Two-Stage Distributed Beamforming Design in Cell-Free Massive MIMO ISAC Systems

Ad\~ao Silva; At\'ilio Gameiro; Daniel Castanheira; Leonardo Leyva

REVIEW 3 major objections 4 minor 28 references

Two-Stage Distributed Beamforming Design in Cell-Free Massive MIMO ISAC Systems

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A two-stage split of beamforming computation cuts cell-free ISAC fronthaul load while keeping performance close to centralized design.

desk verdict A genuinely useful distributed beamforming architecture for CF-ISAC, but the central 'comparable performance' claim leans on an unspecified centralized baseline that needs to be pinned down before the paper is fully verifiable. read the letter →

arxiv 2501.10136 v1 pith:DSBJ5HKP submitted 2025-01-17 eess.SP

Leonardo Leyva , Daniel Castanheira , Adão Silva , Atílio Gameiro This is my paper · ORCID

classification eess.SP

keywords cell-freemassiveMIMOintegratedsensingandcommunicationdistributedbeamformingfronthaulloadmajorization-minimizationnull-spaceprojectionsumSINRoptimizationISAC

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 two-stage distributed beamforming design for cell-free massive MIMO systems that serve communication users and sense a target at the same time. The claim is that splitting each beamforming vector into a local access-point component and a central-unit scalar weight, then exchanging only compact equivalent channels between the two stages, yields performance close to a fully centralized solution while dramatically reducing the fronthaul load. Concretely, the beamforming-design data exchange becomes independent of the number of antennas per access point, removing a key scalability bottleneck for cell-free integrated sensing and communication (ISAC). A careful reader should take away that most of the benefit of joint optimization survives distribution, and that the practical cost is small: roughly three exchange iterations are enough to approach the converged result.

What carries the argument

The engine is the two-stage precoder decomposition combined with null-space interference cancellation and majorization-minimization. Null-space projection turns the multi-user SINR objective into a sum of projected desired-signal powers that each access point can compute from its local channel, eliminating multi-user interference without exchanging full channel matrices. The remaining non-convex problems, one at each access point and one at the central unit, are replaced by convex surrogates built from a first-order Taylor lower bound and a linearized sensing constraint, and the two stages are iterated alternately. The compact scalars exchanged per iteration, namely the equivalent channels $z_{m,k}$ and $g_{m,k}$, the precoder powers $w_{m,k}$, and the central weights $\delta_{m,k}$ (with derived coefficients $\alpha_{m,k}$ and $\beta_{m,k}$), are what make the fronthaul count independent of $N_{\mathrm{tx}}$.

What would settle it

Pick a target angle that coincides with or lies very close to one of the user channel angles, compute the projected steering-vector gains $\|\mathbf{P}_{m,k}^H\mathbf{a}_{N_{\mathrm{tx}}}(\theta_m)\|^2$, and check whether any feasible precoder can meet the sensing threshold; if these gains are near zero, the optimization problem becomes infeasible and the claimed sensing performance cannot be delivered.

Watch

Extended reading notes

Core claim

The central discovery is that the joint beamforming problem decomposes with little loss: every precoding vector is written as $\mathbf{f}_{m,k}=\delta_{m,k}\mathbf{w}_{m,k}$, where $\mathbf{w}_{m,k}$ is computed locally at the $m$-th access point and $\delta_{m,k}$ is a scalar weight computed at the central unit. Inter-user interference is cancelled locally by projecting each local precoder into the null space of the other users' channels, so the access points optimize in parallel with no inter-access-point channel sharing. The two sides exchange only equivalent communication channels, equivalent sensing channels, and precoder powers, giving a total of $6N_{\mathrm{iter}}MK$ complex scalars over $N_{\mathrm{iter}}$ iterations, independent of the antenna count $N_{\mathrm{tx}}$; the centralized baseline exchanges $2N_{\mathrm{tx}}MK$ scalars. In the reported scenario the sum SINR stays close to the centralized solution across the tested sensing thresholds, and three iterations already reach roughly 98--99% of the final value.

Load-bearing premise

The design stands on the assumption that after each access point projects its sensing direction into the null space of its users' interference channels, the remaining energy pointed at the sensing target is still enough to satisfy the sensing SNR requirement.

Editorial extensions

If this is right

Beamforming-design fronthaul becomes independent of the antenna count: the total exchange is $6N_{\mathrm{iter}}MK$ complex scalars versus $2N_{\mathrm{tx}}MK$ for the centralized approach.
Central-unit computational complexity drops from $O(N_{\mathrm{tx}}^3M^3K^3\sqrt{N_{\mathrm{tx}}MK})$ to $O(M^3K^3\sqrt{MK})$, because the central unit optimizes only $MK$ scalar weights.
A small number of exchange iterations is sufficient: about three iterations reach 99% of the converged sum SINR at $\Delta=30$ dB and about 98% at $\Delta=40$ dB.
Increasing the sensing requirement $\Delta$ shifts power toward the target and lowers the communication sum SINR, and the distributed design tracks the centralized trade-off curve closely.

Reading between the lines

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

Because the exchanged quantities depend only on $M$, $K$, and the iteration count, the same compact-exchange pattern should transfer to other distributed beamforming problems that summarize each access point through aggregate equivalent channels, not only cell-free ISAC.
The method's practical limit is likely the null-space projection: if a target direction nearly coincides with a user channel direction, the projected sensing gain may be too small to meet the sensing constraint, so a safeguard that detects low projected gain and reconfigures user-to-AP assignments would be a natural extension.
Extending to multiple targets would presumably require one sensing constraint per target and extra scalar sensing gains per iteration, trading some of the fronthaul saving for broader sensing coverage.

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

Kale, quick take on arXiv:2501.10136. The genuinely new thing is the two-stage split: each precoder is factored into a local AP vector and a central complex weight, and the APs exchange only equivalent channels (3MK scalars per direction) rather than full CSI. That makes fronthaul independent of array size, which is a real structural advantage over the centralized CF-ISAC designs in [21]–[25]. The MM-based alternating solution is competent, the null-space MUI cancellation is a sensible decoupling trick, and the local/central subproblems are convex after the surrogates. Proposition 1 gives a valid (if terse) containment argument for the sensing constraint relaxation. The convergence plots show the algorithm settling in about three iterations.

The soft spot is exactly where the stress-test lands: the 'comparable performance' claim in Fig. 6 rests on a centralized baseline that is never specified. Footnote 1 says the nonconvex centralized problem is handled with MM but 'detailed insights into the centralized optimization problem's solution are not provided.' Since (18) is nonconvex, the gap in Fig. 6 is a comparison to some undisclosed local solution, not to the optimum. That is load-bearing for the abstract's central claim. The missing communication noise variance sigma_k^2 is a smaller but real reproducibility gap — absolute SINR values cannot be re-derived. And there are no error bars on the mean curves, so the 10–15% gap shown could be within noise.

The sensing-feasibility concern is secondary but worth a sentence in revision. With K=2 and Ntx=32, the null-space projection usually leaves enough gain, but the paper should either prove a bound on the projected steering gain or report the distribution of target angles used in the averaged curves. The beampattern plots use a single channel realization and do not address adversarial alignments.

Who is this for? A group working on scalable CF-ISAC beamforming. The architecture idea deserves a serious referee. Send it to peer review, but the revisions must include a spelled-out centralized baseline (initialization, surrogate, iterations) or an appendix with the MM details, plus the missing noise variance and error bars. Without that, Fig. 6 is unverifiable.

Referee Report

3 major / 4 minor

Summary. The paper proposes a two-stage distributed beamforming design for cell-free massive MIMO integrated sensing and communication (ISAC) systems. The precoder at each access point is decomposed into a local precoder computed at the AP and a complex weight computed at the central unit, enabling a division of processing that reduces fronthaul exchange. Inter-user interference is explicitly zero-forced through null-space projection at each AP, and the resulting nonconvex optimization problems are solved via majorization-minimization, with convex surrogates solved by standard tools. Numerical results show convergence within a few iterations, a sensing-communication trade-off, and a fronthaul load independent of the number of transmit antennas, while claiming communication performance close to a centralized MM-based solution.

Significance. If the claims are correct, the paper offers a sensible scalability-oriented design for cell-free ISAC, with two concrete structurally valid advantages: the per-iteration fronthaul exchange scales as O(M K) rather than O(N_tx M K), and the CU complexity scales independently of N_tx. The MM surrogates are theoretically grounded by a proved subset relaxation (Proposition 1), and the algorithmic description is sufficiently detailed to be implemented. The main open questions concern the reproducibility and fairness of the performance comparison against an unspecified centralized baseline, and the feasibility of the sensing constraint after null-space projection in adversarial channel-target alignments.

major comments (3)

[Section IV-F, Fig. 6] The central claim of 'comparable performance to centralized methods' is supported only by an unspecified baseline. Footnote 1 states that the centralized problem (18) is handled with the MM technique, but does not specify the surrogate function, the initialization, the number of iterations, or the convergence criteria. Since (18) is nonconvex, the quality of that baseline depends crucially on these choices, and the small gap reported in Fig. 6 could be an artifact of a poor local optimum. The paper needs to either rigorously specify the centralized algorithm (e.g., in an appendix) or soften the claim to 'close to the centralized MM benchmark' and provide the associated parameters.
[Section V] The communication noise variance σ_k^2 is never specified in the numerical setup, although it appears directly in the SINR expression (9) and therefore determines the absolute values of the sum SINR in Figs. 4 and 6. Without this parameter, the reported SINR values cannot be reproduced by an independent implementation. Please state the value used (e.g., σ_k^2 = 1, 0 dB, or per-user noise power) and specify whether the same value is used for the centralized and distributed solutions.
[Section IV-B, Figs. 5-6] The feasibility of the sensing constraint after null-space projection is not analyzed. When the target angle θ_m is aligned with the subspace spanned by the interference channels used to build P_{m,k}, the projected steering vector a_{m,k} = P_{m,k}^H a_{Ntx}(θ_m) can have a very small norm, making the sensing SNR constraint (19c) or (23c) infeasible for the reported Δ values. The paper does not discuss how infeasible random channel realizations are handled in the Monte Carlo averages, nor does it report the distribution of the projected steering-vector gains for the chosen θ_m. Please address this by either providing a feasibility analysis, reporting the number of discarded realizations, or selecting experimental angles that avoid adversarial alignments.

minor comments (4)

[Equations (23c), (24c)] The steering vector is written as a_k^H(θ_m), but the projected steering vector was defined as a_{m,k} in (21). Use a_{m,k} consistently to avoid confusion with the array response a_{Ntx}(θ_m).
[Algorithm 1] Typographical errors: 'distibuted' in the Algorithm 1 title and 'Distributted' in its caption should be corrected to 'distributed'.
[Algorithm 1, step 4] The initialization g_{m,k} = (1/M) sqrt(Δ̃/K) seems inconsistent with its definition g_{m,k} = a_{m,k}^H w_{m,k} in Section IV-D, which is a channel-dependent quantity. Please clarify the rationale or the intended meaning of this initialization.
[Section V] The paper does not specify the number of Monte Carlo realizations used for the mean curves in Figs. 4 and 6 beyond '100 random realizations' for Fig. 4; please state the number used for Fig. 6 as well.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is self-contained, and the only self-citation is re-proved in the paper.

full rationale

No circular step is present. The derivation chain is self-contained: the system model defines SINR in (9) and sSNR in (14) independently of the proposed algorithm; the factorization F_m = W_m Δ_m in (15) is a mathematical reparameterization, not an assumption of the conclusion. The null-space projection in (21)-(22) is a construction that cancels MUI, and the resulting optimization problems (23) and (26) are solved with MM surrogates whose feasibility guarantee is Proposition 1, proved in the paper itself; the citation to [13] for this relaxation is not load-bearing because the same proposition and proof are given in Section IV-D. The fronthaul comparison in Section IV-F is a direct algebraic count of exchanged scalars (6NiterMK versus 2NtxMK), not a fitted or predicted quantity. The centralized baseline is admittedly not detailed in footnote 1, which is a reproducibility limitation and a possible fairness concern for the comparison, but it is not a case where an output is equivalent to an input by construction: 'comparable performance' is an empirical claim against a self-constructed baseline, not a quantity forced by definition or by fitting. Therefore the paper does not exhibit self-definitional, fitted-input, or self-citation load-bearing circularity.

Assumptions & free parameters 6 free parameters · 8 assumptions · 0 invented entities

The central claim rests on standard domain assumptions (perfect local CSI, single-path LoS sensing, Swerling I independence) and on an algorithmic premise: the alternating AP-CU iterations converge to a near-central stationary point within a few steps. Simulation parameters are hand-chosen and not all reported (notably sigma_k^2). No new physical entities are introduced; the two-stage precoding structure is a processing architecture, not a new law or entity.

free parameters (6)

Communication noise variance sigma_k^2
Required to define SINR in (9) and to give meaning to the absolute sum SINR values in Figs. 4 and 6; never specified in Section V.
Number of channel paths L = 10
Set in Section V for the geometric channel model (4); the trade-off curves may shift with L, and no sensitivity analysis is provided.
Sensing channel variance sigma_m,n^2 = -10 dB
Set in Section V; this value directly controls the left-hand side of the sensing constraint (19c) and hence the feasibility of the sSNR thresholds.
Sensing receiver noise variance sigma_n^2 = -20 dB
Set in Section V for the APrx noise in (14); it scales the denominator of the sSNR.
Per-AP power budget P_m = 1
Set in Section V; the communication and sensing trade-off is power-normalized to this value.
Number of algorithm iterations Niter = 3 (and 10 in Fig. 4)
Chosen in Section V as a convergence-accuracy heuristic; the claim of performance near the centralized solution depends on this choice.

assumptions (8)

domain assumption Perfect CSI is available locally at each APtx.
Stated in Section II-A. All derivations of null-space projection and SINR assume exact channel knowledge; imperfect CSI would break MUI cancellation and bias the comparison.
domain assumption Sensing channel gains alpha_m,n follow the Swerling I model and are independent across AP pairs.
Used in Section II-D to derive the additive sensing power in (13); if the target were specular, the sensing SNR expression would differ and cohere differently.
domain assumption Sensing channel between each APtx and APrx is a single LoS path via the target.
Assumed in Section II-D and used to write the sensing channel as alpha_m,n a_Nrx(phi_n) a^H_Ntx(theta_m); multi-path or direct leakage would change the model.
domain assumption Each APtx can compute the null-space projection of its local channels without coordination.
The interference cancellation in (21)-(22) requires only local channel knowledge, a key enabler of the distributed scheme; if channels were not locally separable, the approach fails.
standard math The MM surrogate (24c) provides a feasible subset of the original sensing constraint, per Proposition 1.
Proposition 1 proves the subset relation using the convexity of the norm; this is a valid sufficient condition, not an unproved assumption.
ad hoc to paper The alternating updates between APs and CU converge to a good stationary point within Niter iterations.
Section IV-C and Algorithm 1 use a fixed iteration count Niter and random initialization; no convergence theorem is given, only the numerical evidence of Fig. 4.
domain assumption The interference channel matrix H_m,k has full row rank so that a non-trivial null space exists.
The null-space projection in (22) requires enough antennas relative to the number of users; the paper's simulation satisfies this, but it is not stated as a condition for the method's validity.
ad hoc to paper The centralized baseline used for comparison is solved to its MM stationary point and is a fair performance reference.
Footnote 1 states the centralized solution is not detailed; the fairness of the comparison cannot be independently assessed.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Two-Stage Distributed Beamforming Design in Cell-Free Massive MIMO ISAC Systems." pith.science (2026). https://pith.science/paper/DSBJ5HKP

@misc{pith2026250110136,
  author       = {Pith},
  title        = {Pith review of: Two-Stage Distributed Beamforming Design in Cell-Free Massive MIMO ISAC Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DSBJ5HKP}},
  note         = {Machine review of arXiv:2501.10136}
}

read the original abstract

Integrating radio-sensing functionalities into future cell-free (CF) wireless networks promises efficient resource utilization and facilitates the seamless roll-out of applications such as public safety and smart infrastructure. While the beamforming design problem for the CF integrated sensing and communication (ISAC) paradigm has been addressed in the literature, existing methods rely on centralized signal processing, leading to fronthaul load and scalability issues. This paper presents a two-stage beamforming design for the CF ISAC paradigm, aiming to significantly reduce the fronthaul load by distributing the signal processing tasks between the central unit (CU) and the access points (APs). The design optimizes the sum signal-to-interference-plus-noise ratio (SINR) for communication users, subject to per-AP power constraints and signal-to-noise ratio (SNR) requirements for radio-sensing purposes. The resulting optimization problems are non-convex and challenging to solve. To address this, we employ a majorization-minimization (MM) approach, which decomposes the problem into simpler convex subproblems. The results show that the two-stage beamforming design achieves performance comparable to centralized methods while substantially reducing the fronthaul load, thus minimizing data transmission requirements over the fronthaul network. This advancement highlights the potential of the proposed method to enhance the efficiency and scalability of cell-free MIMO ISAC systems.

Figures

Figures reproduced from arXiv: 2501.10136 by the authors.

**Figure 2.** Diagram of the generation of the transmitted signal i [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** High-level schematic of the data exchange required fo [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 5.** Figure 5: Transmit beampatterns obtained by the TsDBA. (a) [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 7.** Figure 7: Fronthaul load comparison for beamforming design [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

**Figure 6.** Figure 6: The trade-off between the sSNR constraint value [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

[21]

Demirhan and A

U. Demirhan and A. Alkhateeb, ”Cell-Free ISAC MIMO Syst ems: Joint Sensing and Communication Beamforming,” arXiv:2301 .11328, Feb. 2024. [Online]. Available: https://arxiv.org/abs/2 301.11328

work page 2024
[25]

Cao and Q

Y . Cao and Q. -Y . Y u, ”Design and Performance Analyses of V -OFDM Integrated Signal for Cell-Free Massive MIMO Joint Communi cation and Radar System,” IEEE Systems Journal, vol. 17, no. 4, pp. 5943-5954, Dec. 2023

work page 2023
[1]

B. Paul, A. R. Chiriyath and D. W. Bliss, ”Survey of RF comm unications and sensing convergence research,” IEEE Access , vol. 5, pp. 252-270, Dec. 2017

work page 2017
[2]

F. Liu, C. Masouros, A. P . Petropulu, H. Grifﬁths and L. Ha nzo, ”Joint radar and communication design: applications, state-of-t he-art, and the road ahead,” IEEE Trans. Commun. , vol. 68, no. 6, pp. 3834-3862, Jun. 2020

work page 2020
[3]

Leyva, D

L. Leyva, D. Castanheira, A. Silva, A. Gameiro and L. Hanz o, ”Cooper- ative Multiterminal Radar and Communication: A New Paradig m for 6G Mobile Networks,” IEEE V eh. Technol. Mag. vol. 16, no. 4, pp. 38-47, Dec. 2021

work page 2021
[4]

Y . Cui, F. Liu, X. Jing, and J. Mu, ”Integrating Sensing an d Communi- cations for Ubiquitous IoT: Applications, Trends, and Chal lenges,” IEEE Netw., vol. 35, no. 5, pp. 158-167, Oct. 2021

work page 2021
[5]

J. A. Zhang et al., ”Enabling Joint Communication and Rad ar Sensing in Mobile Networks—A Survey,” IEEE Commun. Surv. Tutor ., vol. 24, no. 1, pp. 306-345, Oct. 2021

work page 2021
[6]

Liu et al., ”Seventy Y ears of Radar and Communications : The road from separation to integration,” IEEE Signal Process

F. Liu et al., ”Seventy Y ears of Radar and Communications : The road from separation to integration,” IEEE Signal Process. Mag. , vol. 40, no. 5, pp. 106-121, Jul. 2023

work page 2023

Show all 28 references

[7]

[Online]

3GPP , ”Feasibility Study on Integrated Sensing and Comm unication, TR 22.837 (Release 19),” 2023. [Online]. Available: https: //portal.3gpp.org/desktopmodules/Speciﬁcations/Spec iﬁcationDetails. aspx?speciﬁcationId=4044

2023
[8]

[Online]

3GPP , ”Service requirements for Integrated Sensing and Communication, TS 22.137 (Release 19),” 2023. [Online]. Available: https://portal.3gpp.org/desktopmodules/Sp eciﬁcations/ SpeciﬁcationDetails.aspx?speciﬁcationId=4198

2023
[9]

Giordani, M

M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Z orzi, ”Toward 6G Networks: Use Cases and Technologies,” IEEE Commun. Mag., vol. 58, no. 3, pp. 55-61, Mar. 2020

2020
[10]

Liu et al

F. Liu et al. , ”Integrated Sensing and Communications: Toward Dual- Functional Wireless Networks for 6G and Beyond,” IEEE J. Sel. Areas Commun., vol. 40, no.6, pp.1728-1767, Jun. 2022

2022
[11]

F. Liu, C. Masouros, A. Li, H. Sun and L. Hanzo, ”MU-MIMO C ommu- nications With MIMO Radar: From Co-Existence to Joint Trans mission,” IEEE Trans. Wirel. Commun. , vol. 17, no. 4, pp. 2755-2770, Apr. 2018

2018
[12]

X. Liu, T. Huang, N. Shlezinger, Y . Liu, J. Zhou and Y . C. E ldar, ”Joint Transmit Beamforming for Multiuser MIMO Communicat ions and MIMO Radar,” IEEE Trans. Signal Process. , vol. 68, pp. 3929-3944, Jun. 2020

2020
[13]

Leyva, D

L. Leyva, D. Castanheira, A. Silva, and A. Gameiro, ”Hyb rid Beam- forming Design for Communication-Centric ISAC,” IEEE Sens. J. , vol. 24, no. 13, pp. 21179-21190, May. 2024

2024
[14]

Li and M

Y . Li and M. Jiang, ”Joint Transmit Beamforming and Rece ive Filters Design for Coordinated Two-Cell Interfering Dual-Functio nal Radar- Communication Networks,” IEEE Trans. V eh. Technol., vol. 71, no. 11, pp. 12362-12367, Nov. 2022

2022
[15]

L. Chen, X. Qin, Y . Chen and N. Zhao, ”Joint Waveform and C luster- ing Design for Coordinated Multi-Point DFRC Systems,” IEEE Trans. Commun., vol. 71, no. 3, pp. 1323-1335, Mar. 2023

2023
[16]

Cheng, Y

G. Cheng, Y . Fang, J. Xu and D. W. K. Ng, ”Optimal Coordina ted Transmit Beamforming for Networked Integrated Sensing and Commu- nications,” IEEE Trans. Wirel. Commun. , vol. 23, no. 8, pp. 8200-8214, Aug. 2024

2024
[17]

H. Q. Ngo, A. Ashikhmin, H. Y ang, E. G. Larsson, and T. L. M arzetta, ”Cell-Free Massive MIMO V ersus Small Cells,” IEEE Trans. Wirel. Commun., vol. 16, no. 3, pp. 1834-1850, Mar. 2017

2017
[18]

Behdad, ¨O

Z. Behdad, ¨O. T. Demir, K. W. Sung, E. Bj¨ ornson and C. Cavdar, ”Power Allocation for Joint Communication and Sensing in Cell-Fre e Massive MIMO,” GLOBECOM 2022 - 2022 IEEE Global Communications Con- ference, Rio de Janeiro, Brazil, Jan. 2023

2022
[19]

Behdad, ¨O

Z. Behdad, ¨O. T. Demir, K. W. Sung, E. Bj¨ ornson and C. Cavdar, ”Multi- Static Target Detection and Power Allocation for Integrate d Sensing and Communication in Cell-Free Massive MIMO,” IEEE Trans. Wirel. Commun., vol. 23, no. 9, pp. 11580-11596, Sept. 2024

2024
[20]

Cao and Q

Y . Cao and Q. -Y . Y u, ”Joint Resource Allocation for User -Centric Cell- Free Integrated Sensing and Communication Systems,” IEEE Commun. Lett., vol. 27, no. 9, pp. 2338-2342, Sept. 2023

2023
[22]

W. Mao, Y . Lu, J. Liu, B. Ai, Z. Zhong and Z. Ding, ”Beamfor ming Design in Cell-Free Massive MIMO Integrated Sensing and Com munica- tion Systems,” GLOBECOM 2023 - 2023 IEEE Global Communications Conference, Kuala Lumpur, Malaysia, 2023

2023
[23]

W. Mao, Y . Lu, C. -Y . Chi, B. Ai, Z. Zhong and Z. Ding, ”Communication-Sensing Region for Cell-Free Massive MIMO ISAC Systems,” IEEE Trans. Wirel. Commun. , early access, Apr. 2024

2024
[24]

S. Liu, R. Liu, M. Li, and Q. Liu, ”Cooperative Cell-Free ISAC Networks: Joint BS Mode Selection and Beamforming Design,” arXiv preprint: 2305.10800, 2024. [Online]. Avai lable: https://arxiv.org/abs/2305.10800

2024 arXiv
[26]

Ayach, S

O. Ayach, S. Rajagopal, S. Surra, Z. Piand and R. Heath, ” Spatially Sparse Precoding in millimeter wave MIMO systems,” IEEE Trans. Wireless Commun., V ol. 13, no. 3, p. 1499–1513, Mar. 2014

2014
[27]

M. A. Richards, J. Scheer, W. A. Holm, and W. L. Melvin, Pr inciples of modern radar. Citeseer, 2010, vol. 1

2010
[28]

CVX: Matlab software for discipli ned convex programming, version 2.1,

M. Grant and S. Boyd, “CVX: Matlab software for discipli ned convex programming, version 2.1,” http://cvxr.com/cvx, Mar. 201 4

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [21]

Demirhan and A

U. Demirhan and A. Alkhateeb, ”Cell-Free ISAC MIMO Syst ems: Joint Sensing and Communication Beamforming,” arXiv:2301 .11328, Feb. 2024. [Online]. Available: https://arxiv.org/abs/2 301.11328

work page 2024

[2] [25]

Cao and Q

Y . Cao and Q. -Y . Y u, ”Design and Performance Analyses of V -OFDM Integrated Signal for Cell-Free Massive MIMO Joint Communi cation and Radar System,” IEEE Systems Journal, vol. 17, no. 4, pp. 5943-5954, Dec. 2023

work page 2023

[3] [1]

B. Paul, A. R. Chiriyath and D. W. Bliss, ”Survey of RF comm unications and sensing convergence research,” IEEE Access , vol. 5, pp. 252-270, Dec. 2017

work page 2017

[4] [2]

F. Liu, C. Masouros, A. P . Petropulu, H. Grifﬁths and L. Ha nzo, ”Joint radar and communication design: applications, state-of-t he-art, and the road ahead,” IEEE Trans. Commun. , vol. 68, no. 6, pp. 3834-3862, Jun. 2020

work page 2020

[5] [3]

Leyva, D

L. Leyva, D. Castanheira, A. Silva, A. Gameiro and L. Hanz o, ”Cooper- ative Multiterminal Radar and Communication: A New Paradig m for 6G Mobile Networks,” IEEE V eh. Technol. Mag. vol. 16, no. 4, pp. 38-47, Dec. 2021

work page 2021

[6] [4]

Y . Cui, F. Liu, X. Jing, and J. Mu, ”Integrating Sensing an d Communi- cations for Ubiquitous IoT: Applications, Trends, and Chal lenges,” IEEE Netw., vol. 35, no. 5, pp. 158-167, Oct. 2021

work page 2021

[7] [5]

J. A. Zhang et al., ”Enabling Joint Communication and Rad ar Sensing in Mobile Networks—A Survey,” IEEE Commun. Surv. Tutor ., vol. 24, no. 1, pp. 306-345, Oct. 2021

work page 2021

[8] [6]

Liu et al., ”Seventy Y ears of Radar and Communications : The road from separation to integration,” IEEE Signal Process

F. Liu et al., ”Seventy Y ears of Radar and Communications : The road from separation to integration,” IEEE Signal Process. Mag. , vol. 40, no. 5, pp. 106-121, Jul. 2023

work page 2023