A Learning Framework For Cooperative Collision Avoidance of UAV Swarms Leveraging Domain Knowledge

Haibo Zhang; Shuangyao Huang; Zhiyi Huang

REVIEW 4 major objections 4 minor 30 references

A Learning Framework For Cooperative Collision Avoidance of UAV Swarms Leveraging Domain Knowledge

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read UAV swarms can learn cooperative collision avoidance without credit assignment or observation sharing, because a contour-following reward on a field with obstacle peaks makes each agent's individual maximization safe; the paper reports…

desk verdict The reward-design idea is genuinely clever—turning an active-contour cost into an individual MARL reward to sidestep credit assignment—but the experiments as reported don't establish the claimed gains because of a train/test ambiguity in how the reward is computed plus unfair baselines. read the letter →

arxiv 2507.10913 v1 pith:OWDA7TD4 submitted 2025-07-15 cs.MA cs.LGcs.RO

Shuangyao Huang , Haibo Zhang , Zhiyi Huang This is my paper

classification cs.MAcs.LGcs.RO

keywords multi-agentreinforcementlearningUAVswarmcollisionavoidanceactivecontourmodeldomain-knowledgerewardpotentialfieldDDPGcooperativenavigation

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

reMARL, the framework proposed here, is an attempt to show that cooperative collision avoidance for UAV swarms does not need coordination machinery inside the learning algorithm. The environment is represented as a two-dimensional potential field with obstacles and the swarm's virtual center as peaks, and each drone's reward is built from the active-contour cost that is minimized by smooth curves following the field's edges. Because contours on such a field never pass through peaks and never intersect, a drone that maximizes its own reward is claimed to avoid collisions and fly smoothly. If the claim holds, large swarms can be trained with independent DDPG agents, sidestepping the credit-assignment and observation-sharing bottlenecks that limit existing MARL methods.

What carries the argument

The load-bearing object is the active contour model from image processing, used here as a trajectory scorer rather than an image segmenter. By placing obstacles at field maxima and minimizing $f(S,\Phi)=\int\left(\frac{1}{2}|S''(\rho)|^2-\frac{1}{2}|\nabla\Phi(S(\rho))|^2\right)d\rho$, trajectories are pulled onto smooth contour lines that, by construction, avoid peaks and other contours. The PSO correction of Eq. 8 enforces the hard inter-drone spacing during training, and a per-agent DDPG learner turns the resulting scalar reward into a policy.

What would settle it

Replay a trained episode and recompute the reward from the executed, unadjusted trajectory; if it disagrees materially with the reward stored during training, the policy was optimized against a fictitious trajectory. A direct comparison between agents trained on PSO-adjusted positions and agents trained on the actually executed positions would settle whether the reported improvements are real.

Watch

Extended reading notes

Core claim

The paper's central claim is that the geometric property of contours—they never cross maxima—can be turned into a per-agent reward that makes cooperative collision avoidance an individual learning problem. Formally, the environment is mapped to a potential field $\Phi(q)=\Phi_s(q)+\sum\Phi_o(q)$, with the swarm's virtual center and every obstacle as maxima; a trajectory $S$ is scored by $f(S,\Phi)=\int\left(\frac{1}{2}|S''(\rho)|^2-\frac{1}{2}|\nabla\Phi(S(\rho))|^2\right)d\rho$, the active-contour cost. The reward $r=-f(S,\Phi)+r_{\text{form}}\cdot r_{\text{collide}}$ adds a formation-preserving and safety-gating swarming term, while PSO adjusts the drones' positions during training so their contour levels respect a hard inter-drone distance. With this reward, each drone is trained by DDPG alone, and the reported experiments show reMARL outperforming VDN, QMIX, COMA, IQL, and MAPPO once the swarm exceeds three UAVs, scaling to ten, and cutting reaction time by 98.75% and energy cost by 85.37% relative to contour-based meta-heuristic search.

Load-bearing premise

The reward used for training is computed from positions that PSO has collectively shifted to keep drones apart, and the paper does not state whether those shifted positions are the positions the drone actually flew or whether the correction is applied at deployment.

Editorial extensions

If this is right

Adding UAVs to the swarm only adds independent DDPG learners, so training cost scales linearly rather than exponentially in swarm size.
The fixed-size observation—self state, swarm virtual center, and detected obstacle states—removes communication and centralized critics from the network design.
The smoothness term in the reward bounds steering to $[-\pi/4, \pi/4]$, so actions are stable and energy-efficient by construction.
Because the agent maximizes the full reward, it can deviate from strict contours when contours are not viable, such as flying between two obstacles instead of taking a long detour.
A deployed policy is a single forward pass, giving a reaction time of about 0.006 seconds versus 0.48 seconds for the meta-heuristic baseline.

Reading between the lines

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

The paper leaves open whether the PSO-adjusted positions used to compute the reward match the positions the drone actually executed; if they do not, the learned policy is scored against a trajectory it did not fly, and the reported gains could be inflated.
The same reward construction should transfer to any multi-agent domain where safety can be encoded as 'never cross a peak' on a shared scalar field, such as surface vessels or ground robots navigating on a risk map.
A natural stronger evaluation would verify the learned policy's safety formally, since the hard distance constraints are enforced by PSO during training and are not guaranteed by the neural network at deployment.
The small-swarm regime's weaker performance suggests a broader design lesson: coordination machinery only pays off when the task's geometry makes independent locally optimal behavior unsafe; in trivial geometries it mostly adds noise.

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

The core idea here is actually neat. The authors convert the active-contour cost function from their earlier PSO-based work into an individual reward for independent DDPG learners. That lets them drop credit assignment and observation sharing entirely, and it's a plausible route to larger swarms. They're also honest about a real limit: when contours are non-viable or non-existent, the learned policy can still navigate, which the meta-heuristic version couldn't do.

Where the paper falls down is the evidence. The reward in Eq. 7 is defined after a PSO adjustment of UAV positions (Eq. 8), and the paper never clarifies whether the adjusted position is the one the executed action actually produces. The Framework section says PSO adjusts positions and the reward is then used to train; the Experiments section says the reward is computed on the actual trajectory after the action. If training rewards are computed on hypothetical adjusted positions rather than on the states actually reached, the policy optimizes a different MDP than the one at deployment, and the performance comparison is meaningless. This is a load-bearing ambiguity. There's also a notation slip in Eq. 8: f_thres is written with dU2O, but the surrounding text says the constraint is on the inter-UAV distance dU2U.

The baseline comparison is unfair. VDN, QMIX, COMA, and IQL are restricted to 10 discrete actions while reMARL uses continuous DDPG, so part of the gap in Fig. 4 is exactly what you'd expect. The energy metric in Table 1 is average curvature, which is the same smoothness term being minimized inside the reward, so that comparison is partly circular. And the error bars are huge—Energy Cost is reported as 19.72 ± 31.8, which makes the 85% improvement hard to trust. No code or data is provided.

The abstract overclaims: it says collisions are 'inherently avoided' because contours never pass through peaks, but the body admits contours may be non-viable and the learned policy sometimes departs from them. The safety claim should be scoped to the reward design's intention, not treated as a guarantee.

That said, the underlying idea is worth taking seriously. All the flaws are addressable in a revision: pin down the PSO/reward relationship, rerun baselines with continuous actions, report seeds and interquartile ranges, and separate the energy claim from the reward's smoothness term. I'd send this to a serious referee rather than desk-reject, because the reward-design insight could genuinely help UAV swarm settings if the train/test mismatch is resolved.

Referee Report

4 major / 4 minor

Summary. The paper proposes reMARL, a multi-agent reinforcement learning framework for cooperative collision avoidance of UAV swarms. The core idea is to construct a potential field in which obstacles and the swarm virtual center are peaks, then define a reward from active-contour theory so that each agent's trajectory approximates a contour on the field. Training uses independent DDPG agents that observe only self, swarm, and obstacle properties, with PSO used during training to adjust UAV positions collectively so that contour distances satisfy inter-UAV thresholds. The paper claims that this reward removes the need for credit assignment or observation sharing, enables training with swarm sizes up to 10, and outperforms VDN, QMIX, COMA, IQL, and MAPPO, while also achieving large reaction-time and energy-efficiency improvements over a meta-heuristic baseline. The empirical evidence, however, is undermined by a train/test mismatch in how the reward is computed, an asymmetric comparison against discretized baselines, and an energy metric that partly duplicates the training objective.

Significance. If the central claims were valid, the paper would make a useful contribution: replacing complex MARL architectures with a domain-knowledge-derived individual reward and a simple DDPG learner is an appealing recipe for scalable UAV swarm training, and the observation design that avoids explicit observation sharing is interesting. The paper also addresses a practically relevant problem and tests multiple swarm/obstacle configurations. However, the current evidence does not support the headline claims. The reward is computed on positions that have been collectively adjusted by PSO during training but not necessarily at deployment, so the learned policy may optimize a different MDP than the one evaluated. The baseline comparison is also unfair because the competing MARL methods are restricted to 10 discrete actions while reMARL uses continuous DDPG. Finally, the energy-efficiency metric is essentially the smoothness term that is already minimized inside the training reward, making the energy claim circular. These issues are load-bearing for the paper's main assertions, so the contribution cannot be accepted in its present form.

major comments (4)

[Framework / Innovative Reward Design (Eqs. 7-8)] The paper states that PSO adjusts UAV positions collectively and that this search is "only needed in training" (Framework), while the reward is computed on the adjusted positions q' via Eq. (8). If at deployment the PSO correction is not applied, the policy is trained on reward signals from adjusted trajectories that never occur under the executed action distribution. The Q-function and policy gradient therefore optimize a different MDP from the one deployed, so the reported gains over MARL baselines are not established. The authors should either apply the same PSO correction at deployment, or train and report using rewards evaluated on the positions actually produced by the agent's action.
[Results] The comparison is asymmetric: reMARL uses DDPG with continuous actions, whereas VDN, QMIX, COMA, IQL, and MAPPO are restricted to 10 discrete actions sampled uniformly from [-π/4, π/4] for algorithms requiring a discrete action space. This confounds algorithm choice with action-space granularity; MAPPO and COMA can be implemented with continuous policies. The claim that reMARL outperforms state-of-the-art MARL therefore requires re-running baselines with comparable continuous action spaces or explicitly controlling for discretization error.
[Results (Table 1, Eq. 10)] The energy-efficiency metric is average curvature defined by Eq. (10), which is essentially the same smoothness term minimized inside the contour reward Eq. (4) (|S''|^2 vs. |S''|). Hence the reported 85.37% energy improvement over the meta-heuristic is partly a reflection of the training objective rather than an independent measure of energy consumption. An independent metric (e.g., simulated power draw from acceleration and turning) is needed to support the energy claim.
[Innovative Reward Design (Eq. 8) and Abstract] The abstract's claim that collisions are "inherently avoided" because contours never pass through peaks or intersect is not supported by the paper's own admission in Results that agents adapt to environments "where contours may be non-viable or non-existent" and navigate between obstacles. Safety in the actual experiments is enforced by the Swarming reward r_collide and by the threshold d_col, not by contour topology alone. Additionally, Eq. (8) defines f_thres using dU2O where the surrounding text says the hard constraint applies to the distance between any two UAVs (dU2U); the notation should be corrected and the actual constraint specified.

minor comments (4)

[Abstract and Title] There are multiple typos and spacing issues: "A voidance" in the title, "counters" should be "contours" in the abstract, and "UA V" is inconsistently spaced throughout the manuscript.
[Conclusion] The conclusion says "Extensive experiments are conducted to proof that reMARL achieves..." which should be "prove" or "demonstrate."
[Results (Fig. 4)] The learning curves in Fig. 4 appear to be single runs without shaded confidence intervals or multiple-seed statistics, making it difficult to judge whether the reported differences are significant; a table of final mean and standard deviation across seeds would strengthen the empirical claims.
[Agent] The policy network outputs a single tanh unit, but the action is defined in the range [-π/4, π/4]; the scaling from the network output to the actual action range is not described, which hampers reproducibility.

Circularity Check

1 steps flagged · score 4.0 of 10

Energy-efficiency metric duplicates the reward's smoothness term; main MARL comparison is external.

self definitional [Sections 'Innovative Reward Design' (Eqs. 4, 7) and 'Experiments' (Eq. 10, Table 1)]
"f (S(ρ)) = ∫ 1/2 |S''(ρ)|² − 1/2 |▽Φ(S(ρ))|²dρ ... We use the average curvature of trajectories to evaluate the UA V's energy costs, which is defined in Eq. 10. En = ∫ | S''(p) | dp"

The Contour part of the reward is r = −f(S,Φ), and f contains the smoothness term ∫(1/2)|S''|². Maximizing r therefore directly minimizes trajectory curvature. The paper's 'Energy Cost' metric, Eq. 10, is En = ∫|S''|dp — the same curvature functional (modulo squaring and factor). The 85.37% energy improvement in Table 1 is thus a measurement of the very quantity the reward is engineered to minimize, not an independent estimate of power or work. The energy-efficiency conclusion is true by construction with respect to this metric rather than being an externally validated prediction.

full rationale

The core comparison against VDN, QMIX, COMA, IQL and MAPPO is external and not circular: baselines are standard algorithms trained with the Swarming reward while reMARL uses the full reward, and the learning curves are reported directly. No fitted parameter is relabeled as a prediction, and no load-bearing uniqueness theorem is imported from the authors' prior work; the active-contour energy is traced to Kass et al. and the prior E2Coop/E2CoPre citations are background. The one genuine circular element is the energy-efficiency evaluation: Eq. 10 defines energy cost as average curvature, which is the smoothness term already minimized inside the reward of Eq. 4, so Table 1's energy claim is partly self-verifying. Concerns about PSO-adjusted positions being used in training but not at deployment, and about contours being non-viable in some environments, are train/test mismatch and overclaiming issues rather than circularity; they affect validity but do not raise the circularity score further.

Assumptions & free parameters 6 free parameters · 6 assumptions · 1 invented entities

The framework rests on the potential-field model from the authors' prior work, on hand-chosen safety and influence parameters, and on an opaque PSO-based reward adjustment. The virtual center is a modeling device. No first-principles derivation is attempted.

free parameters (6)

d_safe = not given
Minimum safe distance in Eq. 1, chosen by hand; defines the plateau of the obstacle repulsive field.
R_o = not given
Influential range of the obstacle repulsive field in Eq. 1.
R_s = not given
Influential range of the swarm repulsive field in Eq. 2.
d_col = 20 (Table 1)
Collision threshold used for the r_collide reward and safety evaluation.
PSO coefficients (mu, c1, c2) = mu uniform in [0,1], c1/c2 unspecified
Coefficients in Eq. 9 for the PSO contour adjustment.
reward combination weights = 1 and 1
Weights of Contour and Swarming terms in Eq. 7 are fixed at 1 with no tuning or ablation reported.

assumptions (6)

domain assumption Obstacles and the swarm virtual center are placed at field peaks; contours never pass through peaks.
Introduced in the Domain Knowledge section; used to argue safety and smoothness of contour-following trajectories.
domain assumption UAV velocities have constant magnitude during avoidance; only direction changes in [-pi/4, pi/4].
Application Scenario section; justifies the action space and fixed trajectory length in Eq. 4.
ad hoc to paper The reward f(S, Phi) evaluated on a combined trajectory S1 is a valid Markovian reward for the agent.
Reward is computed by combining current and previous trajectory segments, creating dependence on history not present in the observation; the paper does not address this non-Markovianity.
ad hoc to paper PSO-based collective position adjustment (Eq. 8) produces a reward signal that correctly reflects the executed policy.
The paper states PSO adjusts UAV positions/contours during training, but does not specify how the adjusted positions relate to actual executed actions or how the reward is assigned.
domain assumption The potential field is fully determined by the swarm virtual center and obstacle positions and velocities, so the specified observation is sufficient for cooperation.
Minimum Requirements on Observation Design; assumes that cooperative behavior can be inferred from virtual center and obstacle properties alone.
domain assumption The contour-following trajectories are smooth and energy-efficient, justifying Eq. 10 as an energy cost.
Used in Table 1 to equate average curvature with energy cost; no physical energy model is given.

invented entities (1)

virtual center p*
purpose: A conceptual point ahead of the swarm used to construct the swarm repulsive field Phi_s.
No independent falsifiable handle; it is a modeling construct assumed to move with constant initial velocity throughout the episode.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Learning Framework For Cooperative Collision Avoidance of UAV Swarms Leveraging Domain Knowledge." pith.science (2026). https://pith.science/paper/OWDA7TD4

@misc{pith2026250710913,
  author       = {Pith},
  title        = {Pith review of: A Learning Framework For Cooperative Collision Avoidance of UAV Swarms Leveraging Domain Knowledge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OWDA7TD4}},
  note         = {Machine review of arXiv:2507.10913}
}

read the original abstract

This paper presents a multi-agent reinforcement learning (MARL) framework for cooperative collision avoidance of UAV swarms leveraging domain knowledge-driven reward. The reward is derived from knowledge in the domain of image processing, approximating contours on a two-dimensional field. By modeling obstacles as maxima on the field, collisions are inherently avoided as contours never go through peaks or intersect. Additionally, counters are smooth and energy-efficient. Our framework enables training with large swarm sizes as the agent interaction is minimized and the need for complex credit assignment schemes or observation sharing mechanisms in state-of-the-art MARL approaches are eliminated. Moreover, UAVs obtain the ability to adapt to complex environments where contours may be non-viable or non-existent through intensive training. Extensive experiments are conducted to evaluate the performances of our framework against state-of-the-art MARL algorithms.

Figures

Figures reproduced from arXiv: 2507.10913 by the authors.

**Figure 2.** Potential field Φ mapped from the environment and its gradient amplitudes. (a) (b) [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Illustration of the environment. PSO simply by minimizing the following cost function. fpso = fthres + fshif t, (8) where fthres = dU2O, dU2O ≥ ¯dU2O, Inf, otherwise, and fshif t = maxi∈[1,N] ||qi , q′ i ||, where qi and q ′ i are the position of UAV i before and after adjustment, respectively. N is the number of UAVs in the swarm. Minimizing fthres ensures that the minimum distance between any two UAVs remains ei… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Learning curves of reMARL in comparison with state-of-the-art MARL algorithms. UAV and an obstacle or between any two UAVs is smaller than a threshold dcol. Fig. 3b illustrates how the trajectories are generated from UAV’s actions. In Fig. 3b, contours are plotted as b…

**Figure 5.** Figure 5: Trajectories and action values. Reaction Time (Sec) Energy Cost min. dU2O min. dU2U Meta-Heuristic 0.48 ± 0.05 134.88 ± 8.3 38.70 40.31 reMARL 0.006 ± 0.03 19.72 ± 31.8 24.78 29.45 Improvement 98.75% 85.37% 35.96% 26.94% [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

30 extracted references · 24 canonical work pages

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Cai, Y.; and Yang, S. X. 2014. A potential field-based PSO approach for cooperative target searching of multi-robots. In Proceeding of the 11th World Congress on Intelligent Control and Automation, 1029--1034. IEEE

work page 2014
[4]

Foerster, J.; Farquhar, G.; Afouras, T.; Nardelli, N.; and Whiteson, S. 2018. Counterfactual multi-agent policy gradients. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32

work page 2018
[5]

L.; Dadios, E

Galvez, R. L.; Dadios, E. P.; and Bandala, A. A. 2014. Path planning for quadrotor UAV using genetic algorithm. In International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management (HNICEM), 1--6. IEEE

work page 2014
[6]

Gao, J.; Zhang, H.; Tan, L.; and Ren, X. 2022. UAV dynamic obstacle avoidance based on improved reciprocal velocity obstacle. In Journal of Physics: Conference Series, volume 2216, 012014. IOP Publishing

work page 2022
[7]

Huang, S.; and Low, K. 2018. A Path Planning Algorithm for Smooth Trajectories of Unmanned Aerial Vehicles via Potential Fields. In International Conference on Control, Automation, Robotics and Vision (ICARCV), 1677--1684. IEEE

work page 2018
[8]

Huang, S.; Zhang, H.; and Huang, Z. 2021. E^2Coop : Energy efficient and cooperative obstacle detection and avoidance for uav swarms. In Proceedings of the International Conference on Automated Planning and Scheduling, volume 31, 634--642

work page 2021

Show all 30 references

[9]

Huang, S.; Zhang, H.; and Huang, Z. 2024 a . CoDe: A Cooperative and Decentralized Collision Avoidance Algorithm for Small-Scale UAV Swarms Considering Energy Efficiency. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 13152--13159. IEEE

2024
[10]

Huang, S.; Zhang, H.; and Huang, Z. 2024 b . E^2CoPre : Energy Efficient and Cooperative Collision Avoidance for UAV Swarms With Trajectory Prediction. IEEE transactions on intelligent transportation systems, 25(7): 6951--6963

2024
[11]

Kass, M.; Witkin, A.; and Terzopoulos, D. 1988. Snakes: Active contour models. International journal of computer vision, 1(4): 321--331

1988
[12]

Koenig, S.; and Likhachev, M. 2005. Fast replanning for navigation in unknown terrain. Transactions on Robotics, 21(3): 354--363

2005
[13]

Konda, V.; and Tsitsiklis, J. 1999. Actor-critic algorithms. Advances in neural information processing systems, 12

1999
[14]

Liu, Y.; Chen, C.; Wang, Y.; Zhang, T.; and Gong, Y. 2024. A fast formation obstacle avoidance algorithm for clustered UAVs based on artificial potential field. Aerospace Science and Technology, 147: 108974

2024
[15]

Lombard, A.; Durand, L.; and Galland, S. 2020. Velocity obstacle based strategy for multi-agent collision avoidance of unmanned aerial vehicles. In 2020 IEEE International Conference on Sensing, Communication and Networking (SECON Workshops), 1--6. IEEE

2020
[16]

Lowe, R.; Wu, Y.; Tamar, A.; Harb, J.; Abbeel, P.; and Mordatch, I. 2017. Multi-agent actor-critic for mixed cooperative-competitive environments. arXiv preprint arXiv:1706.02275

2017 arXiv
[17]

Mete, A.; Mouhoub, M.; and Farid, A. M. 2023. Coordinated Multi-Robot Exploration using Reinforcement Learning. In International Conference on Unmanned Aircraft Systems (ICUAS), 265--272. IEEE

2023
[18]

S.; Jayaram, S.; and Ganapathisubramaniyan, A

Raja, G.; Anbalagan, S.; Narayanan, V. S.; Jayaram, S.; and Ganapathisubramaniyan, A. 2019. Inter-UAV collision avoidance using Deep-Q-learning in flocking environment. In Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON), 1089--1095. IEEE

2019
[19]

S.; Farquhar, G.; Foerster, J.; and Whiteson, S

Rashid, T.; Samvelyan, M.; de Witt, C. S.; Farquhar, G.; Foerster, J.; and Whiteson, S. 2018. QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning. arXiv:1803.11485

2018 arXiv
[20]

M.; Zambaldi, V.; Jaderberg, M.; Lanctot, M.; Sonnerat, N.; Leibo, J

Sunehag, P.; Lever, G.; Gruslys, A.; Czarnecki, W. M.; Zambaldi, V.; Jaderberg, M.; Lanctot, M.; Sonnerat, N.; Leibo, J. Z.; Tuyls, K.; et al. 2017. Value-decomposition networks for cooperative multi-agent learning. arXiv preprint arXiv:1706.05296

2017 arXiv
[21]

Tan, M. 1993. Multi-agent reinforcement learning: Independent vs. cooperative agents. In Proceedings of the international conference on machine learning, 330--337

1993
[22]

Tang, J.; Swamy, G.; Fang, F.; and Wu, S. Z. 2024. Multi-agent imitation learning: Value is easy, regret is hard. Advances in Neural Information Processing Systems, 37: 27790--27816

2024
[23]

Van den Berg, J.; Lin, M.; and Manocha, D. 2008. Reciprocal velocity obstacles for real-time multi-agent navigation. In International conference on robotics and automation, 1928--1935. Ieee

2008
[24]

Van Hasselt, H.; Guez, A.; and Silver, D. 2016. Deep reinforcement learning with double q-learning. In Proceedings of the AAAI conference on artificial intelligence, volume 30

2016
[25]

Viseras, A.; Meissner, M.; and Marchal, J. 2021. Wildfire front monitoring with multiple uavs using deep q-learning. IEEE Access

2021
[26]

Wang, D.; Fan, T.; Han, T.; and Pan, J. 2020. A two-stage reinforcement learning approach for multi-UAV collision avoidance under imperfect sensing. IEEE Robotics and Automation Letters, 5(2): 3098--3105

2020
[27]

Wang, J.; Ren, Z.; Han, B.; Ye, J.; and Zhang, C. 2021. Towards understanding cooperative multi-agent q-learning with value factorization. Advances in Neural Information Processing Systems, 34: 29142--29155

2021
[28]

Wang, X.; Yi, M.; Liu, J.; Zhang, Y.; Wang, M.; and Bai, B. 2023. Cooperative Data Collection with Multiple UAVs for Information Freshness in the Internet of Things. Transactions on Communications

2023
[29]

H.; and Tumer, K

Wolpert, D. H.; and Tumer, K. 2002. Optimal payoff functions for members of collectives. In Modeling complexity in economic and social systems, 355--369. World Scientific

2002
[30]

Zhang, Y.; Wu, Z.; Ma, Y.; Sun, R.; and Xu, Z. 2022. Research on autonomous formation of Multi-UAV based on MADDPG algorithm. In International Conference on Control & Automation (ICCA), 249--254. IEEE

2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Cai, Y.; and Yang, S. X. 2014. A potential field-based PSO approach for cooperative target searching of multi-robots. In Proceeding of the 11th World Congress on Intelligent Control and Automation, 1029--1034. IEEE

work page 2014

[4] [4]

Foerster, J.; Farquhar, G.; Afouras, T.; Nardelli, N.; and Whiteson, S. 2018. Counterfactual multi-agent policy gradients. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32

work page 2018

[5] [5]

L.; Dadios, E

Galvez, R. L.; Dadios, E. P.; and Bandala, A. A. 2014. Path planning for quadrotor UAV using genetic algorithm. In International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management (HNICEM), 1--6. IEEE

work page 2014

[6] [6]

Gao, J.; Zhang, H.; Tan, L.; and Ren, X. 2022. UAV dynamic obstacle avoidance based on improved reciprocal velocity obstacle. In Journal of Physics: Conference Series, volume 2216, 012014. IOP Publishing

work page 2022

[7] [7]

Huang, S.; and Low, K. 2018. A Path Planning Algorithm for Smooth Trajectories of Unmanned Aerial Vehicles via Potential Fields. In International Conference on Control, Automation, Robotics and Vision (ICARCV), 1677--1684. IEEE

work page 2018

[8] [8]

Huang, S.; Zhang, H.; and Huang, Z. 2021. E^2Coop : Energy efficient and cooperative obstacle detection and avoidance for uav swarms. In Proceedings of the International Conference on Automated Planning and Scheduling, volume 31, 634--642

work page 2021