SoccerDiffusion: Toward Learning End-to-End Humanoid Robot Soccer from Gameplay Recordings

Florian Vahl; Jan Gutsche; Jasper G\"uldenstein; Jianwei Zhang; J\"orn Griepenburg

REVIEW 3 major objections 4 minor 28 references

SoccerDiffusion: Toward Learning End-to-End Humanoid Robot Soccer from Gameplay Recordings

T0 review · 3 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A transformer-based diffusion policy trained directly on RoboCup gameplay recordings executes walking, kicking, and fall recovery on a real humanoid in real time after single-step distillation, but tactical play is not yet learned.

desk verdict Honest feasibility study with a useful dataset and a real fall-recovery result, but the abstract oversells walking and kicking since the policy rarely initiates motion from standstill. read the letter →

arxiv 2504.20808 v2 pith:NAXFMLUT submitted 2025-04-29 cs.RO cs.AIcs.LG

Florian Vahl , Jörn Griepenburg , Jan Gutsche , Jasper Güldenstein , Jianwei Zhang This is my paper · ORCID

classification cs.ROcs.AIcs.LG

keywords humanoidrobotsoccerdiffusionpolicybehavioralcloningimitationlearningtransformerknowledgedistillationRoboCupend-to-endcontrol

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 attempts to show that a single transformer-based diffusion policy can learn useful humanoid soccer behaviors, including walking in all directions, kicking, turning, and fall recovery, directly from passive recordings of real RoboCup matches, with no hand-coded skill modules, reward functions, or manually engineered demonstrations. The authors argue this matters because gameplay recordings are cheap and abundant, so a policy that learns from them could serve as a base model for later reinforcement learning or preference optimization instead of starting from scratch. The paper demonstrates the approach with about fifteen hours of match data, a multimodal context of images, joint states, previous joint commands, and orientation readings, and a distillation step that collapses the diffusion chain to one inference pass. The result is a policy that runs onboard in real time and recovers from falls 95% of the time on physical hardware, while high-level tactical behavior is explicitly acknowledged as not yet achieved.

What carries the argument

The load-bearing object is a transformer-based diffusion policy: separate transformer encoders embed joint states, previous joint commands, orientation, and image features into a shared latent space, and a cross-attention decoder denoises a noisy future joint command trajectory conditional on that latent context. The machinery that makes this practical is teacher-student distillation, in which a deterministic 30-step DDIM chain from the teacher generates target trajectories and a single-step student is trained to reproduce them, cutting inference cost enough for onboard real-time execution. The transformer's positional encodings and the diffusion denoising objective together let the model represent multimodal action distributions over long horizons rather than collapsing to a mean action.

What would settle it

Take the released dataset, remove or heavily down-weight all segments longer than a few seconds in which the robot's joints are nearly static, retrain the same architecture from scratch, and count how many times the policy initiates walking from a standing start on the physical robot or in simulation. If the policy still rarely moves, the limitation is not the data balance; if it moves, the standing-still bias was caused by dataset composition.

Watch

Extended reading notes

Core claim

The central claim is that a diffusion-based visuomotor policy can be trained end-to-end from existing gameplay recordings and then deployed on a physical humanoid. Concretely, SoccerDiffusion maps a context window of joint states, previous joint commands, reconstructed roll and pitch rotations, and downsampled images, together with the game state and the diffusion step, to a denoised joint command trajectory. After training on 88 recordings totaling about 15 hours, the model produces stable walking, turning, kicking, head-scanning, and multi-stage fall recovery in simulation and on the physical robot, and a one-step distilled student matches the teacher's behavior in real time. The paper does not claim tactical competence: intentional kicking, ball positioning, and self-initiated walking remain limited, which the authors attribute to dataset composition and model capacity.

Load-bearing premise

The load-bearing premise is that the recorded match data is a sufficiently balanced teacher: if the long stretches of robots standing still dominate the dataset, the model will default to standing still and never learn to initiate movement, which the paper itself observes as a strong bias.

Editorial extensions

If this is right

A base humanoid soccer policy can be pretrained from match recordings alone, without hand-crafted skill demonstrations or reward shaping.
Distillation to a single diffusion step makes the policy fast enough for real-time control on embedded hardware, so the learned behaviors are deployable, not just offline results.
The learned policy transfers from real-world training data to simulation without additional fine-tuning, indicating the behavior representation is not tightly overfit to the exact real-world conditions.
A gameplay-pretrained diffusion model provides a concrete starting point for RL or preference optimization, potentially reducing the exploration burden compared with learning such dynamic motor skills from scratch.
Tactical behavior such as intentional kicking and field positioning does not emerge from this data alone, so reaching full autonomous soccer requires additional mechanisms beyond imitating the recorded distribution.

Reading between the lines

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

If the standing-still bias is confirmed as a data-composition effect, a simple reweighting or filtering of stationary segments could turn the same architecture into a policy that initiates walking, a testable extension directly motivated by the paper's own observation.
The compressed one-step student is claimed to behave like the teacher; one could quantify this by measuring trajectory diversity across repeated samples, since a diffusion teacher's multimodality is exactly what classical behavioral cloning loses.
The successful real-to-sim transfer without adaptation hints that adding simulated game recordings to the training set could cheaply expand coverage of rare events such as falls and kicks, potentially improving the high-level behavior the paper finds missing.
A direct head-to-head comparison of RL fine-tuning from the released base model versus RL from random initialization on the same soccer task would quantify how much the gameplay-pretrained prior actually accelerates learning.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The core of this paper is a feasibility test: can a transformer-based diffusion policy learn useful humanoid soccer behaviors from real match logs? The answer is a qualified yes, and the paper is honest about the qualifications. The released 15-hour dataset from RoboCup is a genuinely useful artifact, and the quantitative fall-recovery results are real evidence—95% physical success versus 100% for the baseline, with 100% in simulation. Real-time inference after distillation is also demonstrated, which matters for embedded deployment. Those are solid contributions.

The soft spots are real and concentrated in the gap between the abstract and the evidence. The stress-test note is right: the standing-still bias reported in §4.2 is not just missing high-level tactics. It is a low-level initiation failure. The policy freezes when the context window has no prior motion, which is exactly the cold-start condition a soccer robot faces after a fall, at kick-off, or after any stoppage. The authors acknowledge this and attribute it to standing-heavy training data, but that attribution actually sharpens the problem: the data distribution is not a sufficient teacher for the most basic useful behavior. The qualitative "walking" and "kicking motions" also fall short of the abstract's promise. No ball contact is shown in the kicking evaluation, and walking is described in terms of similarity to the dataset, not as goal-directed locomotion. The fall-recovery numbers carry the quantitative weight, and they are good, but they do not support the unqualified claim that the model replicates "walking, kicking, and fall recovery" as equivalent achievements.

The citation pattern is clean. Self-citation is limited to the team's own platform and dataset, which is normal. No fitted constants are disguised as predictions, and the policy is tested on a physical robot and in simulation with a domain shift. That is genuinely reproducible work.

Who should read this? The RoboCup community and anyone working on imitation learning for legged robots will get value from the dataset and the negative result about standing-still bias. The paper deserves a serious referee. My recommendation is to send it to peer review, but require a revision that either tempers the abstract to match the evidence or adds a quantitative walking-initiation test (e.g., proportion of cold-start trials where the robot steps within N seconds). As it stands, the central claim is partially supported, and the paper itself provides the clearest reason why.

Referee Report

3 major / 4 minor

Summary. This paper presents SoccerDiffusion, a transformer-based diffusion policy trained end-to-end on approximately 15 hours of real RoboCup gameplay recordings from the Hamburg Bit-Bots. The model consumes image, joint, rotation, and game-state inputs and outputs joint command trajectories; a teacher-student distillation step reduces the DDIM sampling to a single step for real-time inference. The main quantitative result is a fall-recovery evaluation (40 physical trials, 95% success; 40 simulation trials, 100% success) compared with the team's baseline stack. Walking, turning, kicking-like motions, and stabilization are described qualitatively. The authors release the dataset, code, and pretrained models.

Significance. If the claims are appropriately scoped, the paper is a useful feasibility study: it shows that real-world match logs, without hand-designed demonstrations, can yield a policy that reproduces a complex recovery skill at near-baseline reliability and that distillation can make such a policy real-time on a humanoid platform. The public release of the dataset and models is a concrete contribution to the RoboCup and legged-locomotion communities. However, the central abstract claim that walking, kicking, and fall recovery are replicated is only quantitatively supported for fall recovery; the walking and kicking components are qualitative and the paper's own Section 4.2 reports a failure to initiate walking from standstill. The significance is therefore conditional on revising the claims to match the evidence or adding targeted experiments.

major comments (3)

[§4.2 / Abstract] The claim that the model replicates walking is not supported by the evidence presented. Section 4.2 states that there is "a strong bias to keep standing still, if there was no prior motion in the context window" and that "the robot rarely starts to walk on its own." Since a soccer robot must initiate locomotion from a standstill at kick-off, after stoppages, and after falls, the policy's inability to initiate walking means the abstract's "walking" is not demonstrated in the gameplay setting. The qualitative walking description appears to describe continuation of motion once the robot is already stepping. The abstract and conclusion should be revised to limit "walking" to context-conditioned continuation, or a cold-start walking test should be added.
[§4.2] The "kicking" claim is based on observation of "kicking motions" but no evidence is provided that the robot contacts a ball or that the motion has any functional effect. A controlled experiment with a stationary ball, reported as contact rate or ball displacement, is needed to support the abstract's "kicking" claim; otherwise the text should say "kicking-like motions without ball contact."
[§3.4 / §4] The real-time inference claim is load-bearing but unsupported by measurements. The paper states that distillation reduces the multistep diffusion to a single step to enable real-time inference on the Ryzen 7 5700U APU, and the conclusion says "We achieved real-time performance," but no inference latency is reported anywhere. Please include wall-clock inference times on the actual Wolfgang-OP hardware and compare them with the 50 Hz control period (20 ms) used in the data pipeline.

minor comments (4)

[Table 2] Table 2 contains inconsistent labels, such as "J. Transformer Encoder 4L" and later "J. Transformer 4L" with different layer counts, and similarly for rotations, which makes the architecture specification ambiguous. Please unify the table and define each row clearly.
[§4.1] The fall-recovery success rates (95% physical, 100% baseline) are reported without statistical treatment. With 40 trials per condition, a Fisher exact test would not show a significant difference, and the results should be described as "comparable" rather than implying a meaningful gap.
[§3.4] The comparison between the distilled and non-distilled models is only qualitative, with the statement that no qualitative difference was observed. Reporting the same fall-recovery protocol for the teacher model would substantiate the claim that distillation preserves behavior.
[§3.1] The paper mentions that data from other RoboCup teams was obtained but not included in the evaluation; please specify why it was excluded and clarify whether the released dataset contains only Bit-Bots recordings.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the policy is trained on recorded gameplay and evaluated on a physical robot and in simulation; the standing-still bias undercuts the walking claim but is an empirical limitation, not a derivation that reduces to its inputs.

full rationale

This paper is an empirical imitation-learning study. The derivation chain is: real match recordings (from the authors' own team) -> preprocessing -> transformer-based diffusion training on a joint-command denoising objective -> DDIM distillation -> evaluation on a physical Wolfgang-OP robot and in Webots simulation. The only quantitative metric, fall recovery (Table 3), is compared against the baseline software stack and evaluated both physically and in simulation; it is not an output that is mathematically forced by the training objective. The model's objective is to imitate the recorded action distribution, and claiming that it does so is an empirical claim tested on a physical platform, not a tautology. Qualitative claims about walking and kicking are based on observing the policy execute motions, and the paper candidly reports a 'strong bias to keep standing still' that limits autonomous walking; that is a validity or coverage limitation, not circularity. Self-citations are limited to the team's own robot platform [2] and dataset, which are data and hardware provenance rather than load-bearing theoretical results. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via citation. The standing-still bias is an honest limitation that affects the strength of the 'walking' claim, but it does not make the derivation equivalent to its inputs by construction.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities; the game-state encoding and latent representations are internal to the model. All free parameters are hand-chosen architectural and data-processing choices. The axioms are domain assumptions about data quality, reconstruction fidelity, simulation transfer, and distillation faithfulness.

free parameters (7)

Context window lengths = 2 s for joints/rotations, 1 s for images
Chosen to balance temporal context and latency; not derived from data.
Token patch size = 5
Hand-selected hyperparameter for embedding joint and rotation sequences.
Transformer hidden size = 512
Selected to fit real-time inference on a Ryzen 7 5700U; not justified by an ablation.
Diffusion steps = 1000 training, 30 teacher sampling, 1 student
Standard choices; distilling to 1 step enables real-time execution, but parity is only qualitatively checked.
Sensor sampling rates = 50 Hz joints/IMU, 10 Hz images
Chosen to synchronize modalities and limit data volume; not derived from the task.
Image resolution = 224x224
Downsampled from 480x480; chosen for compute, not for task requirements.
Game state simplification = 3 states (play, not play, unknown)
Hand-designed compression of game controller messages.

assumptions (5)

domain assumption The denoising objective of diffusion models provides a valid learned policy and preserves multimodality.
Central method choice; not proven for this domain.
domain assumption The recorded joint commands from the existing team software are a sufficiently expert and diverse teacher.
If the dataset is dominated by standing sequences, the policy collapses to stand still, as the authors report.
domain assumption The reconstructed roll/pitch IMU state from an intermediate representation is accurate enough for stable control.
Section 3.2 notes IMU was not directly recorded in 2024 and was reconstructed; only roll and pitch are available.
domain assumption Webots dynamics approximate the physical robot sufficiently for sim-to-real transfer.
Evaluation relies on simulation transfer without additional training; sim and real performance are close in fall recovery.
domain assumption Single-step distillation preserves the behavioral distribution of the 30-step teacher.
Only a qualitative check is reported; no quantitative distributional comparison.

how reviews work

0 comments

Cite this review

Pith. "Pith review of SoccerDiffusion: Toward Learning End-to-End Humanoid Robot Soccer from Gameplay Recordings." pith.science (2026). https://pith.science/paper/NAXFMLUT

@misc{pith2026250420808,
  author       = {Pith},
  title        = {Pith review of: SoccerDiffusion: Toward Learning End-to-End Humanoid Robot Soccer from Gameplay Recordings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NAXFMLUT}},
  note         = {Machine review of arXiv:2504.20808}
}

read the original abstract

This paper introduces SoccerDiffusion, a transformer-based diffusion model designed to learn end-to-end control policies for humanoid robot soccer directly from real-world gameplay recordings. Using data collected from RoboCup competitions, the model predicts joint command trajectories from multi-modal sensor inputs, including vision, proprioception, and game state. We employ a distillation technique to enable real-time inference on embedded platforms that reduces the multi-step diffusion process to a single step. Our results demonstrate the model's ability to replicate complex motion behaviors such as walking, kicking, and fall recovery both in simulation and on physical robots. Although high-level tactical behavior remains limited, this work provides a robust foundation for subsequent reinforcement learning or preference optimization methods. We release the dataset, pretrained models, and code under: https://bit-bots.github.io/SoccerDiffusion

Figures

Figures reproduced from arXiv: 2504.20808 by the authors.

**Figure 2.** Qualitative evaluation: (a) walking and (b) fall recovery, both performed [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 10 canonical work pages

[1]

In: Machine In- telligence 15: Intelligent Agents, pp

Bain, M., Sammut, C.: A framework for behavioural cloning. In: Machine In- telligence 15: Intelligent Agents, pp. 103–129. Oxford University Press (2000). https://doi.org/10.1093/oso/9780198538677.003.0006

arXiv 2000
[2]

In: 2020 IEEE-RAS 20th International Conference on Humanoid Robots (Humanoids)

Bestmann, M., Güldenstein, J., Vahl, F., Zhang, J.: Wolfgang-OP: A Robust Hu- manoid Robot Platform for Research and Competitions. In: 2020 IEEE-RAS 20th International Conference on Humanoid Robots (Humanoids). pp. 90–97 (2021). https://doi.org/10.1109/HUMANOIDS47582.2021.9555808

arXiv 2021
[3]

arXiv preprint (2023)

Chang,Z.,Koulieris,G.A.,Shum,H.P.H.:OntheDesignFundamentalsofDiffusion Models: A Survey. arXiv preprint (2023). https://doi.org/10.48550/arXiv.2306. 04542

doi:10.48550/arxiv.2306 2023
[4]

In: Proc

Chen, S.F., Wang, H.C., Hsu, M.H., Lai, C.M., Sun, S.H.: Diffusion model- augmented behavioral cloning. In: Proc. of the 41st International Conference on Machine Learning (ICML) (2024). https://doi.org/10.48550/arXiv.2302.13335

arXiv doi:10.48550/arxiv.2302.13335 2024
[5]

The International Journal of Robotics Research (2024)

Chi et al.: Diffusion Policy: Visuomotor Policy Learning via Action Diffusion. The International Journal of Robotics Research (2024). https://doi.org/10.1177/ 02783649241273668

work page 2024
[6]

Curtright, T.L., Fairlie, D.B., Zachos, C.K., et al.: A compact formula for rota- tions as spin matrix polynomials. SIGMA. Symmetry, Integrability and Geometry: Methods and Applications (2014). https://doi.org/10.3842/SIGMA.2014.084

work page doi:10.3842/sigma.2014.084 2014
[7]

Cyberbotics Ltd.: Webots: Open-source Mobile Robot Simulation Software, https: //www.cyberbotics.com

work page
[8]

Foxglove Developers: MCAP: serialization-agnosticlog container file format (2024), https://mcap.dev, available from https://github.com/foxglove/mcap

work page 2024

Show all 28 references

[9]

Science Robotics (2024)

Haarnoja et al.: Learning agile soccer skills for a bipedal robot with deep rein- forcement learning. Science Robotics (2024). https://doi.org/10.1126/scirobotics. adi8022 12 F. Vahl et al

2024 doi
[10]

In: Proc

He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proc. of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016). https://doi.org/10.1109/CVPR.2016.90

2016 doi
[11]

In: Proc

Ho, J., Jain, A., Abbeel, P.: Denoising Diffusion Probabilistic Models. In: Proc. of Advances in Neural Information Processing Systems 33 (NeurIPS). pp. 6840–6851 (2020). https://doi.org/10.48550/arXiv.2006.11239

2020 doi arXiv
[12]

arXiv preprint (2022)

Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv preprint (2022). https: //doi.org/10.48550/arXiv.2207.12598

2022 doi arXiv
[13]

ACM Comput

Hussein, A., Gaber, M.M., Elyan, E., Jayne, C.: Imitation Learning: A Survey of Learning Methods. ACM Comput. Surv. 50(2) (2017). https://doi.org/10.1145/ 3054912

2017
[14]

Science Robotics (2022)

Liu et al.: From motor control to team play in simulated humanoid football. Science Robotics (2022). https://doi.org/10.1126/scirobotics.abo0235

2022 doi
[15]

arXiv preprint (2017)

Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. arXiv preprint (2017). https://doi.org/10.48550/arXiv.1711.05101

2017 doi arXiv
[16]

arXiv preprint (2021)

Luhman, E., Luhman, T.: Knowledge Distillation in Iterative Generative Models for Improved Sampling Speed. arXiv preprint (2021). https://doi.org/10.48550/ arXiv.2101.02388

2021 doi arXiv
[17]

arXiv preprint (2023)

Oquab et al.: Dinov2: Learning robust visual features without supervision. arXiv preprint (2023). https://doi.org/10.48550/arXiv.2304.07193

2023 doi arXiv
[18]

In: The Eleventh International Conference on Learning Representations (ICLR) (2023)

Pearce et al.: Imitating Human Behaviour with Diffusion Models. In: The Eleventh International Conference on Learning Representations (ICLR) (2023). https://doi. org/10.48550/arXiv.2301.10677

2023 doi arXiv
[19]

In: Proc

Pomerleau, D.A.: Alvinn: an autonomous land vehicle in a neural network. In: Proc. of Advances in Neural Information Processing Systems 1 (NIPS). pp. 305– 313 (1988)

1988
[20]

arXiv preprint (2023)

Reuss, M., Li, M., Jia, X., Lioutikov, R.: Goal-Conditioned Imitation Learning using Score-based Diffusion Policies. arXiv preprint (2023). https://doi.org/10. 48550/arXiv.2304.02532

2023 doi arXiv
[21]

In: Proc

Ross, S., Bagnell, D.: Efficient Reductions for Imitation Learning. In: Proc. of the Thirteenth International Conference on Artificial Intelligence and Statistics (AISTATS). vol. 9, pp. 661–668 (2010), https://proceedings.mlr.press/v9/ross10a

2010
[22]

Autonomous Robots 48(2) (2024)

Rossi et al.: The human in the loop Perspectives and challenges for RoboCup 2050. Autonomous Robots 48(2) (2024). https://doi.org/10.1007/s10514-024-10159-3

2024 doi
[23]

In: Proc

Schaal, S.: Learning from demonstration. In: Proc. of Advances in Neural Informa- tion Processing Systems 9 (NIPS). pp. 1040–1046 (1996)

1996
[24]

Schaal, S.: Is imitation learning the route to humanoid robots? Trends in cognitive sciences 3(6), 233–242 (1999)

1999
[25]

In: Interna- tional Conference on Learning Representations (ICLR) (2021)

Song, J., Meng, C., Ermon, S.: Denoising Diffusion Implicit Models. In: Interna- tional Conference on Learning Representations (ICLR) (2021). https://doi.org/10. 48550/arXiv.2010.02502

2021 doi arXiv
[26]

arXiv preprint (2024)

Tirumala et al.: Learning robot soccer from egocentric vision with deep reinforce- ment learning. arXiv preprint (2024). https://doi.org/10.48550/arXiv.2405.02425

2024 doi arXiv
[27]

In: Proc

Vaswani et al.: Attention is all you need. In: Proc. of Advances in Neural Infor- mation Processing Systems 30 (NIPS). pp. 6000–6010 (2017). https://doi.org/10. 48550/arXiv.1706.03762

2017 doi arXiv
[28]

arXiv preprint (2025)

Wolf, R., Shi, Y., Liu, S., Rayyes, R.: Diffusion Models for Robotic Manipulation: A Survey. arXiv preprint (2025). https://doi.org/10.48550/arXiv.2504.08438

2025 doi arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

In: Machine In- telligence 15: Intelligent Agents, pp

Bain, M., Sammut, C.: A framework for behavioural cloning. In: Machine In- telligence 15: Intelligent Agents, pp. 103–129. Oxford University Press (2000). https://doi.org/10.1093/oso/9780198538677.003.0006

arXiv 2000

[2] [2]

In: 2020 IEEE-RAS 20th International Conference on Humanoid Robots (Humanoids)

Bestmann, M., Güldenstein, J., Vahl, F., Zhang, J.: Wolfgang-OP: A Robust Hu- manoid Robot Platform for Research and Competitions. In: 2020 IEEE-RAS 20th International Conference on Humanoid Robots (Humanoids). pp. 90–97 (2021). https://doi.org/10.1109/HUMANOIDS47582.2021.9555808

arXiv 2021

[3] [3]

arXiv preprint (2023)

Chang,Z.,Koulieris,G.A.,Shum,H.P.H.:OntheDesignFundamentalsofDiffusion Models: A Survey. arXiv preprint (2023). https://doi.org/10.48550/arXiv.2306. 04542

doi:10.48550/arxiv.2306 2023

[4] [4]

In: Proc

Chen, S.F., Wang, H.C., Hsu, M.H., Lai, C.M., Sun, S.H.: Diffusion model- augmented behavioral cloning. In: Proc. of the 41st International Conference on Machine Learning (ICML) (2024). https://doi.org/10.48550/arXiv.2302.13335

arXiv doi:10.48550/arxiv.2302.13335 2024

[5] [5]

The International Journal of Robotics Research (2024)

Chi et al.: Diffusion Policy: Visuomotor Policy Learning via Action Diffusion. The International Journal of Robotics Research (2024). https://doi.org/10.1177/ 02783649241273668

work page 2024

[6] [6]

Curtright, T.L., Fairlie, D.B., Zachos, C.K., et al.: A compact formula for rota- tions as spin matrix polynomials. SIGMA. Symmetry, Integrability and Geometry: Methods and Applications (2014). https://doi.org/10.3842/SIGMA.2014.084

work page doi:10.3842/sigma.2014.084 2014

[7] [7]

Cyberbotics Ltd.: Webots: Open-source Mobile Robot Simulation Software, https: //www.cyberbotics.com

work page

[8] [8]

Foxglove Developers: MCAP: serialization-agnosticlog container file format (2024), https://mcap.dev, available from https://github.com/foxglove/mcap

work page 2024