VariAntNet: Learning Decentralized Control of Multi-Agent Systems

Alfred M. Bruckstein; Ariel Barel; Eran Iceland; Erez Koifman; Yigal Koifman

REVIEW 3 major objections 7 minor 23 references

VariAntNet: Learning Decentralized Control of Multi-Agent Systems

T0 review · 3 major / 7 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A learned, decentralized controller gathers agent swarms up to 2.5x faster than a provable analytical rule.

desk verdict The architecture and loss are a real step forward for bearing-only gathering, but the 2.5x speedup is only as good as the unstated step size used for the analytical baseline. read the letter →

arxiv 2509.02271 v1 pith:CERWFWB4 submitted 2025-09-02 cs.LG cs.AIcs.MA

Yigal Koifman , Erez Koifman , Eran Iceland , Ariel Barel , Alfred M. Bruckstein This is my paper · ORCID

classification cs.LGcs.AIcs.MA

keywords swarmroboticsdecentralizedcontrolgatheringproblembearing-onlysensingdeeplearninggraphLaplacianalgebraicconnectivityCTDE

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

VariAntNet is a deep-learning controller for swarms of anonymous robots that sense only the directions to nearby neighbors and never communicate. The paper claims that on the gathering task—converging on a common location without a shared coordinate system—this learned policy converges up to 2.5 times faster than the analytical bearing-only algorithm of Bellaiche et al., while keeping the swarm connected in 90.6–100% of scenarios when the initial visibility ratio is at or below 0.75. The speed comes from a neural policy trained with a loss that balances a task term (max distance from the swarm center) against a cohesiveness term derived from the algebraic connectivity of the visibility graph. That trade-off is the paper's central practical point: in time-critical applications, losing a few agents may be worth gathering the rest quickly.

What carries the argument

The load-bearing mechanism is the cohesiveness loss: one over the algebraic connectivity (second smallest eigenvalue) of the weighted visibility graph, a differentiable lower-bound surrogate for Cheeger's constant. Minimizing this loss during training pushes the swarm away from configurations that are about to split. Around it sit two architectural choices: a rotation-equivariant preprocessing step that aligns each observation to the summed bearing vector, making the policy independent of the agent's orientation, and max pooling over per-neighbor MLP features, making it independent of observation order and swarm size.

What would settle it

Implement both controllers in the same simulator with identical maximum step size and time discretization, then measure average steps to convergence; if the analytical baseline matches VariAntNet's speed, the 2.5x claim fails.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a compact, fully decentralized neural network can solve the bearing-only gathering problem faster than a provably convergent analytical controller. VariAntNet processes each agent's unordered, variable-sized set of bearing vectors through a rotation-invariant preprocessing step and a max-pooling feature extractor, then outputs a direction and a step size. Training minimizes a weighted sum of the maximum distance to the swarm center of mass and the reciprocal of the algebraic connectivity of the weighted visibility graph, which acts as a differentiable surrogate for the Cheeger constant. In simulations over 1,000 random initial constellations per s

Load-bearing premise

The claimed speedup presumes the analytical baseline is run with the same per-step motion budget as VariAntNet; if the baseline uses smaller effective steps, the comparison is unfair. The simulations also assume noise-free bearings, collision-free motion, and a connected initial constellation.

Editorial extensions

If this is right

Learned decentralized policies can make bearing-only swarms converge in about half the steps of an analytical rule, which matters for time-critical missions.
The algebraic-connectivity loss is differentiable and transferable, so other centralized-training/decentralized-execution swarm policies could adopt the same cohesion objective.
The architecture's parameter count stays constant as swarm size grows, since max pooling aggregates arbitrary numbers of neighbor features.
Operators can tune the speed-cohesion trade-off by reweighting the loss terms, choosing a policy that accepts some disconnection for faster gathering.

Reading between the lines

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

A hybrid policy that switches to the conservative analytical rule whenever the estimated algebraic connectivity drops below a threshold could preserve most of the speedup while eliminating worst-case fragmentation.
The same preprocessing and loss could be adapted to other decentralized tasks—coverage, rendezvous, or flocking—that rely on bearing-only local sensing.
Re-benchmarking the analytical baseline with per-step displacement exactly matched to the network's maximum step size would test whether the reported speedup is intrinsic or a comparison artifact.

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

Thought you'd want to know before the next reading group. This paper deserves a serious look, but with one load-bearing caveat.

What's new: they train a compact PointNet-style network (3.8k params) for decentralized bearing-only gathering, with rotation-equivariant preprocessing and a differentiable loss based on the algebraic connectivity / Cheeger bound. That combination is new for this task. The architecture handles variable-size unordered observations and no shared frame. Lemma 1 on rotational equivariance is correct for non-zero resultant observations, and they disclose the zero-resultant exception. The empirical setup is solid: 1,000 shared scenarios across models, up to 2,500 steps, and they report both convergence and connectivity. They also get credit for being transparent about the fragmentation trade-off; the paper's own body scopes the claim to VR<=0.875 and calls 10% disconnection acceptable.

The soft spots are real. First, the headline speedup (up to 2.5x faster than Bellaiche et al.) depends on a comparison fairness detail that is never specified: the analytical baseline is a continuous-time rule, while VariAntNet's action caps per-step displacement at sigma<=1. To count 'steps' for the analytical baseline you must choose a discrete-time step / Euler h, and the paper doesn't state it. If the baseline is run with smaller effective steps, the ratio inflates. This is not a minor point — it directly underpins the main claim. The stress-test note is right to flag it. That said, the qualitative result (a learned controller can gather faster than the conservative analytical rule) is plausible and partially supported by the curve shapes, so I wouldn't dismiss the paper over it; I just wouldn't trust the 2.5x number until this is documented.

Second, the abstract overstates connectivity. It says 'maintaining high swarm connectivity across varying swarm sizes,' but Table 3 shows connectivity dropping to 41.3% for the fastest variant at VR=1, N=30. The body is honest about this; the abstract is not. Third, Table 3 has no error bars, which is a visible omission for a paper whose message is about averages. Minor: Table 1's layer numbering is off, and no code or data is released, so the baseline step size cannot be checked post hoc.

The central argument — that a learned policy can trade a little cohesion for a lot of speed — holds up as scoped. Readers in swarm robotics and learning-based control will get real value. I'd send it to peer review: the idea is timely, the architecture is clean, and the flaws are fixable with a corrected abstract and a documented baseline discretization.

Referee Report

3 major / 7 minor

Summary. The paper proposes VariAntNet, a deep-learning decentralized controller for anonymous agents with bearing-only, limited-range sensing, applied to the multi-agent gathering problem. The controller uses rotation-equivariant preprocessing, order/size-invariant aggregation of neighbor bearing vectors via max pooling, and a loss combining a task term (max distance from the swarm center of mass) with a cohesiveness term based on the reciprocal of the algebraic connectivity of a weighted visibility graph. Training uses a curriculum over visibility ratios; evaluation compares two VariAntNet variants against the analytical Bellaiche et al. controller on 1,000 scenarios per cell for N=10, 20, 30 and five visibility-ratio levels. The headline claim is that VariAntNet converges 1.7-2.5x faster than the analytical baseline while keeping the swarm connected in most but not all configurations, framed as a speed-cohesion tradeoff with a 'golden point' at VR <= 0.875.

Significance. If validated, the result would be a genuinely useful contribution: it demonstrates that a compact learned controller can substantially outperform a provably convergent, conservative analytical rule on a well-studied bearing-only gathering problem, and it offers a reusable graph-Laplacian-based cohesion loss for CTDE-trained swarm controllers. The paper is transparent about fragmentation and the tradeoff, and the use of 1,000 shared scenarios across models is a strength. The main caveat is that the speed comparison depends on the baseline's discrete-time implementation, which is not described in the manuscript.

major comments (3)

[§6.2, Table 3] The baseline discretization is not specified. The analytical rule in [1] is continuous-time; to count 'steps' one must choose an Euler step h or per-step displacement. VariAntNet's action space restricts per-step displacement to sigma <= 1 (§3.2), so the learned policy's motion budget per step is capped. If the analytical baseline is simulated with a smaller effective step, the reported step ratios (e.g., 1629/646 ≈ 2.5x for N=30, VR=1) become an artifact of unequal motion budgets rather than a property of the learned policy. The manuscript does not report the baseline's integration step, per-step displacement, or wall-clock time. Please state this implementation choice, justify matching the same motion budget, or report physical time.
[Abstract, §6.2] The abstract states that VariAntNet achieves 'more than double the convergence rate while maintaining high swarm connectivity across varying swarm sizes.' Table 3 at Marginal (VR=1), N=30 gives Conn.% = 41.3 for the recommended weighted variant; this is not high connectivity. The body appropriately discusses the tradeoff and defines the 'golden point' at VR <= 0.875, but the abstract and conclusion should carry the same qualification. Otherwise readers may take the headline claim to apply to all evaluated regimes.
[§4.1, Lemma 1] The equivariance guarantee is conditional on ||O_i(t) 1_n|| != 0. In the zero-sum case, the paper states 'no rotation is applied.' Such observations can arise naturally, e.g., when an agent sees two neighbors in exactly opposite directions. For an agent with no shared coordinate system, the resulting action would then depend on the arbitrary local frame, breaking the claimed 'guaranteed through analytical transformations' property. The authors should state whether zero-resultant observations can occur under the assumptions and, if so, explain how equivariance is preserved or why it is not needed in those cases.

minor comments (7)

[Table 3, Fig. 7] Report standard errors or confidence intervals for the mean convergence steps. Figure 7 shows a '±1' band but the caption does not define whether it is standard deviation, standard error, or another spread measure.
[§4.2] The action space requires 0 <= sigma <= 1, but the step-size output is 'FC 32 -> 1 ReLU'. A ReLU can output values greater than 1; the clipping or normalization used to enforce the action-space bound is not described.
[§4.2] Notation is inconsistent: the first block is described as f_phi and later as f_theta. Please unify.
[§6.1] The text says 'initial visibility ratio (AR)' but the symbol is VR throughout. Correct the typo.
[Table 1] Layer numbering is inconsistent: rows 3 and 4 are swapped, and two rows are labelled '7'. Renumber the layers to match the architecture description.
[Fig. 9] The caption defines DR as 'expected size of the largest connected subgroup' while §6.2 defines it as expected disconnected ratio (percentage of agents that separate from the largest group). Align the two definitions; the figure labels are also quite dense and hard to read.
[Throughout] Spelling/typographical issues include 'preformed' (should be 'performed') in §6.2 and 'Analitical' in Fig. 7. Also, the paper does not mention code or data availability; for reproducibility, consider providing them in the final version.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central loss derivation and empirical comparison are self-contained; the analytical baseline is an external published algorithm.

full rationale

The paper's central methodological claim is the VariAntNet loss, L = alpha*LCohesiveness + beta*LTask, where LTask is the maximum distance to the swarm centroid and LCohesiveness is 1/(lambda_2 + epsilon), with lambda_2 the algebraic connectivity of a weighted visibility graph. This is justified by a standard Cheeger inequality lower bound (h(G) >= lambda_2/2), not by assuming the target result. The training objective does directly optimize a metric related to the evaluation metric (distance to centroid), but that is standard task fitting, not a renamed prediction or a definitional identity with the claimed finding. The headline comparison against Bellaiche et al. [1] is an empirical benchmark on 1,000 freshly generated constellations; although [1] shares an author (Bruckstein), it is an externally published, independently falsifiable algorithm, so the self-citation is not load-bearing in the derivation. The unspecified discrete-time step size for the analytical baseline and the tuning of alpha,beta are legitimate correctness and generalization concerns, but they are not circular reductions: no equation in the paper defines the predicted speedup in terms of a fitted parameter or a self-citation chain. Under the stated rules, this is a non-finding with score 0.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the sensor and connectivity model (unit-vector observations, visibility graph cohesion), standard spectral graph results (Cheeger bound, differentiability of lambda_2), and several hand-chosen quantities (loss weights, epsilon, network shape). No new physical entities are introduced. The free parameters are few and mostly disclosed, but the loss weights are tuned on the evaluation set used for the headline comparison.

free parameters (3)

loss weights alpha, beta = alpha=1, beta=1; alpha=1, beta=10 (Weighted 1,10)
Hand-chosen and tuned via the ablation study (Section 6.3); they directly set the speed-connectivity trade-off reported in Table 3.
epsilon in L_Cohesiveness = 1e-6
Stability constant added to 1/lambda_2 (Section 4.3.2); hand-chosen.
NN architecture shape = 8 layers, widths 16/32, 3,875 params
Layer widths and activations (Table 1) are design choices, not derived; the total is consistent only after correcting mislabeled rows.

assumptions (6)

domain assumption Initial constellation is connected (visibility graph G(V,E) connected at t=0)
Section 3.1; the problem statement, generator (Algorithm 1), and cohesion goal all build on this, and disconnected agents 'may not be able to rejoin'.
domain assumption Bearing-only sensing: agents observe unit vectors to neighbors only, within range V, without noise
Sections 3.1-3.2; all claims about the policy rest on the sensor model; the real-world firefighting motivation adds noise, occlusion, and obstacles not modeled here.
standard math Cheeger inequality h(G) >= lambda_2/2 for weighted graphs
Section 4.3.2, cited to [12] and [22]; justifies substituting 1/lambda_2 for 1/h(G) as a cohesion loss.
standard math lambda_2 is continuous and differentiable with respect to edge weights
Section 4.3.2; required for gradient training; differentiability fails at eigenvalue crossings (degenerate lambda_2), a caveat the paper does not discuss.
ad hoc to paper Edge weighting w(vi,vj) = V - d(pi,pj) captures cohesion
Section 4.3.2; a design choice used only in training, not justified by a theorem or measured against alternative weightings.
ad hoc to paper Observations with zero resultant (||O_i 1_n|| = 0) can be left unrotated without harming the equivariance property
Section 4.1, Lemma 1 requires ||O_i 1_n|| != 0; the paper states 'no rotation is applied' otherwise, which breaks equivariance on exactly those inputs (e.g., two neighbors at opposite bearings).

how reviews work

0 comments

Cite this review

Pith. "Pith review of VariAntNet: Learning Decentralized Control of Multi-Agent Systems." pith.science (2026). https://pith.science/paper/CERWFWB4

@misc{pith2026250902271,
  author       = {Pith},
  title        = {Pith review of: VariAntNet: Learning Decentralized Control of Multi-Agent Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CERWFWB4}},
  note         = {Machine review of arXiv:2509.02271}
}

read the original abstract

A simple multi-agent system can be effectively utilized in disaster response applications, such as firefighting. Such a swarm is required to operate in complex environments with limited local sensing and no reliable inter-agent communication or centralized control. These simple robotic agents, also known as Ant Robots, are defined as anonymous agents that possess limited sensing capabilities, lack a shared coordinate system, and do not communicate explicitly with one another. A key challenge for simple swarms lies in maintaining cohesion and avoiding fragmentation despite limited-range sensing. Recent advances in machine learning offer effective solutions to some of the classical decentralized control challenges. We propose VariAntNet, a deep learning-based decentralized control model designed to facilitate agent swarming and collaborative task execution. VariAntNet includes geometric features extraction from unordered, variable-sized local observations. It incorporates a neural network architecture trained with a novel, differentiable, multi-objective, mathematically justified loss function that promotes swarm cohesiveness by utilizing the properties of the visibility graph Laplacian matrix. VariAntNet is demonstrated on the fundamental multi-agent gathering task, where agents with bearing-only and limited-range sensing must gather at some location. VariAntNet significantly outperforms an existing analytical solution, achieving more than double the convergence rate while maintaining high swarm connectivity across varying swarm sizes. While the analytical solution guarantees cohesion, it is often too slow in practice. In time-critical scenarios, such as emergency response operations where lives are at risk, slower analytical methods are impractical and justify the loss of some agents within the swarm. This paper presents and analyzes this trade-off in detail.

Figures

Figures reproduced from arXiv: 2509.02271 by the authors.

**Figure 2.** Agent p1 observation is given as unit vectors pointing toward its neighbors, constrained by bearingonly sensing with limited range V . A swarm is said to be cohesive if and only if the graph G(V, E) is a connected graph, i.e., there exists a path between any two vertices νi and νj in G. The agents sensed by agent i at time t are defined as the set of agent i’s neighbors, Ni(t), in the visibility graph G(V, E): Ni(… view at source ↗

**Figure 3.** The VariAntNet pipeline includes a preprocessing and postprocessing rotational equivariant transformation and an [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: Illustrations (a) and (b) demonstrate rotational equivariance. The constellations and input representation are considered equivalent for p1’s decision process. NN block, described by gψ, which produces the agent’s action a ∈ A. The NN can be mathematically described as…

**Figure 5.** Figure 5: NN structure consists of fϕ(ˆui,j ), a shared MLP that encodes observations, a max pooling layer for extracting the agent’s local features, and a final MLP denoted gψ(LF(fϕ(ˆui,j ))), which determines the agent’s next step direction and size. 4.3. Geometric Loss Func…

**Figure 6.** Figure 6: The illustrations depict a successful positioning [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

**Figure 7.** Figure 7: Comparison of the average convergence steps of the swarm between the VariAntNet variants and the Analitical [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

**Figure 8.** Figure 8: Comparison of control models based on Expected Disconnection Ratio (DR, bars) and Average Convergence [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

**Figure 9.** Figure 9: Comparison of VariAntNet variants based on the Expected Disconnection Ratio (DR, bars) and Average Conver [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 22 canonical work pages

[1]

Continuous time gathering of agents with limited visibility and bearing-only sensing,

L. I. Bellaiche and A. Bruckstein, “Continuous time gathering of agents with limited visibility and bearing-only sensing,” Swarm Intelligence, vol. 11, no. 3, pp. 271–293, 2017, Springer

work page 2017
[2]

Dis- tributed memoryless point convergence algorithm for mobile robots with limited visibility,

H. Ando, Y . Oasa, I. Suzuki, and M. Yamashita, “Dis- tributed memoryless point convergence algorithm for mobile robots with limited visibility,”IEEE Transac- tions on Robotics and Automation, vol. 15, no. 5, pp. 818–828, 1999, IEEE

work page 1999
[3]

Stability analysis of swarms,

V . Gazi and K. M. Passino, “Stability analysis of swarms,” IEEE Transactions on Automatic Control , vol. 48, no. 4, pp. 692–697, 2003, IEEE

work page 2003
[4]

Come to- gether: Multi-agent geometric consensus (gathering, rendezvous, clustering, aggregation),

A. Barel, R. Manor, and A. M. Bruckstein, “Come to- gether: Multi-agent geometric consensus (gathering, rendezvous, clustering, aggregation),” arXiv preprint arXiv:1902.01455, 2019

arXiv 1902
[5]

Dis- tributed & decentralized control and task allocation for flexible swarms,

Y . Koifman, A. Barel, and A. Bruckstein, “Dis- tributed & decentralized control and task allocation for flexible swarms,” inThe 8th International Sympo- sium on Swarm Behavior and Bio-Inspired Robotics, pp. 137–144, 2024

work page 2024
[6]

The signal propagation effects on IEEE 802.15.4 radio link in fire environ- ment,

M. D. Chinthaka, N. H. Malka, K. Ramamohanarao, B. Moran, and P. Farrell, “The signal propagation effects on IEEE 802.15.4 radio link in fire environ- ment,” in 2010 Fifth International Conference on Information and Automation for Sustainability , pp. 411–414, 2010, IEEE

work page 2010
[7]

Radio Propagation in Fire Environ- ments,

J. A. Boan, “Radio Propagation in Fire Environ- ments,” Ph.D. dissertation, University of Adelaide, Australia, 2009

work page 2009
[8]

Eigenvalues of the Laplacian and their relationship to the connectedness of a graph,

A. Marsden, “Eigenvalues of the Laplacian and their relationship to the connectedness of a graph,” Uni- versity of Chicago, REU, 2013

work page 2013

Show all 23 references

[9]

Effects on the algebraic connectivity of weighted graphs under edge rotations,

X. Chen, S. Zhang, S. Gao, and X. Song, “Effects on the algebraic connectivity of weighted graphs under edge rotations,” Linear Algebra and its Applications, vol. 703, pp. 289–301, 2024, Elsevier

2024
[10]

A survey on robotic technologies for for- est firefighting: Applying drone swarms to improve firefighters’ efficiency and safety,

J. J. Rold ´an-G´omez, E. Gonz ´alez-Gironda, and A. Barrientos, “A survey on robotic technologies for for- est firefighting: Applying drone swarms to improve firefighters’ efficiency and safety,” Applied Sciences, vol. 11, no. 1, p. 363, 2021, MDPI

2021
[11]

Eigenvalues of the Laplacian of a graph,

W. N. Anderson Jr. and T. D. Morley, “Eigenvalues of the Laplacian of a graph,” Linear and Multilinear Algebra, vol. 18, no. 2, pp. 141–145, 1985, Taylor & Francis

1985
[12]

Eigenvalues of the Laplacian on a Graph,

J. Walchessen, “Eigenvalues of the Laplacian on a Graph,” Technical report, University of Chicago, 2019

2019
[13]

Deep reinforcement learning for swarm systems,

M. H ¨uttnerrauch, A. ˇSoˇsi´c, and G. Neumann, “Deep reinforcement learning for swarm systems,” Journal of Machine Learning Research , vol. 20, no. 54, pp. 1–31, 2019

2019
[14]

Reinforcement learning for swarm robotics: An overview of ap- plications, algorithms and simulators,

M.-A. Blais and M. A. Akhloufi, “Reinforcement learning for swarm robotics: An overview of ap- plications, algorithms and simulators,” Cognitive Robotics, 2023, Elsevier

2023
[15]

PointNet: Deep learning on point sets for 3D classification and segmentation,

C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “PointNet: Deep learning on point sets for 3D classification and segmentation,” in Proceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition , pp. 652–660, 2017

2017
[16]

Self-organized group for cooperative multi-agent reinforcement learning,

J. Shao, Z. Lou, H. Zhang, Y . Jiang, S. He, and X. Ji, “Self-organized group for cooperative multi-agent reinforcement learning,” Advances in Neural Infor- mation Processing Systems, vol. 35, pp. 5711–5723, 2022

2022
[17]

Guided deep reinforcement learning for swarm sys- tems,

M. H ¨uttnerrauch, A. ˇSoˇsi´c, and G. Neumann, “Guided deep reinforcement learning for swarm sys- tems,” arXiv preprint arXiv:1709.06011, 2017

2017 arXiv
[18]

A path planning algorithm for collective monitoring using autonomous drones,

S. Islam and A. Razi, “A path planning algorithm for collective monitoring using autonomous drones,” in 2019 53rd Annual Conference on Information Sci- ences and Systems (CISS), pp. 1–6, 2019, IEEE

2019
[19]

Deep reinforcement learning for multiagent sys- tems: A review of challenges, solutions, and appli- cations,

T. T. Nguyen, N. D. Nguyen, and S. Nahavandi, “Deep reinforcement learning for multiagent sys- tems: A review of challenges, solutions, and appli- cations,” IEEE Transactions on Cybernetics, vol. 50, no. 9, pp. 3826–3839, 2020, IEEE

2020
[20]

Multi-agent deep re- inforcement learning: A survey,

S. Gronauer and K. Diepold, “Multi-agent deep re- inforcement learning: A survey,” Artificial Intelli- gence Review , vol. 55, no. 2, pp. 895–943, 2022, Springer

2022
[21]

A survey of multi-agent deep reinforcement learning with com- munication,

C. Zhu, M. Dastani, and S. Wang, “A survey of multi-agent deep reinforcement learning with com- munication,” Autonomous Agents and Multi-Agent Systems, vol. 38, no. 1, p. 4, 2024, Springer

2024
[22]

A lower bound for the smallest eigen- value of the Laplacian,

J. Cheeger, “A lower bound for the smallest eigen- value of the Laplacian,” inProblems in Analysis, vol. 625, pp. 195–199, Princeton UP, 1970

1970
[23]

Deep decentralized multi-task multi-agent re- inforcement learning under partial observability,

S. Omidshafiei, J. Pazis, C. Amato, J. P. How, and J. Vian, “Deep decentralized multi-task multi-agent re- inforcement learning under partial observability,” in International Conference on Machine Learning , pp. 2681–2690, 2017, PMLR

2017

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Continuous time gathering of agents with limited visibility and bearing-only sensing,

L. I. Bellaiche and A. Bruckstein, “Continuous time gathering of agents with limited visibility and bearing-only sensing,” Swarm Intelligence, vol. 11, no. 3, pp. 271–293, 2017, Springer

work page 2017

[2] [2]

Dis- tributed memoryless point convergence algorithm for mobile robots with limited visibility,

H. Ando, Y . Oasa, I. Suzuki, and M. Yamashita, “Dis- tributed memoryless point convergence algorithm for mobile robots with limited visibility,”IEEE Transac- tions on Robotics and Automation, vol. 15, no. 5, pp. 818–828, 1999, IEEE

work page 1999

[3] [3]

Stability analysis of swarms,

V . Gazi and K. M. Passino, “Stability analysis of swarms,” IEEE Transactions on Automatic Control , vol. 48, no. 4, pp. 692–697, 2003, IEEE

work page 2003

[4] [4]

Come to- gether: Multi-agent geometric consensus (gathering, rendezvous, clustering, aggregation),

A. Barel, R. Manor, and A. M. Bruckstein, “Come to- gether: Multi-agent geometric consensus (gathering, rendezvous, clustering, aggregation),” arXiv preprint arXiv:1902.01455, 2019

arXiv 1902

[5] [5]

Dis- tributed & decentralized control and task allocation for flexible swarms,

Y . Koifman, A. Barel, and A. Bruckstein, “Dis- tributed & decentralized control and task allocation for flexible swarms,” inThe 8th International Sympo- sium on Swarm Behavior and Bio-Inspired Robotics, pp. 137–144, 2024

work page 2024

[6] [6]

The signal propagation effects on IEEE 802.15.4 radio link in fire environ- ment,

M. D. Chinthaka, N. H. Malka, K. Ramamohanarao, B. Moran, and P. Farrell, “The signal propagation effects on IEEE 802.15.4 radio link in fire environ- ment,” in 2010 Fifth International Conference on Information and Automation for Sustainability , pp. 411–414, 2010, IEEE

work page 2010

[7] [7]

Radio Propagation in Fire Environ- ments,

J. A. Boan, “Radio Propagation in Fire Environ- ments,” Ph.D. dissertation, University of Adelaide, Australia, 2009

work page 2009

[8] [8]

Eigenvalues of the Laplacian and their relationship to the connectedness of a graph,

A. Marsden, “Eigenvalues of the Laplacian and their relationship to the connectedness of a graph,” Uni- versity of Chicago, REU, 2013

work page 2013