speedrailwayrotaryandmovableantennasspatialcorrelationchannelorthogonalitynear-fieldbeamtrainingdifferentialevolutionline-of-sightMIMO
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes a configuration framework for very large MIMO systems whose antenna panels on a high-speed train can rotate and whose element spacing can be adjusted. Working in a line-of-sight near-field channel, it derives an analytic condition for the channel matrix to be orthogonal and, for parallel unrotated square panels, a closed-form optimal spacing $d = \sqrt{|\lambda D^3/(N_H x_0 z_0)|}$. It then applies a population-based search algorithm called differential evolution to choose the four panel rotation angles, maximizing the rank of the channel correlation matrix. The authors report that this configuration gives up to about 1.4 times the capacity of fixed panels in simulations of a high-speed rail link.

What carries the argument

The engine of the argument is the factorization $G \propto F_{TX} P^* P F_{TX}^*$; the diagonal phase matrices do not affect diagonalization, so the channel gain matrix $R = P^*P$ carries all spatial multiplexing information. Its $(u,v)$ entry collapses to a product of two Dirichlet-like sums whose arguments contain coefficients $\eta_{ab}$, each a rational function of spacing, rotation angles, and the center coordinates $(x_0, y_0, z_0, D)$. Corollary 1's spacing formula is obtained by setting one such coefficient so that a zero of those sums lands on the off-diagonal indices; for arbitrary angles, the same matrix $R$ supplies the rank objective for the adaptive differential evolution optimizer, a population-based global search over the four rotation angles.

What would settle it

At $x_0 = 40$ m, $y_0 = 4$ m, $z_0 = 10$ m with $20 \times 20$ arrays and $f_c = 20$ GHz, compute the exact channel matrix from equation (4) at the Corollary 1 spacing and inspect the maximum off-diagonal magnitude of $H^*H$. The proof assumes $\eta_{11} \approx 0$, but with these parameters $\eta_{11} = d^2(y_0^2 + z_0^2)/D^3$ is comparable to $\eta_{12} = -d^2 x_0 z_0/D^3$, so nonzero off-diagonal entries would show that the quoted spacing does not actually orthogonalize the channel.

Watch

Extended reading notes

Core claim

Under a line-of-sight near-field model dominated by the direct path, the paper claims that the spatial correlation matrix of a ROMA-based XL-MIMO link can be factored so that its off-diagonal behavior is governed by two geometric sums. When the transmitter and receiver panels are parallel, unrotated, and uniformly spaced with the same spacing $d$ in both directions, Corollary 1 states that choosing $d = \sqrt{|\lambda D^3/(N_H x_0 z_0)|}$ makes the off-diagonal entries vanish, i.e., the channel becomes approximately orthogonal, which in the high-SNR regime maximizes capacity. For general rotations, the paper replaces the exact orthogonality condition with the rank of the channel gain matrix $R = P^*P$ and uses adaptive differential evolution to find the four angles that maximize that rank. Simulation results are presented as validation: the analytic spacing matches the point where capacity converges, and with $20 \times 20$ arrays at $x_0 = 40$ m the two-sided ROMA configuration yields 1.4 times the capacity of fixed panels.

Load-bearing premise

The spacing formula relies on treating the horizontal-axis correlation coefficient, called $\eta_{11}$, as negligible; if it is not negligible, or if the array is not square, the single spacing formula does not guarantee orthogonal channels.

Editorial extensions

If this is right

  • At the Corollary 1 spacing, capacity of a parallel-panel link converges to its limiting value, so designers can set spacing without exhaustive search.
  • Allowing both panels to rotate yields the reported 1.4x capacity gain over fixed panels at $x_0 = 40$ m, with single-sided rotation giving 1.15x.
  • The mobility-aware near-field localization model supports low-frequency beam training, so the optimal ROMA configuration can be updated as the train moves.
  • Because only the rank of the channel correlation matrix is optimized, the rotation-angle search does not require instantaneous channel state information, only position-based channel statistics.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension: evaluate the exact, unapproximated $H^*H$ at the Corollary 1 spacing for non-square arrays; the derivation's single spacing cannot zero both $\eta_{12}$ and $\eta_{21}$ independently, so a two-spacing variant would be the natural fix if off-diagonal leakage appears.
  • The same spacing-and-rotation analysis should transfer to other near-field line-of-sight links whenever the paraxial approximation holds, but the 1.4x gain is demonstrated at one SNR and one geometry, so the trade-off with mechanical rotation overhead remains to be mapped.
  • One could replace the discrete rank objective with a smooth proxy, such as the sum of log singular values, to see whether the differential evolution solution is a sharp peak or a broad plateau; a broad plateau would make the configuration robust to train-position error.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper studies a downlink XL-MIMO system with rotary and movable antenna panels for high-speed railway. It first proposes a localization model based on a mobility-aware near-field beam training algorithm from the authors' previous work [8]. It then derives the spatial correlation matrix for two rotated UPA panels, states Corollary 1 giving optimal antenna spacing d = sqrt(|λD^3/(N_H x0 z0)|) for parallel panels, and uses a differential evolution algorithm to optimize the four panel rotation angles by maximizing the rank of the correlation matrix. Numerical results include localization NMSE, capacity versus spacing, and capacity gains up to 1.4x at x0 = 40 m.

Significance. If the derivation were correct, the paper would provide a simple closed-form spacing rule and a rotation-optimization framework for near-field LoS HSR channels, both useful for system design. The correlation-matrix expression (17)-(22) for general 3D rotations is explicit and the DE formulation is reproducible, and the claimed capacity gain is a concrete falsifiable prediction. However, the central spacing result is not established: the proof of Corollary 1 rests on an approximation that is false in the simulated geometry, and the numerical validation does not directly test channel orthogonality. The localization component depends on an unavailable under-review reference. At present the manuscript is not sufficient for publication.

major comments (4)
  1. [Section III.B, proof of Corollary 1, Eqs. (24)-(28)] The step 'Based on the system model shown in Fig. 1, η11 ≈ 0' is not justified and is false for the geometry used in the simulations. With α1=β1=α2=β2=0, Eq. (18) gives η11 = (D^2 − x0^2)d^2/D^3 = (y0^2 + z0^2)d^2/D^3. For the values in Section IV (x0 = 40 m, y0 = 4 m, z0 = 10 m, D ≈ 41.5 m), η11 ≈ 116 d^2/D^3, which is of the same order as η12 = −x0z0 d^2/D^3 ≈ −400 d^2/D^3 and has the opposite sign. The approximation is not even directionally helpful: taking η11 to be exactly zero would make the first Dirichlet factor in Eq. (22) reduce to N_H for index differences with u2 = v2, yielding maximal rather than zero off-diagonal correlation. Thus Eq. (24) is not derived, and the claim that this spacing 'ensures the channel orthogonality condition is fulfilled' is unsupported.
  2. [Section III.B, Eqs. (22)-(24)] Even if η11 were negligible, the proof does not establish orthogonality for all off-diagonal index differences. The condition |N_H η12/λ| = 1 zeroes the first Dirichlet kernel only when the argument contains η12 in the appropriate way; for pairs with Δu1 ≠ 0 and Δu2 ≠ 0, the second Dirichlet kernel involves η21Δu1 + η22Δu2, and η22 = (D^2 − z0^2)d^2/D^3 is nonzero and unconstrained. With the simplified coefficients (25)-(28), for a square array with N_H = N_V, the product in Eq. (22) is not zero for all off-diagonal entries; e.g., for Δu1 = Δu2 = 1, neither kernel vanishes unless additional conditions on η22 (and η11) are enforced. The proof needs to show that the two one-dimensional summations can be nulled independently, or to state conditions on all four ηab coefficients.
  3. [Section III.A and Fig. 2] The localization model is based entirely on the authors' own beam-training method in reference [8], which is under review and not available to the reader; Eq. (7) simply postulates the receiver position without describing how the near-field beam training estimates (θt, Rt). Fig. 2 reports NMSE against two 'widely used' beam-training algorithms but does not specify their concrete implementations or parameters, and the results appear to be generated from the same model proposed by the authors without an independent error model. Consequently, the first component of the proposed framework cannot be independently assessed, even though the rotation-angle optimization and position-dependent spacing expression rely on accurate real-time position input.
  4. [Section IV, Fig. 3] The numerical validation does not directly test the orthogonality condition (6). Fig. 3 shows capacity versus spacing, and the dashed line from Eq. (24) 'closely approaches' the convergence region, but convergence of capacity does not imply that the channel matrix H is orthogonal; the eigenvalue distribution or the off-diagonal norm of G is not reported. Since Eq. (24) was derived from an incorrect approximation, the apparent agreement in Fig. 3 may be coincidental for the chosen geometry. The paper should either provide a direct test of condition (6) or report the rank and eigenvalue spread of G at the predicted spacing.
minor comments (6)
  1. [Section II, after Eq. (4)] The definition 'k = c/f' has incorrect dimensions; the wavenumber should be k = 2πf/c, and the phase expressions that follow should be checked for consistency.
  2. [Section II, Eq. (4) and capacity (5)] The Doppler frequency offset fd is included in f, making the channel time-dependent, but the correlation matrix G and the capacity expression use instantaneous values; please clarify the time index and how fd enters the numerical results.
  3. [Section II, Eq. (6)] The phrase 'H is an orthography matrix' should read 'H is an orthogonal matrix.'
  4. [Section III.B, Eq. (14)] The exponential expression in Eq. (14) has an unbalanced bracket; the final term is written with a parenthesis that is never closed.
  5. [Algorithm 1, line 7] The expression 'F02lamb' appears to be a formatting error for 'F0 · 2^lamb'; please clarify the intended formula and the notation for the dynamic mutation factor.
  6. [Section IV, Fig. 3] The quantity 'normalized antenna spacing' used as the horizontal axis in Fig. 3 is not defined in the text; specify the reference spacing used for normalization.

Circularity Check

1 steps flagged · score 2.0 of 10

No definitional or fitted-input circularity: the spacing law Eq. (24) is model-derived rather than fit to curves. The only self-referential element is the localization module imported from the authors' under-review [8], which is minor and not load-bearing; Corollary 1's unsupported eta11≈0 step is a correctness concern, not a circularity.

  1. other [Section III.A (Localization Model); Fig. 2]
    "We construct the localization model of the user’s location based on the predictive beam training model of our previous work [8]."

    This is a self-referential input: the localization result that feeds the position-dependent ROMA configuration is not derived or externally verified here, but borrowed from an under-review paper by the same authors. It is not load-bearing for Corollary 1 or the differential-evolution rotation optimization, which are self-contained given the coordinates, and no fitted parameter is renamed as a prediction; hence it contributes only a minor score increment rather than a full circularity.

full rationale

The claimed derivation chain is not circular: Eq. (24) is obtained by substituting the parallel, unrotated UPA geometry into Eqs. (17)-(22) and imposing |N_H eta12/lambda| = 1 from the orthogonality condition Eq. (23); no free parameter is fit to the capacity curves, and the dashed line in Fig. 3 is a computed spacing, not a fitted prediction. The capacity and rank simulations use the same LoS channel model from which the orthogonality condition was derived, so they provide internal consistency rather than independent empirical confirmation, but that is not a circular reduction. The one self-referential element is the localization model of Section III.A, which is taken from the authors' own under-review paper [8]; because Corollary 1 and the DE rotation optimization depend only on the position coordinates and not on the details of [8], this self-citation is minor and not load-bearing. Separately, the proof of Corollary 1 contains an unquantified 'eta11 approx 0' step that is dubious at the paper's simulation geometry and, if taken literally, would leave horizontal-only off-diagonal correlations at their maximum; this is a soundness defect in the derivation, not a case of the result being equivalent to its inputs by construction.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

No new physical entities are introduced; ROMA, XL-MIMO, and differential evolution are existing techniques. The only hand-chosen values that affect the optimization output are the differential evolution settings; no physical constants are fitted.

free parameters (1)
  • DE hyperparameters (F0, CR, population size, generation count) = F0=0.5, CR=0.2; population size and generation count not stated
    Chosen by hand; no sensitivity analysis is provided, and the reported rotation-angle results depend on them.
assumptions (6)
  • domain assumption LoS-dominant channel with amplitude 1/D and Fresnel phase expansion
    Eqs. (4), (8)-(10) ignore NLoS and truncate the spherical-wave phase to second order; plausible for open HSR but unvalidated against measured channels.
  • domain assumption Capacity is maximized when H has orthogonal columns
    Section II asserts that in high SNR LoS, orthogonal columns with equal norms maximize the sum of log eigenvalues; the equal-norm condition is not discussed.
  • standard math Orthogonality admits a solution when at least one eta_ab is zero
    Taken from [10]; used to reduce the 2D nulling problem to a single spacing condition.
  • domain assumption Train position can be predicted from two near-field beam training measurements
    Section III-A, Eq. (7); depends on the under-review reference [8], which is not available.
  • ad hoc to paper eta11 approximately 0 in the parallel-panel proof
    Corollary 1 proof claims eta11 approximately 0 without a quantitative condition; with the paper's own y0=4 m, z0=10 m and x0=40 m, eta11 is about 29% of eta12, so the approximation fails exactly where the capacity gain is reported.
  • domain assumption Square UPA at both ends
    Corollary 1 uses only N_H in the spacing formula; zeroing both dimensions requires the same element count per dimension, which is true in the simulations but not stated in the corollary.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Performance Analysis of XL-MIMO with Rotary and Movable Antennas for High-speed Railway." pith.science (2026). https://pith.science/paper/FCG3JML2

@misc{pith2026241203940,
  author       = {Pith},
  title        = {Pith review of: Performance Analysis of XL-MIMO with Rotary and Movable Antennas for High-speed Railway},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FCG3JML2}},
  note         = {Machine review of arXiv:2412.03940}
}
read the original abstract

The rotary and movable antennas (ROMA) technology is efficient in enhancing wireless network capacity by adjusting both the antenna spacing and three-dimensional (3D) rotation of antenna surfaces, based on the spatial distribution of users and channel statistics. Applying ROMA to high-speed rail (HSR) wireless communications can significantly improve system performance in terms of array gain and spatial multiplexing. However, the rapidly changing channel conditions in HSR scenarios present challenges for ROMA configuration. In this correspondence, we propose a analytical framework for configuring ROMA-based extremely large-scale multiple-input-multiple-output (XL-MIMO) system in HSR scenarios based on spatial correlation. First, we develop a localization model based on a mobility-aware near-field beam training algorithm to determine the real-time position of the train relay antennas. Next, we derive the expression for channel orthogonality and antenna spacing based on the spatial correlation matrix, and obtain the optimal antenna spacing when the transceiver panels are aligned in parallel. Moreover, we propose an optimization algorithm for the rotation angle of the transceiver panels, leveraging the differential evolution method, to determine the optimal angle. Finally, numerical results are provided to validate the computational results and optimization algorithm.

Figures

Figures reproduced from arXiv: 2412.03940 by the authors.

Figure 1
Figure 1. ROMA-enabled HSR XL-MIMO communication systems. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Localization NMSE against the antenna number of the transmitter [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Capacity versus normalized antenna spacing for different antenna [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Normalized channel capacity versus the x-axis location for three cases: parallel receiver and transmitter with FPA, one-sided plane with ROMA, and both planes with ROMA. that increasing the number of antennas reduces the error in the near-field beam training algorithms…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. ROMA: ROtary and Movable Antenna

    eess.SP 2025-01 conditional novelty 6.0 of 10

    ROMA arrays combining panel rotation and antenna movement can improve average spectral efficiency in multi-user MIMO systems.

Reference graph

Works this paper leans on

12 extracted references · 9 canonical work pages · cited by 1 Pith paper

  1. [8]

    Mobility-aware predictive beam training of extremely large-scale MIMO-OFDM systems for high- speed railway,

    W. Yi, J. Zhang, Z. Wang, H. Xiao, and B. Ai, “Mobility-aware predictive beam training of extremely large-scale MIMO-OFDM systems for high- speed railway,” IEEE Trans. Intell. Transp. Syst. , under review, 2024

  2. [1]

    Re- source allocation and slicing strategy for multiple services co-existence in wireless train communication network,

    Q. Ren, S. Lin, Y . Cai, X. Deng, L. Kong, S. Mumtaz, and B. Ai, “Re- source allocation and slicing strategy for multiple services co-existence in wireless train communication network,” IEEE Trans. Wireless Commun. , 2024

  3. [2]

    Reconfigurable intelligent surface- aided near-field communications for 6G: Opportunities and challenges,

    X. Mu, J. Xu, Y . Liu, and L. Hanzo, “Reconfigurable intelligent surface- aided near-field communications for 6G: Opportunities and challenges,” IEEE V eh. Technol. Mag., vol. 19, no. 1, pp. 65–74, Mar. 2024

  4. [3]

    A tutorial on extremely large-scale MIMO for 6G: Fundamentals, signal processing, and applications,

    Z. Wang, J. Zhang, H. Du, D. Niyato, S. Cui, B. Ai, M. Debbah, K. B. Letaief, and H. V . Poor, “A tutorial on extremely large-scale MIMO for 6G: Fundamentals, signal processing, and applications,” IEEE Commun. Surveys Tuts., vol. 26, no. 3, pp. 1560-1605, thirdquarter 2024

  5. [4]

    On the Spectral Efficiency of Movable and Rotary Access Points under Rician Fading,

    E. N. Tominaga, O. L. A. L ´opez, T. Svensson, R. D. Souza, and H. Alves, “On the Spectral Efficiency of Movable and Rotary Access Points under Rician Fading,” arXiv:2408.08112, 2024

  6. [5]

    6D Movable Antenna Enhanced Wireless Network Via Discrete Position and Rotation Opti- mization,

    X. Shao, R. Zhang, Q. Jiang, and R. Schober, “6D Movable Antenna Enhanced Wireless Network Via Discrete Position and Rotation Opti- mization,” arXiv:2403.17122, 2024

  7. [6]

    Fluid Antenna Systems,

    K.-K. Wong, A. Shojaeifard, K.-F. Tong, and Y . Zhang, “Fluid Antenna Systems,” IEEE Transactions on Wireless Communications , vol. 20, no. 3, pp. 1950–1962, Mar. 2021

  8. [7]

    Spatial multiplexing in near-field line-of-sight MIMO commu- nications: Paraxial and non-paraxial deployments,

    J. C. Ruiz-Sicilia, M. Di Renzo, P. Mursia, A. Kaushik, and V . Scian- calepore, “Spatial multiplexing in near-field line-of-sight MIMO commu- nications: Paraxial and non-paraxial deployments,” IEEE Trans. Green Commun. Netw., vol. 28, no. 4, pp. 1234–1245, 2024

Show all 12 references
  1. [9]

    Analog-digital beam focusing for line of sight wide-aperture MIMO with spherical wavefronts,

    Y . J. Yun, H. Rho, and W. Choi, “Analog-digital beam focusing for line of sight wide-aperture MIMO with spherical wavefronts,” arXiv:2404.04842, 2024

  2. [10]

    Optimal design of uniform rectangular antenna arrays for strong line-of-sight MIMO channels,

    F. Bøhagen, P. Orten, and G. E. Øien, “Optimal design of uniform rectangular antenna arrays for strong line-of-sight MIMO channels,” EURASIP J. Wirel. Commun. Netw. , 2007

  3. [11]

    An improved differential evolution algorithm and its application in optimization prob- lem,

    W. Deng, S. Shang, X. Cai, H. Zhao, Y . Song, and J. Xu, “An improved differential evolution algorithm and its application in optimization prob- lem,” Soft Comput. , vol. 25, pp. 5277–5298, 2021

  4. [12]

    Target sensing with intelligent reflecting surface: architecture and performance,

    X. Shao, C. You, W. Ma, X. Chen, and R. Zhang, “Target sensing with intelligent reflecting surface: architecture and performance,” IEEE J. Sel. Areas Commun., vol. 40, no. 7, pp. 2070-2084, Jul. 2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.