Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

CrazyMARL: Decentralized Direct Motor Control Policies for Cooperative Aerial Transport of Cable-Suspended Payloads

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A decentralized multi-agent RL policy, trained on hybrid slack-taut cable dynamics, outperforms a classical rigid-rod controller for multi-UAV cable-suspended payload transport, recovering 80% of harsh trials versus 44% for the baseline…

desk verdict Genuinely new MARL + direct-PWM + hybrid-cable pipeline with a credible simulation win; the zero-shot transfer claim is real but currently over-sold. read the letter →

arxiv 2509.14126 v2 pith:MST4TE6T submitted 2025-09-17 cs.RO cs.MA

Viktor Lorentz , Khaled Wahba , Sayantan Auddy , Marc Toussaint , Wolfgang Hönig This is my paper · ORCID
classification cs.ROcs.MA
keywords multi-agentreinforcementlearningcable-suspendedpayloaddecentralizedcontrolsim-to-realtransferdomainrandomizationdirectmotorquadrotorswarmhybridcabledynamics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper aims to establish that a team of small quadrotors can learn a fully decentralized control policy that directly outputs motor PWM commands to transport a cable-suspended payload, handling the hybrid dynamics of cables that go slack and taut. The authors report that this learned policy outperforms a classical decentralized controller in simulation, recovering from 80% of harsh randomized trials compared with 44% for the baseline, and tracks a figure-eight trajectory with less error. They also report successful zero-shot transfer of the same policy to real Crazyflie quadrotors, which keep a payload stable under wind and pushes. If correct, this demonstrates that end-to-end multi-agent reinforcement learning can replace cascaded low-level controllers for cooperative aerial manipulation on resource-constrained hardware while respecting cable physics that prior multi-UAV work simplified away.

What carries the argument

The load-bearing identity is the normalized action mapping: each agent emits $a_i^t \in [-1,1]^4$, converted to $u_i^t = (a_i^t + 1)/2 \in [0,1]^4$, which the simulator reads as a thrust fraction of each motor's maximum and the real robot reads as a PWM duty cycle; the paper justifies this by asserting that, in the operating range of the micro brushed motors, thrust is approximately proportional to duty cycle. This same-number-different-physics bridge is what makes zero-shot transfer conceivable. The second piece is the hybrid cable model: each quadrotor is connected to the payload by a tendon of length $L$ that exerts tension only when the distance equals $L$ (taut) and zero otherwise (slack), so the policy learns to handle cable mode transitions rather than assuming a rigid rod. The third is the modular reward $r = r_{\text{track}}\, r_{\text{stable}} + r_{\text{safe}}$, with bounded exponential shaping terms that couple tracking with stability, add safety incentives for smooth and balanced motor commands, and discourage collisions and saturation. Training uses IPPO with shared parameters and a Gaussian action distribution, over 16,384 parallel environments on a single GPU, with domain randomization applied to initial states, motor thrust caps, actuator time constants, observations, and external wrenches.

What would settle it

Measure static thrust versus PWM duty cycle on the Crazyflie motors across the battery voltage range (3.7 to 4.2 V) and at different motor temperatures; if the curve deviates by more than a few percent from a straight line through the origin, the zero-shot transfer claim is undermined. A direct behavioral test: run the reported push-recovery trials with a worn or partially discharged battery; if the real-world recovery rate drops sharply while the simulator (which models thrust as proportional to $u_i^t$) still predicts high recovery, the central transfer claim is falsified.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that a single decentralized policy trained with Independent Proximal Policy Optimization (IPPO) and shared parameters across agents can map each quadrotor's local observation—payload error and velocity, its own pose, rates, and previous action, plus the relative positions of teammates—directly to normalized motor commands at 250 Hz, with no low-level attitude or rate controller in the loop. In simulation, the policy achieves an 80% recovery rate (797 of 1000 trials) for a two-quadrotor cable-suspended payload task starting from harsh randomized states, versus 43.5% (435 of 1000) for a rigid-rod trajectory-optimization baseline tracking a precomputed polynomial; the learned policy also reaches the goal at more than twice the mean speed (0.58 m/s vs 0.27 m/s). The same policy transfers zero-shot to Crazyflie 2.1 hardware, where two quadrotors carry a 10 g payload through autonomous takeoff, figure-eight tracking, and disturbance recovery under wind measured at 3.5 m/s. The paper attributes this to training on hybrid slack-taut cable dynamics, heavy domain randomization of motor thrust caps, actuator lag, observation noise, and external disturbances, and a reward that couples tracking with stability and safety.

Load-bearing premise

The zero-shot transfer rests on the assumption that thrust is approximately proportional to PWM duty cycle for the Crazyflie's motors, so the policy's normalized command $u_i^t$ means the same physical thing in simulation and on hardware; if battery voltage, motor wear, or nonlinearity breaks that proportionality, the transferred policy may fail.

Editorial extensions

If this is right

  • Multi-UAV teams can transport suspended payloads through slack-taut cable transitions, enabling maneuvers (e.g., recovery from dropped payloads, wide payload swings) that rigid-rod controllers cannot represent.
  • Direct PWM control at 250 Hz removes the need for cascaded attitude/rate controllers and thrust mixing, allowing the same policy to operate near motor saturation on platforms with low thrust-to-weight ratio.
  • Because the policy is decentralized and communication-free, teams scale without a central coordinator or inter-robot bandwidth, and the failure of one drone does not collapse the whole control loop.
  • Domain randomization on motor thrust caps, actuator lag, observation noise, and external disturbances appears sufficient for zero-shot sim-to-real transfer of low-level multi-robot policies, reducing the need for hardware-in-the-loop tuning.
  • Scaling to larger teams is currently the main limitation: recovery success falls from 81% for two quadrotors to 60% for three and near zero for six, and the paper attributes this to permutation sensitivity in how peer observations are ordered, pointing to order-invariant encodings as the next step.

Reading between the lines

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

  • If the thrust-to-duty-cycle proportionality degrades under battery sag or motor heating, the zero-shot transfer may not survive long flights; the paper's domain randomization over motor thrust caps does not explicitly randomize the proportionality slope itself, leaving a voltage-dependent nonlinearity as an untested failure mode.
  • The headline comparison against a rigid-rod baseline is favorable to the learned method because the baseline cannot even represent slack-taut transitions; a slack-aware model-based controller might narrow the gap, though none exists for multi-UAV teams, which is precisely the gap this paper fills.
  • The modular reward and observation structure could be reparameterized for other payload types (rigid links, nets, or multiple suspended loads), suggesting the framework is a general template for decentralized aerial manipulation rather than a single-task solution.
  • A testable extension: train the same observation and reward on teams of four to eight quadrotors with attention-based or sorted peer encodings; if recovery success degrades gracefully with team size, the permutation-sensitivity diagnosis is confirmed.
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.

Referee Report

4 major / 4 minor

Summary. The paper proposes CrazyMARL, a decentralized multi-agent reinforcement learning framework for cooperative aerial transport of cable-suspended payloads by multiple quadrotors. The policy is trained with IPPO and shared parameters in a GPU-parallelized MJX simulator, uses a hybrid slack-taut cable model, and outputs direct PWM motor commands at 250 Hz without a low-level cascade. The authors report simulation results with 79.7% recovery from harsh initial conditions versus 43.5% for a baseline, qualitative robustness tests under wind and pushes, and zero-shot sim-to-real transfer on Crazyflie 2.1 hardware. The paper also presents generalization sweeps and scalability results for Q=1,2,3,6 quadrotors, and releases code and videos.

Significance. If the claims hold, this is a meaningful step for decentralized motor-level control of multi-UAV cable-suspended payload transport, especially because it explicitly models slack-taut cable transitions, which many prior rigid-link approaches avoid. The high-throughput training pipeline (2e9 environment steps in 70 minutes on a single GPU), the fully decentralized execution, and the public release of code and videos are concrete strengths that support reproducibility. The real-hardware demonstration is valuable even if qualitative. However, the central performance claims currently rest on a single, mismatched baseline and an unvalidated thrust-to-PWM mapping in the sim-to-real chain, so the significance is conditional on fixing these points.

major comments (4)
  1. [Abstract and Section IV.A] The abstract claims that the learned policies outperform 'classical decentralized controllers,' but the only baseline used in Section IV.A is the centralized trajectory-optimization method of [7], which the text itself describes as relying on centralized planning with an online tracker and a rigid-rod cable model. A centralized planner is not a decentralized controller, so the experiments do not support the abstract's comparative claim against decentralized methods. Please either add a decentralized model-based baseline (e.g., a decentralized NMPC from [12]) or revise the abstract and conclusion to state that the comparison is against a centralized rigid-rod baseline.
  2. [Section III.C and Section IV.D] The zero-shot sim-to-real claim depends on the assertion in Section III.C that 'in the operating range of the micro brushed motors, thrust is approximately proportional to duty cycle,' combined with interpreting the same normalized action as a thrust command in simulation and as a PWM duty cycle on hardware. No static thrust measurements, deadband characterization, or battery-voltage compensation are reported, and the domain randomization covers per-motor thrust caps and lag time constants but not a nonlinear thrust-to-PWM relationship. Since the paper emphasizes operation near actuation limits, this mapping error could be largest exactly in the regimes of interest. The hardware demonstrations are qualitative and therefore do not quantify any degradation from this mapping. Please add static thrust calibration data or battery-voltage compensation, and report quantitative real-world tracking or recovery metrics.
  3. [Section IV.A and Appendix A] The central quantitative comparison reports success rates (797/1000 versus 435/1000) but the text does not define the success criterion beyond 'recover within 10 s.' The reward function in Appendix A includes terms for tracking error, stability, taut cables, and safe spacing, which overlap heavily with the likely definition of recovery, so the evaluation metric must be stated independently and precisely (e.g., threshold on payload position error, settling time, and cable-tautness conditions). Without a clear, reward-independent success definition, the 79.7% versus 43.5% result is difficult to interpret.
  4. [Section IV.B] The generalization claim in Section IV.B is based on sweeps over cable length, payload mass, observation noise, and seed, but the text does not report confidence intervals or trial-level variance for the success rates in Figure 4, even though each point is described as 1000 runs. For a stochastic evaluation, reporting only point estimates makes it impossible to judge whether the observed differences (e.g., the slight drop for lighter payloads) are meaningful. Please add error bars or confidence intervals.
minor comments (4)
  1. [Section III.F] Equation (8) and the appendix define the reward components, but the roles of the scalar constants λyaw, λup, λs, and vmax are described only as 'tuning' parameters; a short sensitivity discussion would help the reader understand how robust the reported results are to these choices.
  2. [Section IV.C] In the scalability experiment, the statement that Q=6 'exposes a coordination limit' is useful, but the text does not report the failure modes quantitatively (e.g., fraction of trials ending in collision versus timeout). A breakdown would make the scalability claim more informative.
  3. [Section IV.D] The hardware section states that observations use motion capture for payload and teammate positions. Since the paper emphasizes decentralized onboard execution, it would be helpful to clarify whether the motion-capture data are streamed to each robot (and thus a form of external infrastructure) or whether the policy would run with onboard perception only.
  4. [General] There are a few formatting artifacts in the text (e.g., 'V iktor' in the author list and the unnumbered equation in Section III.C), and the text sometimes switches between 'UAV' and 'UA V' spacing. These should be cleaned up.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical RL claims rest on rollouts and hardware tests, not on a derivation that reduces to its inputs.

full rationale

CrazyMARL is an empirical reinforcement-learning paper, not a first-principles derivation. The central claims—80% vs 44% recovery, robustness sweeps, and zero-shot sim-to-real transfer—are established by simulation rollouts and hardware flights, not by an equation chain that could be equivalent to its inputs. The reward does include tracking and taut-cable incentives similar to evaluation metrics, but that is the standard objective–metric relationship in RL, not a fitted input renamed as a prediction; the success metric (recovery within 10 s) is not identical to any single reward term. The baseline [7] is from the same group (Wahba, Toussaint, Hönig) and [35] is a self-citation for the actuator lag model, but neither is used as an unverified load-bearing theorem: the baseline is actually executed, and the actuator model is an explicitly stated approximation later tested on hardware. The Section III.C claim that thrust is approximately proportional to PWM duty cycle is an empirical assumption and a genuine sim-to-real risk, but it is not circular—it does not make the transfer claim true by definition. No prediction reduces by construction to its inputs, and no self-citation chain forces the result.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim depends on several modeling and deployment assumptions rather than on a derivation: the MuJoCo tendon model must capture slack-taut dynamics; the PWM-to-thrust mapping must be near-linear over the operating range; the domain randomization ranges must cover real hardware variation; and hardware observations require motion capture for payload and teammate positions. The reward and DR settings contain many hand-chosen constants. No new physical entities are introduced.

free parameters (3)
  • Reward shaping constants = s=2, c_f=0.02, c_g=40, c_s=2, c_exp=8, c_swing=0.75, c_coll=c_oob=10, c_b=50, d_min=0.15, d_safe=0.18, lambda_yaw=10…
    Hand-tuned in Appendix A; the learned behavior and the evaluation success depend on these choices, though they do not by themselves determine the result.
  • Domain randomization ranges = motor thrust base U(0.105,0.15) N, motor offset N(0,0.008^2) N, actuator lag tau U(0.004,0.05) s, disturbance forces…
    Chosen by the authors in Section III.E to bridge sim-to-real; if actual hardware variation falls outside these ranges, the zero-shot transfer claim weakens.
  • Training hyperparameters and architecture = learning rate 4e-4, entropy coefficient 0.01, clip 0.2, N=16,384 environments, actor MLP [64,64,64], critic MLP…
    Selected by Bayesian optimization over a fixed compute budget in Section III.G; part of the empirical recipe rather than a derived quantity.
assumptions (5)
  • domain assumption MuJoCo tendon model, with tension only when the cable is taut, faithfully represents the slack-taut cable dynamics of the real cable-payload system.
    Section III.D uses a tendon of length L that exerts tension only when taut; the robustness and sim-to-real claims depend on this model fidelity.
  • domain assumption Thrust is approximately proportional to PWM duty cycle over the Crazyflie operating range, so normalized thrust in simulation equals normalized PWM on hardware.
    Section III.C relies on this proportionality for direct PWM mapping; no calibration data is provided.
  • domain assumption Domain randomization ranges cover the real-world distribution of motor, actuator, and disturbance parameters.
    Section III.E defines the DR intervals; the zero-shot sim-to-real claim depends on these intervals containing actual Crazyflie variability.
  • domain assumption External motion capture provides accurate payload and teammate positions during hardware deployment, and these are part of the local observation.
    Section IV.D describes deployment using measured positions from motion capture; the fully decentralized and unstructured-environment framing assumes this infrastructure is available.
  • domain assumption Rigid-body dynamics, contacts, and actuator models in MuJoCo correctly simulate the physical system.
    MuJoCo is treated as the ground-truth simulator throughout training and evaluation; discrepancies from real physics are only partially addressed by domain randomization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CrazyMARL: Decentralized Direct Motor Control Policies for Cooperative Aerial Transport of Cable-Suspended Payloads." pith.science (2026). https://pith.science/paper/MST4TE6T

@misc{pith2026250914126,
  author       = {Pith},
  title        = {Pith review of: CrazyMARL: Decentralized Direct Motor Control Policies for Cooperative Aerial Transport of Cable-Suspended Payloads},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MST4TE6T}},
  note         = {Machine review of arXiv:2509.14126}
}
read the original abstract

Collaborative transportation of cable-suspended payloads by teams of UAVs has the potential to enhance payload capacity, adapt to different payload shapes, and provide built-in compliance, making it attractive for applications ranging from disaster relief to precision logistics. However, multi-UAV coordination under disturbances, nonlinear payload dynamics, and slack-taut cable modes remains a challenging control problem. To our knowledge, no prior work has addressed these cable mode transitions in the multi-UAV context, instead relying on simplifying rigid-link assumptions. We propose CrazyMARL, a decentralized RL framework for multi-UAV cable-suspended payload transport. Simulation results demonstrate that the learned policies can outperform classical decentralized controllers in terms of disturbance rejection and tracking precision, achieving an 80% recovery rate from harsh conditions compared to 44% for the baseline method. We also achieve successful zero-shot sim-to-real transfer and demonstrate that our policies are highly robust under harsh conditions, including wind, random external disturbances, and transitions between slack and taut cable dynamics. This work paves the way for autonomous, resilient UAV teams capable of executing complex payload missions in unstructured environments. Code and videos can be found on the website: https://imrclab.github.io/CrazyMARL.

Figures

Figures reproduced from arXiv: 2509.14126 by the authors.

Figure 1
Figure 1. Overview of our approach. (Left panel) Training in Mujoco with end-to-end JAX and domain randomization. IPPO with shared parameters j [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. 50 randomized initial states for s0 ∼ρ0; harsh cases (slack cables, ground starts) are included. The target is at the center. the top-right shows one state in MuJoCo. Initial states: The payload is sampled around a nominal target, and quadrotors are placed on a spherical shell clipped by cable length, with randomized attitudes and linear and angular velocities. Challenging cases, such as ground starts and slack cabl… view at source ↗
Figure 4
Figure 4. Evaluation of the learned policy’s generalization capabilities in [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (2 more)
Figure 3
Figure 3. Figure 3: Example recovery trajectories from eight harsh initializations. Top: [PITH_FULL_IMAGE:figures/full_fig_p006_3.png]
Figure 6
Figure 6. Figure 6: Two quadrotors with RL policy cooperating to ensure the payload [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Autonomous Aerial Manipulation via Contextual Contrastive Meta Reinforcement Learning

    cs.LG 2026-06 unverdicted novelty 5.0 of 10

    Aco2 trains a quadrotor policy in simulation that adapts to diverse payload dynamics via latent context encoding and contrastive structuring, enabling zero-shot real-world deployment for autonomous aerial delivery.

Reference graph

Works this paper leans on

36 extracted references · 22 canonical work pages · cited by 1 Pith paper

  1. [7]

    Kinodynamic motion planning for a team of multirotors transporting a cable- suspended payload in cluttered environments,

    K. Wahba, J. Ortiz-Haro, M. Toussaint, and W. H ¨onig, “Kinodynamic motion planning for a team of multirotors transporting a cable- suspended payload in cluttered environments,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 12 750–12 757

    work page 2024

  2. [12]

    Distributed nmpc for cooperative aerial manipulation of cable-suspended loads,

    N. De Carli, R. Belletti, E. Buzzurro, A. Testa, G. Notarstefano, and M. Tognon, “Distributed nmpc for cooperative aerial manipulation of cable-suspended loads,”IEEE Robotics and Automation Letters, 2025

    work page 2025

  3. [1]

    A review of quadrotor unmanned aerial vehicles: applications, architectural design and control algorithms,

    M. Idrissi, M. Salami, and F. Annaz, “A review of quadrotor unmanned aerial vehicles: applications, architectural design and control algorithms,” Journal of Intelligent & Robotic Systems, vol. 104, no. 2, p. 22, 2022

    work page 2022

  4. [2]

    Unmanned aerial vehicles for search and rescue: A survey,

    M. Lyu, Y . Zhao, C. Huang, and H. Huang, “Unmanned aerial vehicles for search and rescue: A survey,”Remote Sensing, vol. 15, no. 13, p. 3266, 2023

    2023

  5. [3]

    Decentralized control of quadrotor swarms with end-to- end deep reinforcement learning,

    S. Batra, Z. Huang, A. Petrenko, T. Kumar, A. Molchanov, and G. S. Sukhatme, “Decentralized control of quadrotor swarms with end-to- end deep reinforcement learning,” inConference on Robot Learning. PMLR, 2022, pp. 576–586

    work page 2022

  6. [4]

    Review of aerial transportation of suspended-cable payloads with quadrotors

    J. Estevez, G. Garate, J. Lopez-Guede, and M. Larrea, “Review of aerial transportation of suspended-cable payloads with quadrotors.” 2024

    work page 2024

  7. [5]

    Champion-level drone racing using deep reinforcement learning,

    E. Kaufmann, L. Bauersfeld, A. Loquercio, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Champion-level drone racing using deep reinforcement learning,”Nature, vol. 620, no. 7976, pp. 982–987, 2023

    2023

  8. [6]

    Learning to fly in seconds,

    J. Eschmann, D. Albani, and G. Loianno, “Learning to fly in seconds,” IEEE Robotics and Automation Letters, vol. 9, no. 7, pp. 6336–6343, 2024

    2024

Show all 36 references
  1. [8]

    Collision avoidance and navigation for a quadrotor swarm using end-to-end deep reinforcement learning,

    Z. Huang, Z. Yang, R. Krupani, B. S ¸enbas ¸lar, S. Batra, and G. S. Sukhatme, “Collision avoidance and navigation for a quadrotor swarm using end-to-end deep reinforcement learning,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 300–306

    2024

  2. [9]

    Dynamics, control and planning for cooperative manipulation of payloads suspended by cables from multiple quadrotor robots,

    K. Sreenath and V . Kumar, “Dynamics, control and planning for cooperative manipulation of payloads suspended by cables from multiple quadrotor robots,” inRobotics: Science and Systems IX, Berlin, Germany, P. Newman, D. Fox, and D. Hsu, Eds., 2013

    2013

  3. [10]

    Aerial co- manipulation with cables: The role of internal force for equilibria, stability, and passivity,

    M. Tognon, C. Gabellieri, L. Pallottino, and A. Franchi, “Aerial co- manipulation with cables: The role of internal force for equilibria, stability, and passivity,”IEEE Robotics and Automation Letters, vol. 3, no. 3, pp. 2577–2583, 2018

    2018

  4. [11]

    Nonlinear mpc for full-pose manipulation of a cable-suspended load using multiple uavs,

    S. Sun and A. Franchi, “Nonlinear mpc for full-pose manipulation of a cable-suspended load using multiple uavs,”arXiv preprint arXiv:2301.08545, 2023

    2023 arXiv

  5. [13]

    pc-dbcbs: Kinodynamic motion planning of physically-coupled robot teams,

    K. Wahba and W. H ¨onig, “pc-dbcbs: Kinodynamic motion planning of physically-coupled robot teams,”arXiv preprint arXiv:2505.10355, 2025

    2025 arXiv

  6. [14]

    Safe and agile transportation of cable-suspended payload via multiple aerial robots,

    Y . Wang, J. Wang, X. Zhou, T. Yang, C. Xu, and F. Gao, “Safe and agile transportation of cable-suspended payload via multiple aerial robots,”arXiv preprint arXiv:2501.15272, 2025

    2025 arXiv

  7. [15]

    Agile and cooperative aerial manipulation of a cable-suspended load,

    S. Sun, X. Wang, D. Sanalitro, A. Franchi, M. Tognon, and J. Alonso- Mora, “Agile and cooperative aerial manipulation of a cable-suspended load,”arXiv preprint arXiv:2501.18802, 2025

    2025

  8. [16]

    Impact-aware planning and control for aerial robots with suspended payloads,

    H. Wang, H. Li, B. Zhou, F. Gao, and S. Shen, “Impact-aware planning and control for aerial robots with suspended payloads,”IEEE Transactions on Robotics, vol. 40, pp. 2478–2497, 2024

    2024

  9. [17]

    Es-hpc-mpc: Exponentially stable hybrid perception constrained mpc for quadrotor with suspended payloads,

    L. F. Recalde, M. Sarvaiya, G. Loianno, and G. Li, “Es-hpc-mpc: Exponentially stable hybrid perception constrained mpc for quadrotor with suspended payloads,”arXiv preprint arXiv:2504.08841, 2025

    2025

  10. [18]

    Reinforcement learning for uav attitude control,

    W. Koch, R. Mancuso, R. West, and A. Bestavros, “Reinforcement learning for uav attitude control,”ACM Transactions on Cyber-Physical Systems, vol. 3, no. 2, pp. 1–21, 2019

    2019

  11. [19]

    Reaching the limit in autonomous racing: Optimal control versus reinforcement learning,

    Y . Song, A. Romero, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Reaching the limit in autonomous racing: Optimal control versus reinforcement learning,”Science Robotics, vol. 8, no. 82, p. eadg1462, 2023

    2023

  12. [20]

    Multi-task reinforcement learning for quadrotors,

    J. Xing, I. Geles, Y . Song, E. Aljalbout, and D. Scaramuzza, “Multi-task reinforcement learning for quadrotors,”IEEE Robotics and Automation Letters, 2024

    2024

  13. [21]

    A new nonlinear control strat- egy embedded with reinforcement learning for a multirotor transporting a suspended payload,

    H. Hua, Y . Fang, X. Zhang, and C. Qian, “A new nonlinear control strat- egy embedded with reinforcement learning for a multirotor transporting a suspended payload,”IEEE/ASME Transactions on Mechatronics, vol. 27, no. 2, pp. 1174–1184, 2021

    2021

  14. [22]

    Flare: Agile flights for quadrotor cable-suspended payload system via reinforcement learning,

    D. Cao, J. Zhou, X. Wang, and S. Li, “Flare: Agile flights for quadrotor cable-suspended payload system via reinforcement learning,”arXiv preprint arXiv:2508.09797, 2025

    2025

  15. [23]

    Multi-agent actor-critic for mixed cooperative-competitive environments,

    R. Lowe, Y . I. Wu, A. Tamar, J. Harb, O. Pieter Abbeel, and I. Mordatch, “Multi-agent actor-critic for mixed cooperative-competitive environments,”Advances in neural information processing systems, vol. 30, 2017

    2017

  16. [24]

    The surprising effectiveness of ppo in cooperative multi-agent games,

    C. Yu, A. Velu, E. Vinitsky, J. Gao, Y . Wang, A. Bayen, and Y . Wu, “The surprising effectiveness of ppo in cooperative multi-agent games,” Advances in neural information processing systems, vol. 35, pp. 24 611– 24 624, 2022

    2022

  17. [25]

    Is independent learning all you need in the starcraft multi-agent challenge?

    C. S. De Witt, T. Gupta, D. Makoviichuk, V . Makoviychuk, P. H. Torr, M. Sun, and S. Whiteson, “Is independent learning all you need in the starcraft multi-agent challenge?”arXiv preprint arXiv:2011.09533, 2020

    2011 arXiv

  18. [26]

    Learning decentralized multi-biped control for payload transport,

    B. Pandit, A. Gupta, M. S. Gadde, A. Johnson, A. K. Shrestha, H. Duan, J. Dao, and A. Fern, “Learning decentralized multi-biped control for payload transport,”arXiv preprint arXiv:2406.17279, 2024

    2024 arXiv

  19. [27]

    Decen- tralized navigation of a cable-towed load using quadrupedal robot team via marl,

    W.-T. Chen, M. Nguyen, Z. Li, G. N. Sue, and K. Sreenath, “Decen- tralized navigation of a cable-towed load using quadrupedal robot team via marl,”arXiv preprint arXiv:2503.18221, 2025

    2025 arXiv

  20. [28]

    Deep reinforcement learning-driven collaborative rounding-up for multiple unmanned aerial vehicles in obstacle environments,

    Z. Zhao, Y . Wan, and Y . Chen, “Deep reinforcement learning-driven collaborative rounding-up for multiple unmanned aerial vehicles in obstacle environments,”Drones, vol. 8, no. 9, p. 464, 2024

    2024

  21. [29]

    Payload transporting with two quadrotors by centralized reinforcement learning method,

    D. Lin, J. Han, K. Li, J. Zhang, and C. Zhang, “Payload transporting with two quadrotors by centralized reinforcement learning method,” IEEE Transactions on Aerospace and Electronic Systems, vol. 60, no. 1, pp. 239–251, 2023

    2023

  22. [30]

    Reinforcement learning based trajectory planning for multi-uav load transportation,

    J. Estevez, J. M. Lopez-Guede, J. del Valle-Echavarri, and M. Gra ˜na, “Reinforcement learning based trajectory planning for multi-uav load transportation,”IEEE Access, 2024

    2024

  23. [31]

    Omnidrones: An efficient and flexible platform for reinforcement learning in drone control,

    B. Xu, F. Gao, C. Yu, R. Zhang, Y . Wu, and Y . Wang, “Omnidrones: An efficient and flexible platform for reinforcement learning in drone control,”IEEE Robotics and Automation Letters, vol. 9, no. 3, pp. 2838–2844, 2024

    2024

  24. [32]

    Decentralized aerial manipulation of a cable-suspended load using multi-agent reinforcement learning,

    J. Zeng, A. M. Gimenez, E. Vinitsky, J. Alonso-Mora, and S. Sun, “Decentralized aerial manipulation of a cable-suspended load using multi-agent reinforcement learning,”arXiv preprint arXiv:2508.01522, 2025

    2025

  25. [33]

    Mujoco: A physics engine for model-based control,

    E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,” in2012 IEEE/RSJ international conference on intelligent robots and systems. IEEE, 2012, pp. 5026–5033

    2012

  26. [34]

    Jaxmarl: Multi-agent rl environments and algorithms in jax,

    A. Rutherford, B. Ellis, M. Gallici, J. Cook, A. Lupu, G. Ingvars- son Juto, T. Willi, R. Hammond, A. Khan, C. Schroeder de Wittet al., “Jaxmarl: Multi-agent rl environments and algorithms in jax,”Advances in Neural Information Processing Systems, vol. 37, pp. 50 925–50 951, 2024

    2024

  27. [35]

    Sim-to-(multi)-real: Transfer of low-level robust control policies to multiple quadrotors,

    A. Molchanov, T. Chen, W. H ¨onig, J. A. Preiss, N. Ayanian, and G. S. Sukhatme, “Sim-to-(multi)-real: Transfer of low-level robust control policies to multiple quadrotors,” in2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2019, pp. 59–66

    2019

  28. [36]

    A benchmark comparison of learned control policies for agile quadrotor flight,

    E. Kaufmann, L. Bauersfeld, and D. Scaramuzza, “A benchmark comparison of learned control policies for agile quadrotor flight,” in 2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 10 504–10 510

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.