Mobile Cell-Free Massive MIMO with Multi-Agent Reinforcement Learning: A Scalable Framework

Bo Ai; Enyu Shi; Jiayi Zhang; Yiyang Zhu; Ziheng Liu

Mobile Cell-Free Massive MIMO with Multi-Agent Reinforcement Learning: A Scalable Framework

T0 review · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A scalable multi-agent reinforcement learning framework jointly optimizes UAV mobility and downlink power control in mobile cell-free massive MIMO, reporting sum spectral-efficiency gains of up to 45.97% over standard MADDPG.

desk verdict The action-decomposition idea is real, but the reported SE gains are confounded by an inner-loop reward search that the baselines don't get. read the letter →

arxiv 2412.02581 v1 pith:NABAIEOK submitted 2024-12-03 cs.IT eess.SPmath.IT

Ziheng Liu , Jiayi Zhang , Yiyang Zhu , Enyu Shi , Bo Ai This is my paper · ORCID

classification cs.ITeess.SPmath.IT

keywords cell-freemassiveMIMOmulti-agentreinforcementlearningUAVmobilitypowercontrolspectralefficiencygraphneuralnetworkpermutationinvariancecreditassignment

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

The paper argues that the bottleneck in mobile cell-free massive MIMO is not only the radio channel but the structure of the learning problem: with many mobile access points, conventional multi-agent reinforcement learning suffers from exponential state-action growth, weak collaboration, and unfair reward sharing. It proposes SF-MADDPG, a scalable MARL framework that jointly optimizes UAV mobility and downlink power control, and reports that combining a graph-neural-network communication layer, permutation-invariant and permutation-equivariant networks, and attention-based credit assignment yields large sum spectral-efficiency gains over standard MADDPG and other baselines. The central technical idea is to compress the observation space by separating order-insensitive actions from order-sensitive actions and to allocate rewards according to each agent's actual contribution. If correct, this means practical mobile cell-free deployments could be optimized online with distributed agents rather than centralized processing.

What carries the argument

The load-bearing object is the SF-MADDPG framework and its three nested mechanisms: a GNN-aided communication architecture that updates each agent's observation by aggregating hidden states from neighboring agents; a dynamic hyper permutation network that separates entity-uncorrelated actions (mobility) into a permutation-invariant branch and entity-correlated actions (AP power and antenna power) into a permutation-equivariant branch, with hypernetworks generating customized weight matrices so the candidate solution space is not constrained; and a directional decoupling architecture that combines an attention-based intrinsic reward network with a hypernetwork-based mixed reward network to distribute the global sum-SE reward according to each agent's contribution. The permutation design is backed by a minimum-modification principle: making the input module permutation-invariant makes the invariant output module automatically invariant, while the equivariant output module needs the original observation order as an additional input. The optimization target is the closed-form downlink sum spectral efficiency derived under MR precoding and channel hardening, where power decisions depend on large-scale fading coefficients rather than instantaneous small-scale fading.

What would settle it

A concrete falsifier would be to rerun the proposed framework with N=2 or N=4 antennas per mobile-AP while keeping the deployment otherwise identical, and to compare both large-scale-only policies and policies that also observe instantaneous CSI; if the reported sum-SE gains over MADDPG shrink or reverse, the channel-hardening assumption, rather than the architecture, is carrying the improvement.

Watch

Extended reading notes

Core claim

The paper establishes that a multi-agent reinforcement learning framework can jointly solve mobility and downlink power control for a cell-free massive MIMO system whose access points are mobile UAVs, and that the architecture's three components work together: a GNN-aided communication layer lets agents share neighbor information, a dynamic hyper permutation network compresses the state-action space by exploiting permutation invariance for mobility actions and permutation equivariance for power actions, and a directional decoupling architecture with an attention-based intrinsic reward network assigns credit to individual agents. The paper reports that this SF-MADDPG framework improves sum spectral efficiency by 45.97% over MADDPG, 25.29% over AIRMN, 18.79% over Co-MADDPG, and 14.41% over Pe-MADDPG under the joint architecture, while also converging about 43.5% faster than MADDPG. It further reports that the performance gap over permutation-only and collaboration-only baselines stays near 10% as the number of mobile access points grows from 9 to 64, and that the gap over some baselines widens as the number of UEs grows. The paper interprets these results as showing that mobility-enabled channel reconstruction, permutation-based observation compression, and contribution-based reward partitioning are jointly responsible for the gains.

Load-bearing premise

The whole optimization runs on the channel-hardening premise: the normalized instantaneous channel gain is taken to converge to its deterministic average, so power decisions can be made from large-scale fading coefficients alone, and the authors rely on adding antennas per mobile-AP to make outdated decisions harmless; in the simulations most mobile-APs use only N=8 antennas, which may be too few for that premise to hold.

Editorial extensions

If this is right

Sum spectral efficiency in mobile cell-free massive MIMO can be improved by roughly 46% over basic MADDPG, 25% over attention-based credit-assignment MARL, and 14% over permutation-based MARL when mobility, power control, GNN collaboration, and reward decomposition are jointly used.
Compressing order-insensitive actions through permutation invariance cuts the observation dimension of those actions from O((K+L)L! + (NK)K!) to O(L + (NK)K!), which is the stated reason the framework converges about 43.5% faster than MADDPG.
The joint mobility-and-power architecture yields higher sum SE than mobility-only control, with reported gains of 8.2% for SF-MADDPG, 15.31% for Pe-MADDPG, and 15.59% for AIRMN when power actions are added.
The performance advantage over permutation-only and collaboration-only baselines remains roughly constant as the number of mobile APs scales from 9 to 64, and it grows with the number of UEs, up to a 14.69% gap over one baseline, indicating that credit assignment matters more in UE-dense deployments.
The framework outperforms centralized multidimensional graph neural network baselines by 21.87% to 31.81%, which the paper attributes to the ability of mobile APs to reconstruct channels through movement rather than relying on global information at a fixed geometry.

Reading between the lines

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

The paper's separation of actions into entity-uncorrelated and entity-correlated groups is a general design rule for distributed resource allocation: the same permutation-invariant/equivariant decomposition could be applied to RIS phase control, uplink power control, or user scheduling in cell-free networks.
Because the reported gains over credit-assignment baselines widen with the number of UEs, a direct test is to vary the UE-to-AP ratio and inspect the intrinsic reward network's per-agent allocations; the framework's advantage should concentrate in regimes where contributions are most uneven.
The channel-hardening premise suggests a stress test: if the same framework is trained with N=2 or N=4 antennas per mobile-AP, where hardening is much weaker, the reported gains over MADDPG would either persist, showing the architecture is robust, or shrink, showing the large-scale-fading approximation is doing the work.
The comparison with centralized MDGNN implies that mobility is the main missing ingredient in fixed-geometry centralized learning; a natural extension would be to give MDGNN predicted mobility or position-aware inputs and see whether the centralized baseline closes the gap.

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.

Desk Editor's Note

read the letter

Quick take: the paper has a genuinely clean architecture-level idea, but the headline numbers don't support it as presented. Algorithm 1 is called at every decision step (Algorithm 2, line 5) and iteratively regenerates candidate actions until the sum-SE reward stops improving. That is an online search over the global objective, not a single-shot policy. At deployment it would need global SE feedback for every candidate action, which contradicts the decentralized MARL story. The baselines (MADDPG, Pe-MADDPG, Co-MADDPG, AIRMN) are not given the same search budget, so the 45.97%, 25.29%, 18.79%, and 14.41% gains do not isolate the proposed architecture. The stress-test note has this right; I checked the algorithm boxes and the loop is there.

What is actually new: the claim that mobility actions should be permutation-invariant while power actions should be permutation-equivariant is a clean inductive-bias choice for cell-free mMIMO, and the paper works it out in enough detail to be implementable. The GNN communication and intrinsic-reward credit assignment are borrowed from prior work, but combining them with the permutation split is a reasonable contribution. The ablations between single-PI and joint-PI+PE architectures are a sensible way to show that the power action matters.

Soft spots, in order of severity. First, the confound above; it is load-bearing. Second, the MDGNN comparison is not a fair architecture test: MDGNN has no mobility, so at least part of the 21-31% gap is just the extra degree of freedom from moving APs. The paper says this in words but still presents it as evidence for the framework. Third, there are no seeds, confidence intervals, or statistical tests anywhere; every figure shows a single run. That would be minor if the effect size were huge, but with the search confound it matters. Fourth, Corollary 1 and Theorem 1 are stated without proof and deferred to [4]; common but keeps the paper from being self-contained. Finally, the channel-hardening assumption is doing work with N=8 antennas per AP, which is on the low side for hardening; the gains may not transfer to smaller antenna arrays.

Bottom line: this is a useful template for people working on MARL for cell-free mMIMO, and the action-decomposition idea is worth discussing. But as written, the empirical claim does not support the architecture's contribution. Send it to peer review with the expectation of major revision: disable or bound the inner search at inference, match baseline capabilities, and add seeds and error bars. With those changes the paper could be solid.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported sum-SE gains are partly forced by Algorithm 1's inner loop, which selects actions by directly optimizing the same sum-SE metric used for evaluation.

fitted input called prediction [Algorithm 1 (Sec. III-D), Algorithm 2 line 5 (Sec. IV-B), and Sec. V numerical comparison]
"Algorithm 1, step 8: 'Update sum reward r(i) ex,t = PK k=1 SE(i) k,t with generated actions [a(i) l,t,inv, a(i) l,t,equiv];' step 9: 'until r(i) ex,t < r(i) ex,t or i ⩾ Imax;' Algorithm 2, step 5: 'Obtain action information al,t generated by HDPN architecture with ol,t based on Algorithm 1.' Sec. V: 'the proposed SF-MADDPG yields a 45.97%, 25.29%, 18.79%, and 14.41% improvement in sum SE.'"

The action chosen at each decision step is not a single-shot policy output: Algorithm 1 iteratively generates candidate actions and evaluates them using the extrinsic sum-SE reward, i.e., precisely the metric reported in the numerical comparisons. The reported gains therefore do not isolate the claimed architecture combination (GNN communication, permutation, directional decoupling); part of the improvement is forced by construction because the method performs an online search over the objective function while the baselines are not described as having the same search budget. This is a fitted input (actions fitted to sum SE) being presented as an architectural prediction.

full rationale

Apart from the Algorithm 1 confound, the paper's SE expression is imported from the external reference [4] (not fitted to the results), and the RL comparison is an empirical benchmark rather than a derived law. Self-citations ([5], [17], [20]) are used as design antecedents and baselines, not as load-bearing uniqueness or derivation theorems; the 'Theorems' on permutation invariance/equivariance are definitional properties of composition, not circular predictions. The channel-hardening assumption is an assumption, not a circular step. Thus the central circularity burden is the inference-time reward-based action search in Algorithm 1, which makes the headline percentage gains partially reduce to direct access to the evaluation metric. If Imax were 1 at test time or baselines were given the same search budget, the architectural claim could be isolated; as written, the evidence is confounded.

Assumptions & free parameters 6 free parameters · 6 assumptions · 4 invented entities

The ledger is dominated by hand-chosen hyperparameters and domain assumptions. No new physical entity is postulated. The algorithmic modules introduced have no external validation outside the in-house simulator.

free parameters (6)

Learning rate and soft update rate = 0.01
Chosen by hand for all networks (Table II); reported gains depend on this choice.
Discount factors gamma, gamma_a, gamma_ex = 0.99
Hand-set in Table II for the mixed, advantage, and extrinsic critics.
Network hidden sizes and attention heads = 128/64/256, 8 and 4 heads
Hand architecture choices (Table II) with no sensitivity analysis.
Experience replay pool capacity = 1024
Hand-set in Table II.
Maximum gradient clipping value = 0.5
Hand-set in Table II.
Number of GNN layers J = not reported
The GNN update in (42)-(43) iterates J layers, but J is never specified; performance depends on this choice.

assumptions (6)

domain assumption Channel hardening: normalized instantaneous channel gain converges to deterministic average gain, so power can be optimized using large-scale fading coefficients only.
Invoked in Sec II-C and Sec IV-B; requires many antennas and is load-bearing for the mobility/power decomposition.
standard math Standard MMSE channel estimation and use-and-then-forget capacity bound from [4].
Corollary 1 and Theorem 1 are stated with proofs omitted and deferred to [4].
domain assumption Correlated Rayleigh fading with covariance Rmk and large-scale fading beta_mk = L(d_mk).
System model Eq. (1) and Sec II; simulation uses the COST Walfish-Ikegami pathloss model.
domain assumption TDD protocol with orthogonal pilots and tau_p = K.
Sec II-A and Sec V; pilot assignment is deterministic but arbitrary.
domain assumption The environment is Markov and each agent's partial observation plus GNN messages is sufficient for near-optimal policies.
MDP model in Sec III-A and GNN communication in Sec IV-A; not proven for this setup.
domain assumption MR precoding is used and the closed-form SE expression remains accurate.
Sec II-B applies MR precoding at all mobile-APs and Theorem 1 gives the closed-form SE.

invented entities (4)

Dynamic Hyper Permutation Network (DHPN)
purpose: Generates per-entity weight matrices via hypernetworks to enforce permutation invariance and equivariance and reduce observation dimension.
Introduced in Sec III-D; only evaluated inside the paper's simulator, no external benchmark.
Attention-based intrinsic reward network (ARN)
purpose: Produces individual intrinsic rewards to allocate shared sum-SE reward among mobile-AP agents.
Introduced in Sec IV-B; no independent evidence outside the reported simulations.
Hypernetwork-based mixed reward network (HRN)
purpose: Combines intrinsic and extrinsic rewards for the mixed critic.
Introduced in Sec IV-B; no independent evidence.
GNN-aided communication architecture
purpose: Lets agents exchange and aggregate local messages instead of global state.
Adapted from cited work [20] into this framework; no independent measurement.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Mobile Cell-Free Massive MIMO with Multi-Agent Reinforcement Learning: A Scalable Framework." pith.science (2026). https://pith.science/paper/NABAIEOK

@misc{pith2026241202581,
  author       = {Pith},
  title        = {Pith review of: Mobile Cell-Free Massive MIMO with Multi-Agent Reinforcement Learning: A Scalable Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NABAIEOK}},
  note         = {Machine review of arXiv:2412.02581}
}

read the original abstract

Cell-free massive multiple-input multiple-output (mMIMO) offers significant advantages in mobility scenarios, mainly due to the elimination of cell boundaries and strong macro diversity. In this paper, we examine the downlink performance of cell-free mMIMO systems equipped with mobile-APs utilizing the concept of unmanned aerial vehicles, where mobility and power control are jointly considered to effectively enhance coverage and suppress interference. However, the high computational complexity, poor collaboration, limited scalability, and uneven reward distribution of conventional optimization schemes lead to serious performance degradation and instability. These factors complicate the provision of consistent and high-quality service across all user equipments in downlink cell-free mMIMO systems. Consequently, we propose a novel scalable framework enhanced by multi-agent reinforcement learning (MARL) to tackle these challenges. The established framework incorporates a graph neural network (GNN)-aided communication mechanism to facilitate effective collaboration among agents, a permutation architecture to improve scalability, and a directional decoupling architecture to accurately distinguish contributions. In the numerical results, we present comparisons of different optimization schemes and network architectures, which reveal that the proposed scheme can effectively enhance system performance compared to conventional schemes due to the adoption of advanced technologies. In particular, appropriately compressing the observation space of agents is beneficial for achieving a better balance between performance and convergence.

Figures

Figures reproduced from arXiv: 2412.02581 by the authors.

**Figure 4.** The two permutation types of actions in the cell-free mMIMO system [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 6.** CDF of sum SE under a single PI architecture over different [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

Figures from the paper (5 more)

**Figure 7.** Figure 7: CDF of sum SE under the joint PI and PE architecture over different [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

**Figure 9.** Figure 9: Sum SE against the number of mobile-APs M under a single PI architecture over different MARL-based optimization schemes with K = 8, N = 8, τp = K, and ∆ = λ/2. in severe interference situations, due to the initial exploration space being compressed by permutation archi…

**Figure 10.** Figure 10: Sum SE against the number of mobile-APs M under the joint PI and PE architecture over different MARL-based optimization schemes with K = 8, N = 8, τp = K, and ∆ = λ/2. 2 4 8 16 32 64 Number of antennas per mobile-AP 30 40 50 60 70 80 90 Sum SE [bits/s/Hz] Proposed Pe-…

**Figure 13.** Figure 13: Convergence examples of the proposed SF-MADDPG and MADDPG [PITH_FULL_IMAGE:figures/full_fig_p011_13.png]

**Figure 14.** Figure 14: Convergence examples of the proposed SF-MADDPG and MDGNN [PITH_FULL_IMAGE:figures/full_fig_p012_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 30 canonical work pages

[4]

Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,

E. Bj ¨ornson and L. Sanguinetti, “Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,” IEEE Trans. Wireless Commun. , vol. 19, no. 1, pp. 77–90, Jan. 2020

work page 2020
[1]

On the road to 6G: Visions, requirements, key technologies, and testbeds,

C.-X. Wang, X. You, X. Gao, X. Zhu, Z. Li, C. Zhang, H. Wang, Y . Huang, Y . Chen, H. Haas, J. S. Thompson, E. G. Larsson, M. D. Renzo, W. Tong, P. Zhu, X. Shen, H. V . Poor, and L. Hanzo, “On the road to 6G: Visions, requirements, key technologies, and testbeds,”IEEE Commun. Surveys Tuts., vol. 25, no. 2, pp. 905–974, Secondquarter 2023

work page 2023
[2]

Machine learning for large-scale optimization in 6G wireless networks,

Y . Shi, L. Lian, Y . Shi, Z. Wang, Y . Zhou, L. Fu, L. Bai, J. Zhang, and W. Zhang, “Machine learning for large-scale optimization in 6G wireless networks,” IEEE Commun. Surveys Tuts. , vol. 25, no. 4, pp. 2088–2132, Fourthquarter 2023

work page 2023
[3]

Prospective multiple antenna technologies for beyond 5G,

J. Zhang, E. Bj ¨ornson, M. Matthaiou, D. W. K. Ng, H. Yang, and D. J. Love, “Prospective multiple antenna technologies for beyond 5G,” IEEE J. Sel. Areas Commun. , vol. 38, no. 8, pp. 1637–1660, Aug. 2020

2020
[5]

Cell-free XL-MIMO meets multi-agent reinforcement learning: Architectures, challenges, and future directions,

Z. Liu, J. Zhang, Z. Liu, H. Du, Z. Wang, D. Niyato, M. Guizani, and B. Ai, “Cell-free XL-MIMO meets multi-agent reinforcement learning: Architectures, challenges, and future directions,” IEEE Wireless Com- mun., vol. 31, no. 4, pp. 155–162, Aug. 2024

work page 2024
[6]

Graph representation learning for wireless communications,

M. Mohsenivatani, S. Ali, V . Ranasinghe, N. Rajatheva, and M. Latva- Aho, “Graph representation learning for wireless communications,” IEEE Commun. Mag. , vol. 62, no. 1, pp. 141–147, Jan. 2024

work page 2024
[7]

Machine learning- based cell-free support in the O-RAN architecture: An innovative converged optical-wireless solution toward 6G networks,

J. S. Vardakas, K. Ramantas, E. Vinogradov, M. A. Rahman, A. Girycki, S. Pollin, S. Pryor, P. Chanclou, and C. Verikoukis, “Machine learning- based cell-free support in the O-RAN architecture: An innovative converged optical-wireless solution toward 6G networks,”IEEE Wireless Commun., vol. 29, no. 5, pp. 20–26, Oct. 2022

work page 2022
[8]

Design and optimization of RSMA for coexisting HTC/MTC in 6G and future networks,

S. Zhang, J. Liu, Z. Shi, J. Wang, and N. Kato, “Design and optimization of RSMA for coexisting HTC/MTC in 6G and future networks,” IEEE Trans. Wireless Commun., vol. 22, no. 12, pp. 9533–9548, Dec. 2023

work page 2023

Show all 34 references

[9]

Performance of cell-free massive MIMO with rician fading and phase shifts,

O. ¨Ozdogan, E. Bj ¨ornson, and J. Zhang, “Performance of cell-free massive MIMO with rician fading and phase shifts,” IEEE Trans. Wireless Commun., vol. 18, no. 11, pp. 5299–5315, Nov. 2019

2019
[10]

Uplink performance of RIS-aided cell-free massive MIMO system with electromagnetic interference,

E. Shi, J. Zhang, D. W. K. Ng, and B. Ai, “Uplink performance of RIS-aided cell-free massive MIMO system with electromagnetic interference,” IEEE J. Sel. Areas Commun. , vol. 41, no. 8, pp. 2431– 2445, Aug. 2023

2023
[11]

Air-ground collaborative resource optimization in UA V empowered cell-free massive MIMO systems,

L. Xu, Q. Zhu, W. Xia, T. Q. S. Quek, and H. Zhu, “Air-ground collaborative resource optimization in UA V empowered cell-free massive MIMO systems,” IEEE Trans. Commun., vol. 72, no. 4, pp. 2485–2499, Apr. 2024

2024
[12]

UxNB-enabled cell-free massive MIMO with HAPS-assisted Sub-THz backhauling,

O. Abbasi and H. Yanikomeroglu, “UxNB-enabled cell-free massive MIMO with HAPS-assisted Sub-THz backhauling,” IEEE Trans. Veh. Technol., vol. 73, no. 5, pp. 6937–6953, May. 2024. 13

2024
[13]

Analysis of UA V communications in cell-free massive MIMO systems,

C. D’Andrea, A. Garcia-Rodriguez, G. Geraci, L. G. Giordano, and S. Buzzi, “Analysis of UA V communications in cell-free massive MIMO systems,” IEEE Open J. Commun. Soc. , vol. 1, pp. 133–147, Jan. 2020

2020
[14]

Joint power control and LSFD for wireless-powered cell-free massive MIMO,

O. T. Demir and E. Bj ¨ornson, “Joint power control and LSFD for wireless-powered cell-free massive MIMO,” IEEE Trans. Wireless Com- mun., vol. 20, no. 3, pp. 1756–1769, Mar. 2021

2021
[15]

Joint power control and access point scheduling in fronthaul-constrained uplink cell-free massive MIMO systems,

M. Guenach, A. A. Gorji, and A. Bourdoux, “Joint power control and access point scheduling in fronthaul-constrained uplink cell-free massive MIMO systems,” IEEE Trans. Commun., vol. 69, no. 4, pp. 2709–2722, Apr. 2021

2021
[16]

Learning decentralized power control in cell-free massive MIMO networks,

D. Yu, H. Lee, S.-E. Hong, and S.-H. Park, “Learning decentralized power control in cell-free massive MIMO networks,” IEEE Trans. Veh. Technol., vol. 72, no. 7, pp. 9653–9658, Jul. 2023

2023
[17]

Double-layer power control for mobile cell-free XL-MIMO with multi-agent reinforcement learning,

Z. Liu, J. Zhang, Z. Liu, H. Xiao, and B. Ai, “Double-layer power control for mobile cell-free XL-MIMO with multi-agent reinforcement learning,” IEEE Trans. Wireless Commun., vol. 23, no. 5, pp. 4658–4674, May 2024

2024
[18]

In- teraction pattern disentangling for multi-agent reinforcement learning,

S. Liu, J. Song, Y . Zhou, N. Yu, K. Chen, Z. Feng, and M. Song, “In- teraction pattern disentangling for multi-agent reinforcement learning,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 12, pp. 8157–8172, Dec. 2024

2024
[19]

Emergent communication in multi-agent reinforcement learning for future wireless networks,

M. Chafii, S. Naoumi, R. Alami, E. Almazrouei, M. Bennis, and M. Debbah, “Emergent communication in multi-agent reinforcement learning for future wireless networks,” IEEE Internet Things Mag. , vol. 6, no. 4, pp. 18–24, Dec. 2023

2023
[20]

Graph neural network meets multi-agent reinforcement learning: fundamentals, applications, and future directions,

Z. Liu, J. Zhang, E. Shi, Z. Liu, D. Niyato, B. Ai, and X. S. Shen, “Graph neural network meets multi-agent reinforcement learning: fundamentals, applications, and future directions,” IEEE Wireless Commun. , early access, 2024

2024
[21]

Multi-agent graph-attention communication and teaming,

Y . Niu, R. R. Paleja, and M. C. Gombolay, “Multi-agent graph-attention communication and teaming,” in AAMAS, 2021

2021
[22]

Graph neural networks for wireless communications: From theory to practice,

Y . Shen, J. Zhang, S. H. Song, and K. B. Letaief, “Graph neural networks for wireless communications: From theory to practice,” IEEE Trans. Wireless Commun., vol. 22, no. 5, pp. 3554–3569, May 2023

2023
[23]

The road to 6G: Ten physical layer chal- lenges for communications engineers,

M. Matthaiou, O. Yurduseven, H. Q. Ngo, D. Morales-Jimenez, S. L. Cotton, and V . F. Fusco, “The road to 6G: Ten physical layer chal- lenges for communications engineers,” IEEE Communications Maga- zine, vol. 59, no. 1, pp. 64–69, Jan. 2021

2021
[24]

Deep reinforcement learning: A brief survey,

K. Arulkumaran, M. P. Deisenroth, M. Brundage, and A. A. Bharath, “Deep reinforcement learning: A brief survey,” IEEE Signal Process. Mag., vol. 34, no. 6, pp. 26–38, Nov. 2017

2017
[25]

Decentralized computation offloading with cooperative UA Vs: Multi-agent deep reinforcement learning perspective,

S. Hwang, H. Lee, J. Park, and I. Lee, “Decentralized computation offloading with cooperative UA Vs: Multi-agent deep reinforcement learning perspective,”IEEE Wireless Commun., vol. 29, no. 4, pp. 24–31, Aug. 2022

2022
[26]

Multi-agent reinforcement learning-based joint precoding and phase shift optimization for RIS- aided cell-free massive MIMO systems,

Y . Zhu, E. Shi, Z. Liu, J. Zhang, and B. Ai, “Multi-agent reinforcement learning-based joint precoding and phase shift optimization for RIS- aided cell-free massive MIMO systems,” IEEE Trans. Veh. Technol. , vol. 73, no. 9, pp. 14 015–14 020, Sep. 2024

2024
[27]

Effective commu- nications: A joint learning and communication framework for multi- agent reinforcement learning over noisy channels,

T.-Y . Tung, S. Kobus, J. P. Roig, and D. G ¨und¨uz, “Effective commu- nications: A joint learning and communication framework for multi- agent reinforcement learning over noisy channels,” IEEE J. Sel. Areas Commun., vol. 39, no. 8, pp. 2590–2603, Aug. 2021

2021
[28]

Learning random access schemes for massive machine-type communication with MARL,

M. A. Jadoon, A. Pastore, M. Navarro, and A. Valcarce, “Learning random access schemes for massive machine-type communication with MARL,” IEEE Trans. Mach. Learn. Commun. Netw., vol. 2, pp. 95–109, 2024

2024
[29]

Permutation invariant policy optimization for mean-field multi-agent re- inforcement learning: A principled approach,

Y . Li, L. Wang, J. Yang, E. Wang, Z. Wang, T. Zhao, and H. Zha, “Permutation invariant policy optimization for mean-field multi-agent re- inforcement learning: A principled approach,” arXiv:2105.08268, 2021

2021 arXiv
[30]

Boosting multiagent reinforcement learning via permutation invariant and permutation equivariant networks,

J. Hao, X. Hao, H. Mao, W. Wang, Y . Yang, D. Li, Y . Zhang, and Z. Wang, “Boosting multiagent reinforcement learning via permutation invariant and permutation equivariant networks,” in ICLR, 2023

2023
[31]

LIIR: Learn- ing individual intrinsic reward in multi-agent reinforcement learning,

Y . Du, L. Han, M. Fang, J. Liu, T. Dai, and D. Tao, “LIIR: Learn- ing individual intrinsic reward in multi-agent reinforcement learning,” NeurIPS, vol. 32, 2019

2019
[32]

Generating individual intrinsic reward for cooperative multiagent reinforcement learning,

H. Wu, H. Li, J. Zhang, Z. Wang, and J. Zhang, “Generating individual intrinsic reward for cooperative multiagent reinforcement learning,” International Journal of Advanced Robotic Systems, vol. 18, no. 5, 2021

2021
[33]

Attention-based intrinsic reward mixing network for credit assignment in multi-agent reinforcement learning,

W. Li, W. Liu, S. Shao, S. Huang, and A. Song, “Attention-based intrinsic reward mixing network for credit assignment in multi-agent reinforcement learning,” IEEE Trans. Games , vol. 16, no. 2, pp. 270– 281, Jun. 2024

2024
[34]

Multidimensional graph neural networks for wireless communications,

S. Liu, J. Guo, and C. Yang, “Multidimensional graph neural networks for wireless communications,” IEEE Trans. Wireless Commun., vol. 23, no. 4, pp. 3057–3073, Apr. 2024

2024

Pith tools

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

[1] [4]

Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,

E. Bj ¨ornson and L. Sanguinetti, “Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,” IEEE Trans. Wireless Commun. , vol. 19, no. 1, pp. 77–90, Jan. 2020

work page 2020

[2] [1]

On the road to 6G: Visions, requirements, key technologies, and testbeds,

C.-X. Wang, X. You, X. Gao, X. Zhu, Z. Li, C. Zhang, H. Wang, Y . Huang, Y . Chen, H. Haas, J. S. Thompson, E. G. Larsson, M. D. Renzo, W. Tong, P. Zhu, X. Shen, H. V . Poor, and L. Hanzo, “On the road to 6G: Visions, requirements, key technologies, and testbeds,”IEEE Commun. Surveys Tuts., vol. 25, no. 2, pp. 905–974, Secondquarter 2023

work page 2023

[3] [2]

Machine learning for large-scale optimization in 6G wireless networks,

Y . Shi, L. Lian, Y . Shi, Z. Wang, Y . Zhou, L. Fu, L. Bai, J. Zhang, and W. Zhang, “Machine learning for large-scale optimization in 6G wireless networks,” IEEE Commun. Surveys Tuts. , vol. 25, no. 4, pp. 2088–2132, Fourthquarter 2023

work page 2023

[4] [3]

Prospective multiple antenna technologies for beyond 5G,

J. Zhang, E. Bj ¨ornson, M. Matthaiou, D. W. K. Ng, H. Yang, and D. J. Love, “Prospective multiple antenna technologies for beyond 5G,” IEEE J. Sel. Areas Commun. , vol. 38, no. 8, pp. 1637–1660, Aug. 2020

2020

[5] [5]

Cell-free XL-MIMO meets multi-agent reinforcement learning: Architectures, challenges, and future directions,

Z. Liu, J. Zhang, Z. Liu, H. Du, Z. Wang, D. Niyato, M. Guizani, and B. Ai, “Cell-free XL-MIMO meets multi-agent reinforcement learning: Architectures, challenges, and future directions,” IEEE Wireless Com- mun., vol. 31, no. 4, pp. 155–162, Aug. 2024

work page 2024

[6] [6]

Graph representation learning for wireless communications,

M. Mohsenivatani, S. Ali, V . Ranasinghe, N. Rajatheva, and M. Latva- Aho, “Graph representation learning for wireless communications,” IEEE Commun. Mag. , vol. 62, no. 1, pp. 141–147, Jan. 2024

work page 2024

[7] [7]

Machine learning- based cell-free support in the O-RAN architecture: An innovative converged optical-wireless solution toward 6G networks,

J. S. Vardakas, K. Ramantas, E. Vinogradov, M. A. Rahman, A. Girycki, S. Pollin, S. Pryor, P. Chanclou, and C. Verikoukis, “Machine learning- based cell-free support in the O-RAN architecture: An innovative converged optical-wireless solution toward 6G networks,”IEEE Wireless Commun., vol. 29, no. 5, pp. 20–26, Oct. 2022

work page 2022

[8] [8]

Design and optimization of RSMA for coexisting HTC/MTC in 6G and future networks,

S. Zhang, J. Liu, Z. Shi, J. Wang, and N. Kato, “Design and optimization of RSMA for coexisting HTC/MTC in 6G and future networks,” IEEE Trans. Wireless Commun., vol. 22, no. 12, pp. 9533–9548, Dec. 2023

work page 2023