Low-Complexity CSI Acquisition Exploiting Geographical Diversity in Fluid Antenna System

Christos Masouros; David Morales-Jimenez; Hao Jiang; Jian Dang; Zaichen Zhang; Zhentian Zhang

REVIEW 3 major objections 5 minor 1 cited by

Low-Complexity CSI Acquisition Exploiting Geographical Diversity in Fluid Antenna System

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A geographical-prior EM-AMP variant estimates fluid-antenna channels more accurately, converges faster, and cuts CSI acquisition cost by roughly 40–85%.

desk verdict The distance-based EM-AMP idea is worth considering, but the variance update has a sign error that invalidates the core derivation and the reported results. read the letter →

arxiv 2509.08598 v1 pith:P3Y7O5X4 submitted 2025-09-10 cs.IT math.IT

Zhentian Zhang , David Morales-Jimenez , Jian Dang , Zaichen Zhang , Christos Masouros , Hao Jiang This is my paper

classification cs.ITmath.IT

keywords fluidantennasystemchannelestimationapproximatemessagepassingexpectationmaximizationCSIacquisitiongeographicalpriormassiveconnectivitylarge-scalefading

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

Fluid antenna systems (FAS) need accurate channel estimates to choose among many switchable ports, but existing greedy and model-free estimators either lean on fragile signal assumptions or scale poorly. This paper argues that a well-known iterative Bayesian receiver, EM-AMP (expectation-maximization approximate message passing), can be adapted to FAS by folding in a cheap geographical prior: for a given user, all ports share the same large-scale channel variance, which is set by the user's distance through an inverse-square path-loss law. The proposed update replaces the free variance learning step of ordinary EM-AMP with a clamped, closed-form variance derived from the EM objective, using intermediate statistics from the AMP recursion. Simulations show the result estimates channels more precisely at low SNR, detects active users at least as well, converges faster, and lowers complexity by roughly 40–85% compared with least-squares and AoA-codebook baselines. If right, it makes scalable massive-connectivity CSI acquisition for FAS practical without extra piloting or prior channel statistics.

What carries the argument

The load-bearing object is the EM-AMP variance update with a geographical constraint. EM-AMP alternates an approximate message passing step that computes posterior means and variances of a Bernoulli-Gaussian sparse signal with an expectation-maximization step that learns the prior parameters. The paper's contribution is to replace the unconstrained EM variance estimate with φ^{t+1} = Σ V_{k,n} / Σ π_{k,n}, derived by maximizing the EM objective under the assumption that all ports of user k share the variance ς_k = f(d_k) = d_k^{-2}, and to clamp this value to [φ_min, φ_max] computed from the known user-distance interval [d_ref, d_max]. This single identity turns geographical knowledge into a

What would settle it

A controlled measurement or simulation in which a user at fixed distance has independently shadowed ports: if the per-port variances differ by more than [φ_min, φ_max] and the proposed clamped update then shows worse NMSE than conventional EM-AMP, the identical-variance premise is falsified.

Watch

Extended reading notes

Core claim

The central claim is that the geographical structure of FAS—each user's channel variance across all No ports being the same large-scale fading coefficient ςk = d_k^{-2}, with the user distance known to lie between d_ref and d_max—can be injected into EM-AMP as a constrained variance update and that this is the EM-optimal way to use that prior. Concretely, the paper derives the maximizer of the EM objective for the prior variance as φ = Σ_n V_{k,n} / Σ_n π_{k,n}, Eq. (21), where V_{k,n} is computed from the posterior mean and variance already available from the AMP step, Eq. (19b), and then clamps the result to [φ_min, φ_max] implied by the distance bounds. This replaces the unconstrained per

Load-bearing premise

The method assumes every fluid-antenna port of a given user sees exactly the same channel variance, equal to the user's large-scale fading coefficient following an inverse-square distance law, and that the user distance is known to lie in a prespecified range.

Editorial extensions

If this is right

FAS channel acquisition becomes tractable at massive-connectivity scale: per-iteration complexity O(4KGNo + 3KNo + 2K) does not depend on the number of active users Ka, so cost stays flat as traffic grows.
The error floor that appears in greedy and conventional EM-AMP NMSE at higher SNR can be pushed down by using the distance-derived variance clamp, not by adding pilots.
With the proposed update, the BS can estimate channels and detect active users simultaneously from the same compressed measurements, since the EM-AMP framework already carries activity probabilities λk.
Antenna-gap and angular information change the achievable estimation precision, so the same framework can be tuned by physical FAS geometry rather than by heuristic regularization.
Compared with least-squares at No=62, complexity drops 53.46% for the proposed scheme (85.03% for conventional EM-AMP), making the geographic prior a cheap add-on rather than a computational burden.

Reading between the lines

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

Inference: if per-port shadowing breaks the identical-variance premise, the clamp could be replaced by per-port variance bounds or by estimating an effective distance from the variance—the closed-form update would likely survive with a wider search interval.
Inference: The same geographical prior should transfer to cell-free or distributed FAS deployments, where each access point sees the same user distance geometry; a multi-AP variant could use the variance clamp per AP and potentially improve user separation.
One could test directly whether the gain comes from the EM-optimal variance formula or from the clamping alone by comparing Eq. (21) with a version that clamps the conventional E3 update to the same [φ_min, φ_max]; the paper does not isolate these two effects.
The paper's own results hint that AoA information is complementary: at many ports, an AoA-codebook estimator achieves the lowest NMSE, so a hybrid that feeds both angle and distance priors into EM-AMP is a natural next step.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

Quick take: the paper applies EM-AMP to FAS channel estimation and adds a distance-based variance bound. That is a reasonable niche to explore, and the complexity discussion is fine. But the central update is not derived correctly, so the numerical claims don't stand as written.

The new bit is narrow: it's Vila-Schniter EM-AMP with a clamped, port-common variance learned from the assumption phi = d^{-2}. The authors do cite the relevant FAS CSI prior art and compare against sensible baselines (SOMP+LS, AoA codebook, conventional EM-AMP). The runtime numbers and the observation about existing error floors are useful.

The soft spot is load-bearing. Eq. (19b) computes V = (E[x] - mu)^2 - Var(x). The correct second-moment identity is (E[x] - mu)^2 + Var(x). With the minus sign, the update in Eq. (21) and Algorithm 1 line E3 can go negative, and the clamp to phi_min then turns what should be EM into a heuristic. The sign error isn't cosmetic; it's the whole "geographical prior" step. There are also normalization problems in the pseudo-code: the lambda update uses 1/K instead of 1/N_o, and the mu update sums over the wrong index. I haven't seen code or data, so I can't check whether the simulations reflect the printed algorithm; and the synthetic channels are generated with the same d^{-2} law the algorithm assumes, which makes the comparison against non-geographical baselines partially self-referential.

Who is this for? Someone working on FAS CSI acquisition might find the direction worth pursuing: the idea of constraining the EM variance by distance-derived bounds is salvageable with a corrected update, e.g. phi_new = phi_tilde + (x_tilde - mu)^2, or by computing the active-component integral properly. But this manuscript doesn't establish the method. I would not cite it in its current form. If the editor is willing to require a corrected derivation and better normalization, it's worth one round of referee time; otherwise it's a revise-and-resubmit or reject. I lean toward sending it out with a clear request to fix the math, because the underlying idea is legitimate and the subfield is active.

Referee Report

3 major / 5 minor

Summary. The manuscript proposes an EM-AMP-based channel estimation and active-user detection algorithm for fluid antenna systems (FAS). The key innovation is a geographical-prior variance update: since the large-scale fading coefficient is modeled as ς_k = f(d_k) = d_k^{-2}, the prior variance of each user's channel is updated via an EM step that is clamped to [φ_min, φ_max] computed from the known distance range [d_ref, d_max]. The paper claims improved NMSE and activity-detection error, faster convergence, and lower complexity than greedy, least-squares, AoA-codebook, and conventional EM-AMP baselines. Simulations are provided for a synthetic FAS uplink with Bernoulli-Gaussian activity.

Significance. If the proposed update were correctly derived, the paper would make a useful contribution: it offers a low-complexity O(GKNo) model-based alternative for FAS CSI acquisition and targets the error-floor issues previously reported for greedy FAS estimators. The geographical-prior idea is well motivated, and the complexity analysis is straightforward. However, the central EM derivation contains a sign error in the second-moment identity, and the EM updates for λ and μ contain normalization/index errors. These are load-bearing: they invalidate the claimed EM-optimality of the variance update and the numerical results as presented. The idea is salvageable with corrected updates, but the manuscript in its current form does not establish the claimed performance.

major comments (3)

The identity in Eq. (19b) is incorrect. For a random variable X and constant μ, E[|X−μ|²] = Var(X) + |E[X]−μ|², not Var(X) subtracted from (E[X]−μ)². The published expression also mixes full-posterior statistics with the active-component integral over B̄_ε; the correct EM update for the Gaussian variance is φ = Σ_n π(ν+|γ−μ|²) / Σ_n π. With the printed minus sign, the update can return negative variances, especially at low SNR, and the subsequent clipping to φ_min turns the claimed EM-optimal geographical update into a clamped heuristic. Because Eq. (21), Eq. (15b), and Algorithm 1 line E3 all inherit this error, the reported NMSE/ADE and convergence results cannot be traced to a correct derivation.
The activity-probability update is λ_k^{t+1} = (1/K) Σ_{n=1}^{No} π_{k,n}, but the average should be over the No ports, i.e., (1/No) Σ_n π_{k,n}. The use of 1/K scales λ incorrectly and biases the Bernoulli-Gaussian prior for massive K. The mean update μ_{k,n}^{t+1} = Σ_{k=1}^{K} π_{k,n} γ_{k,n} / (λ_k K) is also inconsistent: the right-hand side sums over k and is independent of the left-hand index k,n. These errors affect both the EM parameter estimates and the final activity detection, so they are not merely typographical.
The simulation setup is a matched-model test. The synthetic channels are generated with ς_k = f(d_k) = d_k^{-2}, and the algorithm is given exactly this f and the distance interval [d_ref, d_max] to set φ_min and φ_max. This is circular for the purpose of demonstrating an advantage over baselines that do not use the geographical prior. The paper provides no experiments under shadowing, a mismatched path-loss exponent, or per-port variance variation. Without such robustness checks, the claimed benefit of the 'geographical diversity' prior over model-agnostic methods is not established.

minor comments (5)

The entry 'ddef Far field lower range' appears to be a typo for d_ref. The values of φ_min and φ_max are never explicitly stated; they can be inferred as f(d_max) and f(d_ref), but should be defined.
Algorithm 1 is captioned 'Algorithm Baseline I, EM-AMP for FAS', which is confusing. It appears to be the proposed algorithm with geographical clipping, not a baseline. The authors should clarify which pseudocode corresponds to the conventional EM-AMP and which to the proposed scheme.
There is a typographical error: 'ˆφx k.n' should be 'ˆφx_{k,n}'.
In Fig. 2, the inset text 'SNR=0 dB' is unexplained and appears misplaced. The running-time measurements are hardware-specific and should be described more carefully (e.g., averaged over how many runs).
No code or data are provided. Given the algorithmic errors above, this makes it impossible for the reader to verify whether the reported curves are produced by the equations in the paper or by a different implementation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the EM-AMP derivation is self-contained (an algebraic sign error in Eq. 19b and the model-matched simulation are correctness/evaluation concerns, not circularity).

full rationale

I walked the claimed derivation chain: the BG prior (5), the AMP posterior (11)-(13), the EM objective (16), the split integral (18), and the variance update (21). The update is obtained by maximizing the expected complete-data log-likelihood with respect to the prior variance; the variance is an optimization variable, not the target prediction. The geographical prior enters as the modeling assumption phi = f(d_k) with hard bounds from [d_ref, d_max]. This is a stated model assumption, not a fitted parameter, and the EM step still estimates phi from data before clamping. The simulation uses the same f(d)=d^{-2} in the generative LSFC and in the prior, making the benchmark favorable to the proposed method, but the NMSE/ADE results are empirical simulation outcomes rather than identities forced by the derivation. The paper's self-citations (e.g., [15], [18]-[21], [27]) appear as background, baseline, or implementation references and are not load-bearing for the claimed EM-AMP optimality. One non-circular but important correctness issue is Eq. (19b): the active-region integral equals pi(nu + |gamma-mu|^2), whereas the printed expression computes (E[x]-mu)^2 - Var[x]; the sign is wrong. This is a mathematical error in the derivation, not an equivalence of output to input, so it does not make the paper circular. On the circularity axis the paper scores 0: no step reduces a prediction to its inputs by construction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim rests on a domain assumption about the channel prior, an unproven finite-size validity of AMP, and a false algebraic identity in the variance update. The clipping bounds φmin and φmax and the initialization formula are hand-chosen inputs.

free parameters (3)

φmin and φmax (variance clipping bounds) = not stated numerically; derived from [d_ref,d_max] via f(d)=d^{-2}
The algorithm clamps the EM variance update to this interval; the bounds encode the geographical prior and are hand-selected from the assumed distance range.
Initial activity probability λ^1_k = computed by a max-formula in Algorithm 1 (I1)
The initialization rule is stated without derivation and contains a maximization over an auxiliary variable a; no justification or reference is given.
Noise variance ψ = input parameter
The algorithm takes ψ as known, whereas the abstract claims model-free adaptive learning; this is an input requirement.

assumptions (3)

domain assumption The FAS channel matrix X follows a Bernoulli-Gaussian prior with per-user activity and per-port variance equal to the large-scale fading coefficient ς_k = f(d_k)
Introduced in Sec. II-B and Sec. III-A; the algorithm's variance update is built on this. Real FAS channels may include shadowing or per-port variations that break the exact equality.
domain assumption AMP's Gaussian approximations and EM convergence hold for the finite system sizes simulated (G=400, K=1000, No up to 62)
The AMP framework from [17] assumes large i.i.d. random matrices; no state-evolution proof is given for these finite sizes.
ad hoc to paper The identity in Eq. (19b), V = ∫ p(x|y)|x-μ|^2 dx = (E[x|y]-μ)^2 - Var(x|y), is valid
This is asserted in the derivation of the variance update but is algebraically false; the correct identity uses the plus sign. The central claim depends on it.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Low-Complexity CSI Acquisition Exploiting Geographical Diversity in Fluid Antenna System." pith.science (2026). https://pith.science/paper/P3Y7O5X4

@misc{pith2026250908598,
  author       = {Pith},
  title        = {Pith review of: Low-Complexity CSI Acquisition Exploiting Geographical Diversity in Fluid Antenna System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P3Y7O5X4}},
  note         = {Machine review of arXiv:2509.08598}
}

read the original abstract

The fluid antenna system (FAS) employs reconfigurable antennas for high spatial gains in compact spaces, enhancing physical layer flexibility. Channel state information (CSI) acquisition is vital for port selection and FAS optimization. Greedy algorithms rely on signal assumptions, and model-free methods face high complexity. A flexible, low-complexity solution is needed for massive connectivity in FAS. Based on expectation maximization-approximate message passing (EM-AMP) framework, efficient matrix computations and adaptive learning without prior model knowledge naturally suit CSI acquisition for FAS. We propose a EM-AMP variant exploiting FAS geographical priors, improving estimation precision, accelerating convergence, and reducing complexity in large-scale deployment. Simulations validate the efficacy of the proposed algorithm.

Figures

Figures reproduced from arXiv: 2509.08598 by the authors.

**Figure 2.** Illustration of ADE, NMSE and running time (s) of diffe [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Illustration of ADE, NMSE and running time (s) of diffe [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

Beyond Covariance: Generative Spatial Correlation Modeling and Channel Interpolation for Fluid Antenna Systems
cs.IT 2026-04 unverdicted novelty 7.0 of 10

FAS channels are represented as AR(p) Gauss-Markov processes to derive the optimal MMSE interpolator, a tight lower bound on required observations, and a Kalman filter achieving that optimum with O(N) complexity.

Reference graph

Works this paper leans on

27 extracted references · 1 canonical work pages · cited by 1 Pith paper

[1]

Massive access for 5G and beyond,

X. Chen, et al. , “Massive access for 5G and beyond,” IEEE J. Select. Areas Commun. , vol. 39, no. 3, pp. 615–637, Mar. 2021

2021
[2]

A tutorial on ﬂuid antenna system for 6G networks: Encompassing communication theory, optimization methods and hard- ware designs,

W. K. New, et al. , “A tutorial on ﬂuid antenna system for 6G networks: Encompassing communication theory, optimization methods and hard- ware designs,” IEEE Comm. Surveys & Tutorials , Early Access , doi: 10.1109/COMST.2024.3498855

arXiv 2024
[3]

Fluid antenna system for 6G: When Bruce Lee inspires wireless communications,

K.-K. Wong, et al. , “Fluid antenna system for 6G: When Bruce Lee inspires wireless communications,” Elect. Lett. , vol. 56, no. 24, pp. 1288–1290, Nov. 2020

2020
[4]

Performance limits of ﬂuid antenna systems,

K. K. Wong, et al., “Performance limits of ﬂuid antenna systems,” IEEE Commun. Lett. , vol. 24, no. 11, pp. 2469–2472, Nov. 2020

2020
[5]

On fundamental limits of ﬂuid antenna-assisted inte- grated sensing and communications for unsourced random acc ess,

Z. Zhang, et al. , “On fundamental limits of ﬂuid antenna-assisted inte- grated sensing and communications for unsourced random acc ess,” IEEE J. Sel. Area Commun. , Early Access , doi: 10.1109JSAC.2025.3608113

arXiv 2025
[6]

Fluid antenna systems,

K.-K. Wong, et al. , “Fluid antenna systems,” IEEE Trans. Wireless Commun., vol. 20, no. 3, pp. 1950–1962, Mar. 2021

1950
[7]

A new spatial block-correlation model for ﬂuid antenna systems,

P . Ram´ ırez-Espinosa,et al. , “A new spatial block-correlation model for ﬂuid antenna systems,” IEEE Trans. Wireless Commun. , vol. 23, no. 11, pp. 15829-15843, Nov. 2024

2024
[8]

Dynamic channel modeling of ﬂuid antenna systems in UA V communications,

H. Jiang, et al., “Dynamic channel modeling of ﬂuid antenna systems in UA V communications,”IEEE Wirel. Commun. Lett. , Early Access, doi: 10.1109/LWC.2025.3588223

arXiv 2025

Show all 27 references

[9]

Fluid antenna multiple access,

K. K. Wong and K. F. Tong, “Fluid antenna multiple access, ” IEEE Trans. Wireless Commun. , vol. 21, no. 7, pp. 4801–4815, Jul. 2022

2022
[10]

On fundamental limits of slow-ﬂuid antenna multiple access for unsourced random access,

Z. Zhang, et al. , “On fundamental limits of slow-ﬂuid antenna multiple access for unsourced random access,” IEEE Wireless Commun. Lett. , Early Access , doi: 10.1109/LWC.2025.3594112

2025
[11]

Sparsity-exploiting channel estimation for un- sourced random access with ﬂuid antenna

K. Zhou, et al. , “Sparsity-exploiting channel estimation for un- sourced random access with ﬂuid antenna”, arxiv, Available: https://doi.org/10.48550/arXiv.2504.17634

doi arXiv
[12]

Channel estimation for FAS-assisted multiuser mmWave systems,

H. Xu, et al. , “Channel estimation for FAS-assisted multiuser mmWave systems,” IEEE Commun. Lett. , vol. 28, no. 3, pp. 632-636, March 2024

2024
[13]

Fluid antenna with linear MMSE channel estimation for large-scale cellular networks,

C. Skouroumounis, et al. , “Fluid antenna with linear MMSE channel estimation for large-scale cellular networks,” IEEE Trans. Commun. , vol. 71, no. 2, pp. 1112-1125, Feb. 2023

2023
[14]

Sparse bayesian learning-based channel estimation for ﬂuid antenna systems,

B. Xu, et al. , “Sparse bayesian learning-based channel estimation for ﬂuid antenna systems,” IEEE Wirel. Commun. Lett. , vol. 14, no. 2, pp. 325-329, Feb. 2025

2025
[15]

Successive bayesian reconstructor for channel estima- tion in ﬂuid antenna systems,

Z. Zhang, et al., “Successive bayesian reconstructor for channel estima- tion in ﬂuid antenna systems,” IEEE Trans. Wireless Commun. , vol. 24, no. 3, pp. 1992-2006, March 2025

1992
[16]

Message passing algorithms for compressed sensing: I. motivation and construction,

D. L. Donoho, et al. , “Message passing algorithms for compressed sensing: I. motivation and construction,” in Proc. IEEE Information Theory W orkshop on Information Theory , Cairo, Egypt, 2010, pp. 1- 5

2010
[17]

Expectation-Maximization Gaussian-Mixture Approximate Message Passing,

J. P . Vila and P . Schniter, “Expectation-Maximization Gaussian-Mixture Approximate Message Passing,” IEEE Trans. Signal Process. , vol. 61, no. 19, pp. 4658-4672, Oct.1, 2013

2013
[18]

Unsourced random access via random scattering with turbo probabilistic data association detector and treatin g collision as interference,

Z. Zhang, et al. , “Unsourced random access via random scattering with turbo probabilistic data association detector and treatin g collision as interference,” IEEE Trans. Wireless Commun., vol. 23, no. 12, pp. 17899- 17914, Dec. 2024

2024
[19]

Uncoupled unsourced random access: Exploiting geographical diversity of access points,

Z. Zhang, et al. , “Uncoupled unsourced random access: Exploiting geographical diversity of access points,” IEEE Trans. V eh. Technol., vol. 74, no. 6 , pp. 9882-9887, June 2025

2025
[20]

Probabilistic ODMA receiver with low-complexity algorithm for MIMO unsourced random access,

Z. Zhang, et al , “Probabilistic ODMA receiver with low-complexity algorithm for MIMO unsourced random access,” IEEE Trans. V eh. Technol., doi: 10.1109/TVT.2025.3570708

2025
[21]

Joint channel estimation and active user detection for cell-free massive access system exploiting coarse user location information,

J. Dang, et al. , “Joint channel estimation and active user detection for cell-free massive access system exploiting coarse user location information,” IEEE Internet Things J. , vol. 11, no. 8, pp. 14985-14999, 15 April15, 2024

2024
[22]

Massive access in cell-free massive MIMO-based Internet of Things: Cloud computing and edge computing paradigms,

M. Ke, et al., “Massive access in cell-free massive MIMO-based Internet of Things: Cloud computing and edge computing paradigms,” IEEE J. Sel. Areas Commun. , vol. 39, no. 3, pp. 756–772, Mar. 2021

2021
[23]

Compressive sensing based adaptive active user detectio n and channel estimation: Massive access meets massive MIMO,

M. Ke, et al., “Compressive sensing based adaptive active user detectio n and channel estimation: Massive access meets massive MIMO, ” IEEE Trans. Signal Process. , vol. 68, pp. 764–779, Jan. 2020

2020
[24]

Channel hardening and favora ble propagation in cell-free massive MIMO with stochastic geometry,

Z. Chen and E. Bj¨ ornson, “Channel hardening and favora ble propagation in cell-free massive MIMO with stochastic geometry,” IEEE Trans. Commun., vol. 66, no. 11, pp. 5205–5219, Nov. 2018

2018
[25]

What is the expectation maximiz ation algo- rithm?

C. Do and S. Batzoglou, “What is the expectation maximiz ation algo- rithm?” Nat. Biotechnol. , vol. 26, no. 8, pp. 897–899, Aug. 2008

2008
[26]

A view of the EM algorithm that jus tiﬁes incremental, sparse, and other variants,

R. Neal and G. Hinton, “A view of the EM algorithm that jus tiﬁes incremental, sparse, and other variants,” in Learning in Graphical Models, Cambridge, MA, USA: MIT Press, 1999, pp. 355–368

1999
[27]

Efﬁcient ODMA for unsourced random access in MIMO and hybrid massive MIMO,

Z. Zhang, et al. , “Efﬁcient ODMA for unsourced random access in MIMO and hybrid massive MIMO,” IEEE Internet Things J. , vol. 11, no. 23, pp. 38846-38860, Dec.1, 2024

2024

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

Massive access for 5G and beyond,

X. Chen, et al. , “Massive access for 5G and beyond,” IEEE J. Select. Areas Commun. , vol. 39, no. 3, pp. 615–637, Mar. 2021

2021

[2] [2]

A tutorial on ﬂuid antenna system for 6G networks: Encompassing communication theory, optimization methods and hard- ware designs,

W. K. New, et al. , “A tutorial on ﬂuid antenna system for 6G networks: Encompassing communication theory, optimization methods and hard- ware designs,” IEEE Comm. Surveys & Tutorials , Early Access , doi: 10.1109/COMST.2024.3498855

arXiv 2024

[3] [3]

Fluid antenna system for 6G: When Bruce Lee inspires wireless communications,

K.-K. Wong, et al. , “Fluid antenna system for 6G: When Bruce Lee inspires wireless communications,” Elect. Lett. , vol. 56, no. 24, pp. 1288–1290, Nov. 2020

2020

[4] [4]

Performance limits of ﬂuid antenna systems,

K. K. Wong, et al., “Performance limits of ﬂuid antenna systems,” IEEE Commun. Lett. , vol. 24, no. 11, pp. 2469–2472, Nov. 2020

2020

[5] [5]

On fundamental limits of ﬂuid antenna-assisted inte- grated sensing and communications for unsourced random acc ess,

Z. Zhang, et al. , “On fundamental limits of ﬂuid antenna-assisted inte- grated sensing and communications for unsourced random acc ess,” IEEE J. Sel. Area Commun. , Early Access , doi: 10.1109JSAC.2025.3608113

arXiv 2025

[6] [6]

Fluid antenna systems,

K.-K. Wong, et al. , “Fluid antenna systems,” IEEE Trans. Wireless Commun., vol. 20, no. 3, pp. 1950–1962, Mar. 2021

1950

[7] [7]

A new spatial block-correlation model for ﬂuid antenna systems,

P . Ram´ ırez-Espinosa,et al. , “A new spatial block-correlation model for ﬂuid antenna systems,” IEEE Trans. Wireless Commun. , vol. 23, no. 11, pp. 15829-15843, Nov. 2024

2024

[8] [8]

Dynamic channel modeling of ﬂuid antenna systems in UA V communications,

H. Jiang, et al., “Dynamic channel modeling of ﬂuid antenna systems in UA V communications,”IEEE Wirel. Commun. Lett. , Early Access, doi: 10.1109/LWC.2025.3588223

arXiv 2025