Pith. sign in

REVIEW 4 major objections 5 minor 33 references

Embodiment-Agnostic Navigation Policy Trained with Visual Demonstrations

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a diffusion-based navigation policy trained on 1.5 hours of human-held depth-camera demonstrations, with no robot in the loop, can be deployed on a legged robot to reach and track a human while avoiding obstacles.

desk verdict A useful data-collection trick for navigation, wrapped in an overbroad 'embodiment-agnostic' claim; the real-robot results are real but the evaluation is thinner than the title suggests. read the letter →

arxiv 2412.20226 v1 pith:5O76K4NY submitted 2024-12-28 cs.RO

Nimrod Curtis , Osher Azulay , Avishai Sintov This is my paper · ORCID
classification cs.RO
keywords depth-basednavigationdiffusionpolicybehaviorcloningembodiment-agnostichumandemonstrationtargettrackingleggedrobotvisual
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

ViDEN claims that a robot navigation policy can be trained entirely from demonstrations recorded by a person walking a handheld depth camera, with no robot in the loop, and then deployed on a different embodiment. The paper argues that representing actions as normalized waypoints in the camera frame, conditioning on the target's relative position, and generating trajectories with a diffusion model makes the transfer possible. On a Unitree Go2 quadruped, the resulting policy reaches and tracks a human while avoiding obstacles, with success rates of 100%, 96%, and 84% on easy, moderate, and hard tasks using only about 1.5 hours of collected data. This matters because demonstration-based navigation is usually tied to a specific robot or needs 100+ hours of robot-collected data, while ViDEN offers a cheaper, robot-independent route to goal-directed navigation. The same framework also shows zero-shot transfer to a new indoor environment at 55% success, improving with modest fine-tuning data.

What carries the argument

The load-bearing mechanism is the combination of an embodiment-agnostic action space and a goal-conditioned diffusion policy. Actions are trajectories of camera-frame waypoints normalized by Eq. (1), so the policy outputs the same normalized relative motion regardless of robot; deployment rescales by the robot's maximum velocity. Depth frames are encoded with a CNN, target position and a randomly sampled intermediate goal are encoded with MLPs, and the transformer-encoded latent is fed to a U-Net diffusion model that denoises an action trajectory over K iterations. The paper's contribution is that this pipeline transfers from a human-held camera at 0.3–0.5 m height to a quadruped, because the policy learns task-centric geometry rather than robot-specific dynamics.

What would settle it

Run the same trained policy on a second embodiment, such as a wheeled robot or a taller legged robot, in the same easy, moderate, and hard tasks. If success rates drop far below the reported 100%, 96%, and 84%, the embodiment-agnostic transfer claim fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that depth images plus relative target position provide enough information for a diffusion-based behavior-cloning policy to synthesize collision-free, target-directed trajectories, and that the demonstration data need not come from the robot at all. A human collects 120 trajectories by walking with a hand-held stereo camera toward a person while passing obstacles; the target is extracted with a detector, the camera trajectory is recorded as $SE(2)$ poses, and the policy is trained to output 32-step normalized waypoint trajectories conditioned on depth and on an intermediate goal. When deployed on a quadruped with a PD controller, the policy completes reaching and tracking tasks at higher success rates than a convolutional MLP baseline and than a goal-conditioned ViNT variant adapted to relative targets, across easy, moderate, and hard obstacle configurations. The paper further reports that task-agnostic exploration data, goal masking, dropout of target position, and a memory token each contribute to robustness, and that the policy can partially infer targets from depth without explicit labels.

Load-bearing premise

The paper assumes that demonstrations recorded by a person walking with a handheld camera at 0.3–0.5 m height can be executed by a different robot after linear velocity normalization, without modeling differences in kinematics, dynamics, sensor height, or field of view, and this premise is tested on only one quadruped.

Editorial extensions

If this is right

  • Navigation policies can be trained without any robot in the loop, so a new target-following capability can be obtained with about 1.5 hours of human walking instead of 100+ hours of robot teleoperation.
  • Because goals are specified by relative position rather than goal images, the policy can follow a moving human and does not require a topological map or pre-specified target image.
  • Goal masking lets a single policy switch between directed navigation and exploration, giving it some ability to infer targets from depth alone when explicit target labels are absent.
  • Deploying in a new environment needs only about 50% of the demonstration data of the original environment to match training performance, suggesting rapid adaptation to new scenes.

Reading between the lines

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

  • The same data-collection scheme should extend to other target classes, such as animals, vehicles, or arbitrary objects, and to other robots with different camera placements, but only if viewpoint differences are handled; testing on a second embodiment would clarify the limits.
  • A natural next step the paper leaves implicit is to split the learned policy into a high-level goal and trajectory planner and a per-robot low-level controller, which would likely improve robustness to physical perturbations, where the reported success drops to 30%.
  • Because the policy sees depth only, its success in low light (73%) may reflect the depth sensor's resilience; an RGB-only variant would isolate whether depth, rather than the diffusion architecture, drives the transfer.
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 / 5 minor

Summary. The paper proposes ViDEN, a behavior-cloning framework that trains a diffusion-based navigation policy entirely from human demonstrations recorded with a hand-held depth camera. The policy consumes depth images plus a relative target position and outputs a short waypoint trajectory, which is then re-scaled to the robot's maximum velocity. The authors claim that this makes the policy embodiment-agnostic. The main experiments deploy the policy on a Unitree Go2 quadruped for human reaching/tracking in easy, moderate, and hard tasks, and compare against a BC-ConvMLP baseline and a modified ViNT baseline. Additional experiments measure robustness to low light, physical perturbations, and dynamic obstacles, as well as generalization to a new indoor environment with varying amounts of fine-tuning data. The central claim is that ViDEN achieves higher success rates than the baselines with only about one hour of robot-independent data.

Significance. If the empirical results hold, the paper makes a useful contribution to demonstration-based navigation: it shows that a diffusion policy trained on data collected without any robot can be deployed on a legged platform, and that depth plus relative target conditioning can reduce the data requirement compared to image-goal methods such as ViNT. The open-source code and the emphasis on low-cost data collection are concrete strengths. However, the significance is currently limited by three issues: the 'embodiment-agnostic' claim is supported by a single robot and a scalar re-scaling; the success-rate evaluation lacks an explicit success criterion, error bars, and statistical detail; and the ViNT baseline is modified in a way that may not be a faithful comparison. These issues are addressable, but they are load-bearing for the paper's central claims.

major comments (4)
  1. [II-B, III-B] The embodiment-agnostic claim is the load-bearing premise of the title and Section II-B, but it rests on a single experiment on a Unitree Go2. The only embodiment adaptation described is a scalar re-scaling (Eq. (1) and the 'denormalization with its maximum velocity' in Section II-E), which does not account for differences in kinematics, dynamics, camera height, field of view, or control bandwidth. As written, Eq. (1) also has a units problem: p_t is a position, while β_min and β_max are described as 'motion velocities,' so the normalized quantity is dimensionally inconsistent unless β are actually displacements or the text is rephrased. To support the title's claim, please test on at least one additional embodiment with significantly different scale or kinematics, or explicitly scope the claim to e.g., 'legged platforms with comparable camera geometry.'
  2. [III-C, Table I] The success rates are the paper's central quantitative evidence, but the manuscript never defines what counts as a success: there is no distance threshold to the human, no time limit, no collision tolerance, and no specification of whether the trial ends when the target is lost. With 25 trials per condition and 5 per scene, reporting only aggregate percentages without error bars or per-scene variability makes it impossible to judge whether the differences between ViDEN (96%, 84%) and its simplified version (88%, 72%) or between ViDEN and ViNT are meaningful. Please report the exact success definition, the number of trials per cell, confidence intervals, and preferably per-scene results.
  3. [III-C, baselines] The ViNT baseline is modified by replacing its target image with the relative distance to the target. This is a substantive alteration to the model's conditioning mechanism, and ViNT is designed and trained for image goals, so the comparison may not reflect ViNT's intended performance. The manuscript should state exactly how the modification was implemented: which modules were replaced, what was retrained, and with what data and hyperparameters. Without this, the claim that ViDEN outperforms 'existing methods' is not fully supported.
  4. [III-A, III-E] The reported data volume is inconsistent: Section I states '1.5 hours of robot-independent data,' while Section III-A reports 120 demonstrations with an average duration of 30 seconds, which is about 60 minutes. In addition, several critical hyperparameters (λ=2.5 m set by 'trial and error,' the β scaling, goal masking probability pm=0.7, task-agnostic fraction κ=20%, and target dropout 50%) are tuned for the evaluation environment without any sensitivity analysis. Since these choices directly influence the success rates in Table I and Figure 7, please clarify the actual data budget and provide at least a minimal sensitivity study or explicitly state that the main results are not sensitive to these values.
minor comments (5)
  1. [II-B] Equation (1) and the surrounding text should use consistent units and explicitly state whether the normalized quantity is a displacement, velocity, or pose waypoint; the current phrasing 'position' versus 'motion velocity' is confusing.
  2. [III-A] The data amount should be aligned: Section I says 1.5 hours, while Section III-A says approximately 60 minutes; please make the numbers consistent and clarify whether task-agnostic data is additional.
  3. [III-C, Table II] The robustness table reports success rates but does not specify the number of trials per disturbance type or the magnitude and duration of the perturbations; please add this information so the rates are interpretable.
  4. [III-C] The claim of outperforming 'existing methods' should be qualified: the only named baseline from the recent literature is ViNT, and it is used in modified form; GNM and NoMaD are discussed but not evaluated.
  5. [III-C, Figure 7] The generalization plot has no error bars or trial counts; given that the zero-shot success rate is 55%, please provide confidence intervals to support the conclusion that zero-shot transfer is feasible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ViDEN's success rates come from external real-robot trials, and the few self-citations are not load-bearing.

full rationale

The paper's central claim is empirical: a diffusion policy trained on human-held depth-camera demonstrations is evaluated by measured success rates on a Unitree Go2 quadruped in real-world trials. These reported numbers are not derived from the training data by construction, nor are they fitted parameters renamed as predictions. The action-space normalization in Section II-B, Eq. (1), is a fixed input transformation using dataset velocity extrema and a robot-specific max-velocity scaling; it is an assumption about transfer, not a fitted quantity that forces the reported success rates. The goal distance lambda is said to be set by trial and error, but it is a deployment hyperparameter, not the target result. The ablations and baselines are compared on the same external trials, so the central comparison does not reduce to an identity. The self-citations, refs. [2] and [22], are background statements about RL data requirements and behavior cloning; they are not invoked to justify the central 'embodiment-agnostic' claim or to forbid alternative approaches. No uniqueness theorem or prior result by the same authors is load-bearing. The manuscript's own limitations, such as testing on a single embodiment and noting weaker performance under physical perturbations, are external-validity caveats rather than circularity. The 'embodiment-agnostic' generalization is under-tested, but under-testing is not circular reasoning. The derivation chain from demonstrations to deployment is self-contained in the sense that the evaluated outcomes are externally measured, not constructed from the inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

No new physical entities; all components are standard. However, several hyperparameters are hand-tuned, and the normalization scheme embodies a strong cross-embodiment transfer assumption.

free parameters (6)
  • beta_min/beta_max = from dataset min/max velocities
    Normalization in Eq. (1) is computed from the collected demonstration velocities, so the action space depends on the training data distribution.
  • lambda (goal distance) = 2.5 m
    Set in deployment 'based on trial and error' (Section III-A), this defines the intermediate goal placement and affects behavior.
  • pm (goal masking probability) = 0.7
    Training hyperparameter for goal masking, chosen by the authors.
  • kappa (task-agnostic data fraction) = 20%
    Fraction of exploration data, chosen by the authors.
  • target positional dropout = 50%
    Dropout applied to target positional data during training, chosen by the authors.
  • robot scaling factor beta = per-robot max velocity
    Denormalization factor used at deployment to adapt to a specific robot's velocity limits.
assumptions (5)
  • domain assumption Human demonstrations are expert behavior for navigation
    The framework treats trajectories of a human holding a camera as optimal demonstrations for a robot to imitate (Section II-A).
  • domain assumption Depth images plus relative target position provide sufficient state representation
    The policy conditions only on depth and target-relative position, ignoring RGB except for detection, assuming this suffices for collision-free navigation (Section II-C).
  • domain assumption YOLOv5 detection and depth centroid projection yield a reliable target position
    Target st is derived from a detected bounding box projected onto depth (Section II-A); errors or missed detections are handled by a memory, but the assumption is load-bearing.
  • ad hoc to paper Linear normalization of waypoints transfers trajectories across embodiments
    Eq. (1) and the per-robot scaling assume a simple linear map from human camera motion to robot motion, which is not physically derived (Section II-B).
  • standard math The diffusion model and denoising procedure converge to good trajectories
    Uses standard diffusion policy framework from [28] without new theory.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Embodiment-Agnostic Navigation Policy Trained with Visual Demonstrations." pith.science (2026). https://pith.science/paper/5O76K4NY

@misc{pith2026241220226,
  author       = {Pith},
  title        = {Pith review of: Embodiment-Agnostic Navigation Policy Trained with Visual Demonstrations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5O76K4NY}},
  note         = {Machine review of arXiv:2412.20226}
}
read the original abstract

Learning to navigate in unstructured environments is a challenging task for robots. While reinforcement learning can be effective, it often requires extensive data collection and can pose risk. Learning from expert demonstrations, on the other hand, offers a more efficient approach. However, many existing methods rely on specific robot embodiments, pre-specified target images and require large datasets. We propose the Visual Demonstration-based Embodiment-agnostic Navigation (ViDEN) framework, a novel framework that leverages visual demonstrations to train embodiment-agnostic navigation policies. ViDEN utilizes depth images to reduce input dimensionality and relies on relative target positions, making it more adaptable to diverse environments. By training a diffusion-based policy on task-centric and embodiment-agnostic demonstrations, ViDEN can generate collision-free and adaptive trajectories in real-time. Our experiments on human reaching and tracking demonstrate that ViDEN outperforms existing methods, requiring a small amount of data and achieving superior performance in various indoor and outdoor navigation scenarios. Project website: https://nimicurtis.github.io/ViDEN/.

Figures

Figures reproduced from arXiv: 2412.20226 by the authors.

Figure 1
Figure 1. (Top) task-centric and task-agnostic demonstrations [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Architecture of the proposed Visual Demonstration-based Embodiment-agnostic Navigation (ViDEN) framework. The [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Policy deployment in a hard level outdoor environment. The robot avoids the obstacles and reaches the human target. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Policy deployment in a hard level outdoor environment with a dynamic human target. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Demonstrations of various challenging scenarios [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Policy deployment in a hard level indoor environment. The policy was fine-tuned with 30 minutes of demonstrations [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Success rate with regard to the time length of [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 26 canonical work pages

  1. [1]

    Survey of deep reinforcement learning for motion planning of autonomous vehicles,

    S. Aradi, “Survey of deep reinforcement learning for motion planning of autonomous vehicles,”IEEE Transactions on Intelligent Transporta- tion Systems, vol. 23, pp. 740–759, 02 2022

    work page 2022

  2. [2]

    Learning to throw with a handful of samples using decision transformers,

    M. Monastirsky, O. Azulay, and A. Sintov, “Learning to throw with a handful of samples using decision transformers,” IEEE Robotics and Automation Letters, vol. 8, no. 2, pp. 576–583, 2023

    work page 2023

  3. [3]

    Recent advances in robot learning from demonstration,

    H. Ravichandar, A. S. Polydoros, S. Chernova, and A. Billard, “Recent advances in robot learning from demonstration,” Annual review of control, rob. and auto. sys. , vol. 3, no. 1, pp. 297–330, 2020

    work page 2020

  4. [4]

    MimicPlay: Long-horizon imitation learning by watching human play,

    C. Wang, L. J. Fan, J. Sun, R. Zhang, L. Fei-Fei, D. Xu, Y . Zhu, and A. Anandkumar, “MimicPlay: Long-horizon imitation learning by watching human play,” in Conference on Robot Learning , 2023

    work page 2023

  5. [5]

    Intelligent mobile robot navigation in unknown and complex environment using reinforcement learning technique,

    R. Raj and A. Kos, “Intelligent mobile robot navigation in unknown and complex environment using reinforcement learning technique,” Scientific Reports, vol. 14, no. 1, p. 22852, 2024

    work page 2024

  6. [6]

    A comprehensive study for robot navigation techniques,

    W. R. Faiza Gul and S. S. N. Alhady, “A comprehensive study for robot navigation techniques,” Cogent Engineering , vol. 6, no. 1, p. 1632046, 2019

    work page 2019

  7. [7]

    Using occupancy grids for mobile robot perception and navigation,

    A. Elfes, “Using occupancy grids for mobile robot perception and navigation,” Computer, vol. 22, no. 6, pp. 46–57, 1989

    1989

  8. [8]

    OctoMap: An efficient probabilistic 3d mapping framework based on octrees,

    A. Hornung, K. M. Wurm, M. Bennewitz, C. Stachniss, and W. Bur- gard, “OctoMap: An efficient probabilistic 3d mapping framework based on octrees,” Auto. Robots, vol. 34, no. 3, pp. 189–206, 2013. Fig. 6: Policy deployment in a hard level indoor environment. The policy was fine-tuned with 30 minutes of demonstrations after pre-training in an outdoor enviro...

    work page 2013

Show all 33 references
  1. [9]

    Past, present, and future of simultaneous localization and mapping: Toward the robust-perception age,

    C. Cadena, L. Carlone, H. Carrillo, Y . Latif, D. Scaramuzza, J. Neira, I. Reid, and J. J. Leonard, “Past, present, and future of simultaneous localization and mapping: Toward the robust-perception age,” IEEE Transactions on Robotics , vol. 32, no. 6, pp. 1309–1332, 2016

    2016

  2. [10]

    S. M. LaValle, Planning Algorithms . Cambridge University Press, 2006

    2006

  3. [11]

    Information-theoretic exploration with bayesian optimization,

    S. Bai, J. Wang, F. Chen, and B. Englot, “Information-theoretic exploration with bayesian optimization,” in IEEE/RSJ International Conference on Intelligent Robots and Systems , 2016, pp. 1816–1822

    2016

  4. [12]

    A frontier-based approach for autonomous exploration,

    B. Yamauchi, “A frontier-based approach for autonomous exploration,” in IEEE Int. Symp. Comp. Intel. in Rob. and Aut. , 1997, pp. 146–151

    1997

  5. [13]

    Thrun, W

    S. Thrun, W. Burgard, and D. Fox, Probabilistic Robotics. MIT Press, 2005

    2005

  6. [14]

    Sampling-based algorithms for optimal motion planning,

    S. Karaman and E. Frazzoli, “Sampling-based algorithms for optimal motion planning,” The International Journal of Robotics Research , vol. 30, no. 7, pp. 846–894, 2011

    2011

  7. [15]

    Motion planning and control for mobile robot navigation using machine learning: a survey,

    X. Xiao, B. Liu, G. Warnell, and P. Stone, “Motion planning and control for mobile robot navigation using machine learning: a survey,” Autonomous Robots, vol. 46, no. 5, pp. 569–597, 2022

    2022

  8. [16]

    LeLaN: Learning a language-conditioned navigation policy from in- the-wild video,

    N. Hirose, C. Glossop, A. Sridhar, D. Shah, O. Mees, and S. Levine, “LeLaN: Learning a language-conditioned navigation policy from in- the-wild video,” in Conference on Robot Learning , 2024

    2024

  9. [17]

    Learning robotic navigation from experience: principles, methods and recent results,

    S. Levine and D. Shah, “Learning robotic navigation from experience: principles, methods and recent results,” Philosophical Transactions of the Royal Society B , vol. 378, no. 1869, p. 20210447, 2023

    2023

  10. [18]

    Gervet, S

    T. Gervet, S. Chintala, D. Batra, J. Malik, and D. S. Chaplot,

  11. [19]

    Improved deep reinforcement learning “Navigating to objects in the real world,

    H. Liu, Z. Huang, and C. Lv, “Improved deep reinforcement learning “Navigating to objects in the real world,” Science Robotics , vol. 8, no. 79, p. eadf6991, 2023. with expert demonstrations for urban autonomous driving,” IEEE Intelligent Vehicles Symposium, pp. 921–928, 2021

    2023

  12. [20]

    Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots,

    C. Chi, Z. Xu, C. Pan, E. Cousineau, B. Burchfiel, S. Feng, R. Tedrake, and S. Song, “Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots,” in Rob.: Sci. and Sys. , 2024

    2024

  13. [21]

    Goal-conditioned imitation learning,

    Y . Ding, C. Florensa, M. Phielipp, and P. Abbeel, “Goal-conditioned imitation learning,” Advances in Neural Info. Processing Sys. , 2019

    2019

  14. [22]

    Survey of learning-based approaches for robotic in-hand manipulation,

    A. I. Weinberg, O. Azulay, A. Shirizly, and A. Sintov, “Survey of learning-based approaches for robotic in-hand manipulation,”Frontiers in Robotics and AI , vol. 11, 2024

    2024

  15. [23]

    Improving autonomous robotic navigation using imitation learning,

    B. C `esar-Tondreau, G. Warnell, E. Stump, K. Kochersberger, and N. R. Waytowich, “Improving autonomous robotic navigation using imitation learning,” Frontiers in Rob. and AI, vol. 8, p. 627730, 2021

    2021

  16. [24]

    Coarse-to-fine imitation learning: Robot manipulation from a single demonstration,

    E. Johns, “Coarse-to-fine imitation learning: Robot manipulation from a single demonstration,” in IEEE International Conference on Robotics and Automation (ICRA) , 2021, pp. 4613–4619

    2021

  17. [25]

    GNM: A general navigation model to drive any robot,

    D. Shah, A. K. Sridhar, A. Bhorkar, N. Hirose, and S. Levine, “GNM: A general navigation model to drive any robot,” in IEEE International Conference on Robotics and Automation , 2023, pp. 7226–7233

    2023

  18. [26]

    ViNT: A foundation model for visual navigation,

    D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine, “ViNT: A foundation model for visual navigation,” in Annual Conference on Robot Learning , 2023

    2023

  19. [27]

    NoMaD: Goal masked diffusion policies for navigation and exploration,

    A. Sridhar, D. Shah, C. Glossop, and S. Levine, “NoMaD: Goal masked diffusion policies for navigation and exploration,” in IEEE Inter. Conference on Robotics and Automation , 2024, pp. 63–70

    2024

  20. [28]

    Diffusion policy: Visuomotor policy learning via action diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song, “Diffusion policy: Visuomotor policy learning via action diffusion,” The International Journal of Robotics Research , 2024

    2024

  21. [29]

    Goal conditioned imitation learning using score-based diffusion policies,

    M. Reuss, M. Li, X. Jia, and R. Lioutikov, “Goal conditioned imitation learning using score-based diffusion policies,” in Robotics: Science and Systems, 2023

    2023

  22. [30]

    UMI on legs: Making manipulation policies mobile with manipulation-centric whole-body controllers,

    H. Ha, Y . Gao, Z. Fu, J. Tan, and S. Song, “UMI on legs: Making manipulation policies mobile with manipulation-centric whole-body controllers,” in Conference on Robot Learning , 2024

    2024

  23. [31]

    Ldp: A local diffusion planner for efficient robot navigation and collision avoidance,

    W. Yu, J. Peng, H. Yang, J. Zhang, Y . Duan, J. Ji, and Y . Zhang, “Ldp: A local diffusion planner for efficient robot navigation and collision avoidance,” arXiv preprint arXiv:2407.01950 , 2024

    2024 arXiv

  24. [32]

    U-Net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer-Assisted Intervention , 2015, pp. 234–241

    2015

  25. [33]

    Efficientnet: Rethinking model scaling for con- volutional neural networks,

    M. Tan and Q. Le, “Efficientnet: Rethinking model scaling for con- volutional neural networks,” in International conference on machine learning, 2019, pp. 6105–6114

    2019

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.