Pith. sign in

REVIEW 4 major objections 5 minor 26 references

A single receiver can track time-varying mmWave channels while simultaneously decoding data and performing over-the-air computing, approaching perfect-CSI bounds without perfect CSI.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 22:06 UTC pith:VIQWXGCJ

load-bearing objection The integration of known JCDE and AirComp components is legitimate, but Eq. (34a) appears to use the true channel inside the channel estimator, which if literal makes the whole no-perfect-CSI claim genie-aided. the 4 major comments →

arxiv 2509.07482 v1 pith:VIQWXGCJ submitted 2025-09-09 eess.SP

Integrated Communication and Computing in Time-Varying mmWave Channels

Joan \c{C}ollaku , Kuranage Roche Rayan Ranasinghe , Giuseppe Thadeu Freitas de Abreu , Takumi Takahashi This is my paper
classification eess.SP
keywords integrated communication and computingover-the-air computingbilinear Gaussian belief propagationchannel trackingchannel predictionmmWave channelsjoint channel and data detectiontime-varying channels
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Integrated communication and computing usually assumes the base station knows the channel perfectly at every instant. This paper tries to remove that assumption for millimeter-wave links by tracking the time-varying channel and decoding data in parallel while computing a target function over the air. The proposed receiver builds on bilinear Gaussian belief propagation for joint channel and data estimation, adds a channel-prediction step before each estimation window, and then applies an MMSE combiner to the residual signal to recover the desired over-the-air computation. Simulation results at 60 GHz with relative velocities of 10–40 km/h show bit error rate, channel-estimation NMSE, and computing NMSE close to genie-aided perfect-CSI bounds, which would make AirComp feasible in high-mobility mmWave scenarios such as vehicle-to-everything links.

Core claim

The paper claims that a BiGaBP-based joint channel and data estimation algorithm, seeded by a Kalman-like channel predictor, can keep up with a time-varying mmWave channel while simultaneously detecting QPSK communication symbols and supporting an over-the-air computation. The detection treats the superimposed computing signal as effective noise; the channel estimator propagates beliefs over time windows and combines them across antennas; the AirComp stage then applies an MMSE combiner to the residual after subtracting the detected communication contribution. With only the initial channel and the channel correlation coefficient r given, the scheme is shown to reach BER, channel NMSE, and com

What carries the argument

The load-bearing mechanism is bilinear Gaussian belief propagation (BiGaBP): message passing on a tripartite graph in which channel coefficients and data symbols are treated as Gaussian unknowns, enabling joint channel and data detection (JCDE). A channel-prediction (CP) step computes the conditional expectation of the channel given the lowest-MSE prior estimate in the current window, seeding the message passing with better starting points. The AirComp operation is carried by an MMSE combiner applied to the residual signal after the detected communication symbols are subtracted, with the combiner built from the estimated channel, the data-estimation error covariance, and the noise power.

Load-bearing premise

The tracking loop assumes the receiver knows the channel correlation coefficient r between adjacent OFDM symbols (and the exact channel at time zero), and the promised estimation procedure for r is never provided.

What would settle it

Run the receiver with a deliberately mismatched correlation coefficient—say, the r corresponding to 40 km/h while the true channel evolves at 10 km/h—and record BER, channel NMSE, and computing NMSE. If the gap to the genie-aided bounds becomes large, the near-genie results depend on privileged knowledge of mobility rather than on the tracking itself.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • A 60 GHz uplink with two single-antenna users moving at up to 40 km/h can sustain BER within a small margin of the perfect-CSI bound while devoting only 1% of transmit power to computing symbols.
  • The computing NMSE of the estimated sum function also tracks the genie-aided bound, meaning AirComp does not require known data symbols or known channel at run time.
  • Channel tracking and pilot-free operation become possible after an initial channel estimate, because each window is seeded by prediction rather than new pilots.
  • The same framework supports integrated communication and computing in high-mobility settings such as vehicle-to-everything links, and can be extended to multi-stream or general nomographic computations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A natural test is to replace the assumed correlation coefficient r with an online estimate; the paper promises such a procedure but never gives it, so the sensitivity of the near-genie results to r is the main open question.
  • Because r is a hyperparameter of the AR(1) channel model, the same belief-propagation machinery could be extended to infer r jointly with the channel, which would remove the strongest remaining assumption.
  • The AirComp stage currently runs after the JCDE loop; feeding the computing estimate back into detection could improve both tasks, since the detection already treats computing signals as noise.
  • The beamformer is fixed from the initial SVD; updating it from tracked channel estimates would likely push the operating range beyond 40 km/h.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes an integrated communication and computing (ICC) receiver for a time-varying millimeter-wave SIMO uplink. The receiver combines a bilinear Gaussian belief propagation (BiGaBP) algorithm for joint channel estimation/tracking and data detection with a channel prediction (CP) step, and then computes an over-the-air arithmetic-sum function from the residual signal via an MMSE combiner. The channel is modeled as a clustered mmWave channel with an AR(1) time dependence controlled by a correlation coefficient r; the BS is assumed to know the initial channel H[0] and r. Simulations at 60 GHz for 2 users over 128 time slots report BER, channel NMSE, and AirComp NMSE for velocities 10–40 km/h, claiming near-genie performance.

Significance. If the receiver is implementable as described, the paper would be a useful demonstration that over-the-air computing can coexist with channel tracking in time-varying mmWave channels without perfect CSI at every time instant. The main strength is the end-to-end simulation of a nontrivial setup: clustered mmWave channels, channel prediction, joint data/channel inference, and residual-based AirComp. However, the algorithmic core is largely carried over from the authors' prior work ([19], [20], [21]); the present contribution is primarily an integration and a simulation study. No code or reproducibility material is provided, and the text contains a load-bearing ambiguity in the channel update that must be resolved before the simulation claims can be accepted.

major comments (4)
  1. [Section III-C, Eq. (34a)] As written, the channel soft-replica update uses the true channel h_m[k] at the same time index being estimated: h'_m,k = Omega_m,k Lambda^{-1}_m,k hbar_m,k + r^k Psi^h_k,m Lambda^{-1}_m,k h_m[k]. Throughout Section III, h_m[k] denotes the true channel, whereas the soft replica is hhat_m,k, and Algorithm 1 does not list h_m[k] as an input. Taken literally, this makes the channel estimator non-causal and genie-aided, which would invalidate the central 'no perfect CSI' claim and the Figure 2 JCCCT curves. If this is a typo (e.g., h_m[0] or a past estimate was intended), it must be corrected and the notation reconciled with Algorithm 1; otherwise, the feasibility of the tracking loop is not established.
  2. [Section II-A, after Eq. (4)] The text states 'The procedure for the estimation of r follows consecutively,' but no such procedure appears anywhere in the manuscript. The correlation r is an input to Algorithm 1 and enters the second-order statistics (15), the channel prediction (17)–(19), the message variances (30), and the channel denoiser (34). Since all tracking and detection performance depends on r, the absence of an estimator is a significant gap. The authors should either provide a concrete estimation procedure or clearly state that r is assumed known and discuss the sensitivity of the results to r mismatch.
  3. [Section III-D, Eq. (38)] The combiner is called 'MMSE' and 'optimal' in the abstract and in Eq. (37), but the closed form in Eq. (38) is derived under the assumption that the estimated channel Hhat[k] is exact, as the preceding sentence acknowledges ('computed without considering channel estimation error'). In a no-perfect-CSI setting, the true MMSE combiner should account for the channel estimation error covariance. The authors should either soften the optimality claim or quantify the impact of channel estimation error on the AirComp NMSE, especially because Figure 2(c) compares against genie-aided bounds.
  4. [Section III-B, Eqs. (20)–(22)] The computing signal term H[k]s[k] is folded into an effective white noise term with variance N0 + E_c. Since s[k] has covariance E_c I_M, the actual covariance of H[k]s[k] is E_c H[k]H[k]^H, which is not generally proportional to the identity after the quasi-SVD beamformer. The scalar approximation may be acceptable, but it should be justified, and its effect on the reported BER/NMSE should be discussed.
minor comments (5)
  1. [Section IV] The simulation paragraph states 'N_RX = 16 receive antennas, P = 2 receive antennas.' This is confusing: P is presumably the number of UPA elements in one dimension, not a second count of receive antennas. Please clarify the relationship between N_RX, P, and N in Eq. (2) and the simulation setup.
  2. [Section IV] The sentence 'The transmit power was set to E_d = 0.99 for communications and E_c = 0.01 for communications' contains a typo: E_c should be 'for computing.'
  3. [Algorithm 1, line 26] The data-symbol term is written as dhat*_{nm,s}; the symbol estimate should not depend on the receive-antenna index n. This is likely a typesetting issue, but it should be fixed to dhat*_{m,s} or similar.
  4. [Section II-A, Eq. (4)] The AR model parameter r is defined as the correlation between adjacent OFDM symbols, but Eq. (7) derives it from coherence time via exp(ln(0.5)/K_max). The connection between the two definitions and the specific OFDM parameters should be made explicit, since K_max depends on Ts and fc.
  5. [General] The simulation results have no error bars, number of Monte Carlo runs, or complexity measurements. Adding these would strengthen the reproducibility of the claims.

Circularity Check

1 steps flagged

Channel-tracking update (34a) feeds the true channel h_m[k] into the estimator, so the claimed no-perfect-CSI JCCCT results are obtained with a genie input; the missing r-estimation procedure is a separate unfulfilled promise.

specific steps
  1. self definitional [Section III-C, 'Denoising and damping', Eq. (34a); Algorithm 1, line 30]
    "Then, the new replicas can be generated as h'_m,k = Omega_m,k Lambda^-1_m,k hbar_m,k + r^k Psi^h_k,m Lambda^-1_m,k h_m[k], (34a)"

    The channel soft replica h'_m,k at time k is updated using h_m[k], the true channel at time k, not an estimate or a past value. The paper's notation fixes h_m[k] as the actual effective channel; Algorithm 1 takes only y[k], H[0], r, N~0, and design parameters as inputs; and the abstract's premise is that no perfect CSI is needed. Feeding the ground-truth channel into the channel-estimation recursion means the 'tracked' channel is defined in terms of the unknown it claims to produce. Consequently the BER, channel-NMSE, and AirComp-NMSE curves in Figure 2 are not predictions from observations; they are constructed with the genie value. If h_m[k] was intended to be a prior estimate, the text does not say so and no code is provided to disambiguate; as written, this load-bearing step is circula

full rationale

The central circularity is Eq. (34a): the channel estimate h'_m,k is an affine function of the true channel h_m[k], so the no-perfect-CSI tracking claim reduces to having the truth as an input. This directly undermines the headline result that JCCCT approaches genie-aided bounds: the receiver is genie-aided at that line. Separately, Section II-A promises 'The procedure for the estimation of r follows consecutively' but never delivers it; this is an unfulfilled completeness/robustness assumption rather than a circular step, since r could in principle be obtained externally. The paper also relies heavily on the authors' prior BiGaBP and AirComp papers ([19], [20], [10], [21]); that self-citation is substantial but would not by itself be circular because the algorithms are cited as building blocks and the integration is evaluated by simulation. However, the literal genie term in (34a) makes the central claim forced by definition, outweighing those considerations and giving a score of 8.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The framework combines algorithms from prior papers [20], [10], [21]; the only new contribution is the integration. The central inputs assumed rather than derived are the channel correlation r, the initial channel H[0], and the algorithm design parameters (W, D, G, t_max, beta). No invented physical entities are introduced.

free parameters (4)
  • channel correlation r = known in simulation, computed from velocity via (5)-(7); no estimation method given
    Input to Algorithm 1; governs channel prediction (19) and statistics (15). In practice r is unknown and must be estimated, but no estimator is provided.
  • window size W, update D, neighborhood G = (8, 3, 6)
    Algorithm design parameters that trade tracking capability vs. complexity; chosen without sensitivity analysis.
  • damping coefficient beta = 0.5
    Chosen to ease convergence; no criterion given.
  • power split E_d, E_c = 0.99, 0.01
    Chosen to prioritize communications; computing signal treated as noise.
axioms (5)
  • domain assumption AR(1) model for small-scale fading with known correlation r (eq. 4)
    The channel time variation is modeled as first-order autoregressive with parameter r; performance depends on this model.
  • domain assumption Perfect initial channel H[0] known at time zero
    Used to construct the quasi-SVD beamformer (12) and to initialize channel prediction (17).
  • domain assumption Perfect synchronization among users
    Stated in Section II-B; unsynchronized transmissions would break the superposition model (8).
  • domain assumption BiGaBP and CP algorithms from [20] are correct and applicable
    The paper builds entirely on [20] without re-derivation; the message-passing approximations (VGA, SGA) are borrowed.
  • domain assumption Computing symbols are zero-mean Gaussian with variance E_c
    Enables treating them as effective noise; real computing symbols may not be Gaussian.

pith-pipeline@v1.3.0-alltime-deepseek · 11075 in / 11508 out tokens · 114030 ms · 2026-08-04T22:06:58.111353+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Integrated Communication and Computing in Time-Varying mmWave Channels." pith.science (2026). https://pith.science/paper/VIQWXGCJ

@misc{pith2026250907482,
  author       = {Pith},
  title        = {Pith review of: Integrated Communication and Computing in Time-Varying mmWave Channels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VIQWXGCJ}},
  note         = {Machine review of arXiv:2509.07482}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We propose a novel framework for integrated communication and computing (ICC) transceiver design in time-varying millimeter-wave (mmWave) channels. In particular, in order to cope with the dynamics of time-varying mmWave channels, the detection of communication symbols and the execution of an over-the-air computing (AirComp) operation are performed in parallel with channel tracking, as opposed to existing state-of-the-art (SotA) on ICC where perfect knowledge of the channel at all time instances is typically assumed. For clarity of exposition, we consider a single-input multiple-output (SIMO) uplink scenario where multiple single-antenna user equipment (UE) transmit to a base station (BS) equipped with multiple antennas, such that each UE, or edge device (ED), precodes its own transmit signal, while the BS, or access points (APs), also performs receive beamforming. The proposed transceiver framework then estimates channel state information (CSI) and data symbols in parallel, using a bilinear Gaussian belief propagation (BiGaBP) algorithm for joint channel and data detection (JCDE), aided by a channel prediction (CP) algorithm executed before each estimation window at the BS. The AirComp operation is then executed by means of an optimal combination of the residual signal. Simulation results demonstrate the effectiveness of the proposed scheme in performing ICC in challenging time-varying mmWave channels, with minimal degradation to both communication and computing performance.

Figures

Figures reproduced from arXiv: 2509.07482 by Giuseppe Thadeu Freitas de Abreu, Joan \c{C}ollaku, Kuranage Roche Rayan Ranasinghe, Takumi Takahashi.

Figure 1
Figure 1. Figure 1: Uplink mmWave SIMO ICC system. h´m[k] = X L l=1 X Cl c=1 σl,c,m[k] √ LCl aNRX (θ RX l,c,m, ϕRX l,c,m), (1) where aNRX (θ RX l,c,m, ϕRX l,c,m) is the array response of the BS/AP receive antennas to the m-th UEs/EDs signal, θRX represents the elevation angle of arrival (AoA), ϕRX represents the azimuth AoA and σl,c,m represents the time-varying small￾scale fading coefficient. Assuming that the BS/AP is equip… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages · 2 internal anchors

  1. [1]

    Integrated sensing and communication with mmwave massive mimo: A compressed sampling perspective,

    Z. Gao, Z. Wan, D. Zheng, S. Tan, C. Masouros, D. W. K. Ng, and S. Chen, “Integrated sensing and communication with mmwave massive mimo: A compressed sampling perspective,”IEEE Transactions on Wireless Communications, vol. 22, no. 3, pp. 1745–1762, 2023

    work page 2023

  2. [2]

    Millimeter-wave vehicular communication to support massive automotive sensing,

    J. Choi, V . Va, N. Gonzalez-Prelcic, R. Daniels, C. R. Bhat, and R. W. Heath, “Millimeter-wave vehicular communication to support massive automotive sensing,”IEEE Communications Magazine, vol. 54, no. 12, pp. 160–167, 2016

    work page 2016

  3. [3]

    Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,

    F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, and S. Buzzi, “Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,”IEEE Journal on Selected Areas in Communications, vol. 40, no. 6, pp. 1728–1767, 2022

    2022

  4. [4]

    Joint channel, data, and radar parameter estimation for afdm systems in doubly-dispersive channels,

    K. R. R. Ranasinghe, H. Seok Rou, G. Thadeu Freitas de Abreu, T. Taka- hashi, and K. Ito, “Joint channel, data, and radar parameter estimation for afdm systems in doubly-dispersive channels,”IEEE Transactions on Wireless Communications, vol. 24, no. 2, pp. 1602–1619, 2025

    work page 2025

  5. [5]

    Computation over multiple-access channels,

    B. Nazer and M. Gastpar, “Computation over multiple-access channels,” IEEE Transactions on Information Theory, vol. 53, no. 10, pp. 3498– 3516, 2007

    work page 2007

  6. [6]

    Low-complexity and high-performance combiners for over the air computing,

    K. Ando and G. T. F. de Abreu, “Low-complexity and high-performance combiners for over the air computing,” in2023 IEEE 9th International Workshop on Computational Advances in Multi-Sensor Adaptive Pro- cessing (CAMSAP), 2023, pp. 126–130

    work page 2023

  7. [7]

    Optimal receive beam- forming for over-the-air computation,

    W. Fang, Y . Zou, H. Zhu, Y . Shi, and Y . Zhou, “Optimal receive beam- forming for over-the-air computation,” in2021 IEEE 22nd International Workshop on Signal Processing Advances in Wireless Communications (SPA WC), 2021, pp. 61–65

    work page 2021

  8. [8]

    Over-the-air computation via broadband channels,

    T. Qin, W. Liu, B. Vucetic, and Y . Li, “Over-the-air computation via broadband channels,”IEEE Wireless Communications Letters, vol. 10, no. 10, pp. 2150–2154, 2021

    work page 2021

  9. [9]

    Over-the-air computation systems: Optimization, analysis and scaling laws,

    W. Liu, X. Zang, Y . Li, and B. Vucetic, “Over-the-air computation systems: Optimization, analysis and scaling laws,”IEEE Transactions on Wireless Communications, vol. 19, no. 8, pp. 5488–5502, 2020

    work page 2020

  10. [10]

    From theory to reality: A design framework for integrated communication and computing receivers,

    K. R. Rayan Ranasinghe, K. Ando, and G. T. Freitas de Abreu, “From theory to reality: A design framework for integrated communication and computing receivers,” in2025 International Conference on Computing, Networking and Communications (ICNC), 2025, pp. 865–870

    work page 2025

  11. [11]

    Millimeter-wave cellular wireless networks: Potentials and challenges,

    S. Rangan, T. S. Rappaport, and E. Erkip, “Millimeter-wave cellular wireless networks: Potentials and challenges,”Proceedings of the IEEE, vol. 102, no. 3, pp. 366–385, 2014

    2014

  12. [12]

    A survey of 5g channel measurements and models,

    C.-X. Wang, J. Bian, J. Sun, W. Zhang, and M. Zhang, “A survey of 5g channel measurements and models,”IEEE Communications Surveys & Tutorials, vol. 20, no. 4, pp. 3142–3168, 2018

    work page 2018

  13. [13]

    Joint spatial division and multiplexing—the large-scale array regime,

    A. Adhikary, J. Nam, J.-Y . Ahn, and G. Caire, “Joint spatial division and multiplexing—the large-scale array regime,”IEEE Transactions on Information Theory, vol. 59, no. 10, pp. 6441–6463, 2013

    work page 2013

  14. [14]

    Channel estimation and hybrid precoding for millimeter wave cellular systems,

    A. Alkhateeb, O. El Ayach, G. Leus, and R. W. Heath, “Channel estimation and hybrid precoding for millimeter wave cellular systems,” IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 5, pp. 831–846, 2014

    2014

  15. [15]

    Novel 3-d nonstationary mmwave massive mimo channel models for 5g high- speed train wireless communications,

    Y . Liu, C.-X. Wang, J. Huang, J. Sun, and W. Zhang, “Novel 3-d nonstationary mmwave massive mimo channel models for 5g high- speed train wireless communications,”IEEE Transactions on V ehicular Technology, vol. 68, no. 3, pp. 2077–2086, 2019

    work page 2077

  16. [16]

    Bilinear generalized approxi- mate message passing—part i: Derivation,

    J. T. Parker, P. Schniter, and V . Cevher, “Bilinear generalized approxi- mate message passing—part i: Derivation,”IEEE Transactions on Signal Processing, vol. 62, no. 22, pp. 5839–5853, 2014

    work page 2014

  17. [17]

    Gaussian Belief Propagation for Solving Systems of Linear Equations: Theory and Application

    O. Shental, D. Bickson, P. H. Siegel, J. K. Wolf, and D. Dolev, “Gaussian belief propagation for solving systems of linear equations: Theory and application,” 2008. [Online]. Available: https://arxiv.org/abs/0810.1119

    work page internal anchor Pith review Pith/arXiv arXiv 2008

  18. [18]

    Distributed inference with variational message passing in gaussian graphical models: Tradeoffs in message schedules and convergence conditions,

    B. Li, N. Wu, and Y .-C. Wu, “Distributed inference with variational message passing in gaussian graphical models: Tradeoffs in message schedules and convergence conditions,”IEEE Transactions on Signal Processing, vol. 72, pp. 2021–2035, 2024

    work page 2021

  19. [19]

    Design of adaptively scaled belief in multi-dimensional signal detection for higher-order modulation,

    T. Takahashi, S. Ibi, and S. Sampei, “Design of adaptively scaled belief in multi-dimensional signal detection for higher-order modulation,” IEEE Transactions on Communications, vol. 67, no. 3, pp. 1986–2001, 2019

    work page 1986

  20. [20]

    Bayesian bilinear inference for joint channel tracking and data detection in millimeter-wave mimo systems,

    T. Takahashi, H. Iimori, K. Ishibashi, S. Ibi, and G. T. F. de Abreu, “Bayesian bilinear inference for joint channel tracking and data detection in millimeter-wave mimo systems,”IEEE Transactions on Wireless Communications, vol. 23, no. 9, pp. 11 136–11 153, 2024

    work page 2024

  21. [21]

    A Flexible Design Framework for Integrated Communication and Computing Receivers

    K. R. R. Ranasinghe, K. Ando, H. S. Rou, G. T. F. de Abreu, T. Takahashi, M. Di Renzoet al., “A flexible design framework for integrated communication and computing receivers,”arXiv preprint arXiv:2506.05944, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  22. [22]

    Millimeter-wave beamformed full- dimensional mimo channel estimation based on atomic norm minimiza- tion,

    Y . Tsai, L. Zheng, and X. Wang, “Millimeter-wave beamformed full- dimensional mimo channel estimation based on atomic norm minimiza- tion,”IEEE Transactions on Communications, vol. 66, no. 12, pp. 6150– 6163, 2018

    work page 2018

  23. [23]

    Iterative eigenvalue decomposi- tion and multipath-grouping tx/rx joint beamformings for millimeter- wave communications,

    Z. Xiao, X.-G. Xia, D. Jin, and N. Ge, “Iterative eigenvalue decomposi- tion and multipath-grouping tx/rx joint beamformings for millimeter- wave communications,”IEEE Transactions on Wireless Communica- tions, vol. 14, no. 3, pp. 1595–1607, 2015

    work page 2015

  24. [24]

    An overview of signal processing techniques for millimeter wave mimo systems,

    R. W. Heath, N. Gonz ´alez-Prelcic, S. Rangan, W. Roh, and A. M. Say- eed, “An overview of signal processing techniques for millimeter wave mimo systems,”IEEE Journal of Selected Topics in Signal Processing, vol. 10, no. 3, pp. 436–453, 2016

    work page 2016

  25. [25]

    Frame theory and fractional programming for sparse recovery-based mmwave channel estimation,

    R.-A. Stoica, H. Iimori, G. T. Freitas de Abreu, and K. Ishibashi, “Frame theory and fractional programming for sparse recovery-based mmwave channel estimation,”IEEE Access, vol. 7, pp. 150 757–150 774, 2019

    work page 2019

  26. [26]

    Ieee 802.11ad: introduction and performance evaluation of the first multi-gbps wifi technology,

    C. Cordeiro, D. Akhmetov, and M. Park, “Ieee 802.11ad: introduction and performance evaluation of the first multi-gbps wifi technology,” inProceedings of the 2010 ACM International Workshop on MmWave Communications: From Circuits to Networks, ser. mmCom ’10, 2010, p. 3–8

    work page 2010