Pith. sign in

REVIEW 3 major objections 6 minor 46 references

DefFusionNet: Learning Multimodal Goal Shapes for Deformable Object Manipulation via a Diffusion-based Probabilistic Model

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read DefFusionNet replaces deterministic goal prediction with a diffusion model that samples from the distribution of valid goal shapes, producing multimodal, physically plausible goals from as few as 10 demonstrations.

desk verdict A credible, incremental application of conditional diffusion to deformable-object goal generation, with a striking 10-demonstration result in simulation, but the physical-robot evidence is currently asserted rather than shown. read the letter →

arxiv 2506.18779 v1 pith:2XDPG7LI submitted 2025-06-23 cs.RO

Bao Thach , Siyeon Kim , Britton Jordan , Mohanraj Shanthi , Tanner Watts , Shing-Hei Ho , James M. Ferguson , Tucker Hermans
show 1 more author
Alan Kuntz
This is my paper
classification cs.RO
keywords deformableobjectmanipulationshapeservoingdiffusionprobabilisticmodelgoalgenerationmultimodaldistributionlearningfromdemonstrationpointcloudrobot
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

Shape servoing for deformable objects requires a goal shape, but specifying one is usually impractical. This paper argues that the goal should be treated as a sample from a learned distribution of all shapes that would complete the task, not as a unique deterministic output. DefFusionNet learns the conditional distribution $p_\theta(P_g \mid P_c, P_T)$ with a diffusion probabilistic model, conditioned on the current object point cloud and a task context. The paper shows that in simulated surgical retraction and object packaging, the generative model produces realistic, bimodal goals from as few as 10 demonstrations, outperforming the deterministic DefGoalNet trained on 1000, and it reports success on a physical robot.

What carries the argument

The central machinery is a conditional denoising diffusion probabilistic model: DefFusionNet. A PointNet encoder turns the current and contextual point clouds into feature vectors, a VAE-style encoder compresses the goal cloud into a latent vector $z$, and a noise-predictor network, structured with LSTM-style gates, estimates the noise added to a goal cloud at each diffusion step. Training minimizes a noise-reconstruction MSE plus a KL divergence that keeps the latent close to a unit Gaussian. At inference, reverse diffusion, conditioned on $z$ sampled from a Gaussian and on the encoded context, converts pure noise into a goal point cloud $P_g$, which is then executed by the separately trained DeformerNet controller without fine-tuning.

What would settle it

Run the full pipeline on a task with two clearly separated valid goal modes and track whether the robot's final shape matches the particular goal that was sampled. If for many samples the end shape is closer to the average of the two modes than to the sampled mode, or if task success does not depend on which goal was sampled, then the model either does not capture distinct modes or the controller cannot realize them. A concrete version: for 100 sampled goals from a bimodal context, measure the Chamfer distance between the reached cloud and the sampled goal versus the distance to the other mode; a median comparable to the cross-mode distance would falsify the claim that sampled goals are executed faithfully.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a diffusion-based generative model can replace deterministic goal prediction for deformable object manipulation. Given the current object point cloud $P_c$ and a task-specific contextual point cloud $P_T$, DefFusionNet learns the conditional distribution $p_\theta(P_g \mid P_c, P_T)$ over goal point clouds that lead to task success. At inference it samples a goal from this distribution and passes it to the pretrained DeformerNet controller, which carries out the deformation in a closed loop. In the experiments, DefFusionNet trained on 10 demonstrations achieves a median collision-avoidance rate above 70% and success percentage above 95% on simulated tissue retraction, beating DefGoalNet trained on 1000 demonstrations; it also outperforms DefGoalNet on object packaging. The authors describe this as the first generative model to produce a diverse, multimodal set of deformable object goals for real-world robotic applications.

Load-bearing premise

The load-bearing premise is that the terminal point clouds of the scripted demonstrations are genuinely valid goal shapes, and that the separately trained DeformerNet can deform the real object to whatever goal $P_g$ DefFusionNet samples, especially on physical hardware.

Editorial extensions

If this is right

  • Deployment of shape servoing no longer requires a manually engineered or manually demonstrated goal shape: the robot only needs the current object point cloud and a contextual cloud at runtime.
  • Tasks with multiple equally valid solutions, such as retracting tissue left or right, become tractable because the predictor outputs a distribution rather than an average that is physically meaningless.
  • Goal generation becomes sample-efficient: in these experiments 10 demonstrations suffice to beat a deterministic baseline trained on 1000, which lowers the cost of collecting demonstration data.
  • Dividing the problem into goal generation and goal-conditioned control lets each part be trained independently, with the controller trained at scale on cheap data and the goal generator on a small curated set.
  • The same approach should extend to continuous, high-variance goal distributions, as suggested by the packaging task where rotation angles are sampled uniformly over $[-\pi, \pi]$.

Reading between the lines

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

  • The decoupling of goal sampling from control suggests a testable extension: evaluate whether DeformerNet actually reaches the exact sampled goal, not just whether the task succeeds; if it does not, the effective goal distribution may be narrower than the modeled one and closed-loop goal correction would be needed.
  • Because all demonstrations are scripted and only terminal point clouds become goals, the learned distribution is anchored to the scripted policies; transferring to human demonstrations, which are noisier and more varied, could either produce richer multimodality or expose sensitivity to demonstration quality, an untested regime.
  • The idea of modeling a distribution over valid goals rather than a single action or state is not limited to point clouds; it could be applied to other goal representations such as target end-effector poses or mesh deformations, and to other deformable-object tasks like folding or cable routing, provided a conditional generator is available.
  • The 'first generative model' claim is specifically about deformable object goals in real-world robotic pipelines; earlier generative models for goal images or videos exist, so the novelty rests on applying generative point-cloud modeling within a shape-servoing loop, which is a narrow but concrete contribution.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes DefFusionNet, a diffusion-based conditional generative model that learns a distribution over valid goal point clouds for deformable-object manipulation, conditioned on the current object point cloud and a task-context point cloud. The generated goals are passed to a frozen DeformerNet controller for goal-conditioned shape servoing. The authors evaluate the approach in simulation on a surgical tissue-retraction task and an object-packaging task, comparing against their prior deterministic DefGoalNet baseline across several data-set sizes (10, 100, and 1000 demonstrations). They report that DefFusionNet with as few as 10 demonstrations outperforms DefGoalNet trained on 1000 demonstrations on collision avoidance, success percentage, coverage, and Chamfer distance, and they claim to demonstrate the approach on physical dVRK and KUKA robot hardware.

Significance. If the central claims hold, the paper addresses a real limitation of deterministic goal generation for deformable-object manipulation: multi-modality is common in practice, and averaging over modes can produce physically infeasible goals. The 10-demonstration result is striking because it suggests that a probabilistic model of goal distributions can be substantially more sample-efficient than a deterministic one for this problem. The authors also commit to releasing code and data, which is a concrete strength. However, the significance is tempered by the absence of quantitative physical-robot results, the lack of a second baseline, and a technical mismatch between the stated probabilistic sampling formulation and the actual inference procedure.

major comments (3)
  1. [Sec. V (intro) and Sec. IV-F] The abstract and Section V state that the method is demonstrated on a physical dVRK and on two KUKA iiwa arms, but the experiments subsections (V-B) report only simulation metrics, with no number of physical trials, task-success criteria, quantitative success rates, or failure cases for either hardware setup. Because the pipeline uses a frozen DeformerNet without fine-tuning, the simulation results do not establish that goals sampled by DefFusionNet are reachable by DeformerNet on real hardware. This gap is load-bearing for the claim of physical-robot validation and should be addressed by either providing the missing hardware results or by clearly limiting the paper's claim to simulation.
  2. [Sec. IV-E, Algorithm 2, and Eqs. (3)-(4)] The generative model is defined probabilistically in Eq. (3) as a Gaussian transition with variance β_t I, and Eq. (4) expresses the full generative distribution as a product of such transitions. Algorithm 2, however, performs a deterministic update that subtracts the predicted noise and never samples from those Gaussian transitions. This is a DDIM-like deterministic sampler, not the DDPM sampler implied by the formulation, and the discrepancy is not discussed. The diversity of the generated goals, which is central to the paper's contribution, depends on the actual sampling procedure, so the authors should either sample from Eq. (3) or explicitly state and justify the deterministic sampler.
  3. [Sec. IV-B, Eq. (1), and Sec. IV-C] Equation (1) defines the conditional goal distribution as a product over points, p_theta(X|c) = ∏_i p_theta(x_i|c), and the noise-predictor description in Section IV-C suggests that h is a series of fully-connected layers applied pointwise, with the global context injected only as a per-point gate and bias. If the denoising network has no cross-point interactions, then the joint structure of the generated point cloud is captured only through the single latent vector z and the global context features. The paper should clarify whether h contains any cross-point or attention layers; if it does not, it should justify why independent per-point generation is sufficient to represent coherent deformable-object goal shapes, since that is the core modeling claim of the paper.
minor comments (6)
  1. [Figs. 8-10 and 14-15] The figures should state the number of trials per box and define the box elements (median, quartiles, whiskers). In addition, the headline comparison of DefFusionNet-1000 versus DefGoalNet-1000 is based on a single training run; reporting variance across multiple seeds for all dataset sizes would strengthen the claim.
  2. [Sec. V-B.1] The success percentage metric is computed only for the subset of scenarios that already avoid collisions. This conditional metric can be misleading because a method that rarely avoids collisions may still show a high success percentage on its remaining runs; the authors should report joint success or report both components over all test scenarios.
  3. [Sec. V-B.2] The coverage percentage is described as the percentage of object volume contained by the container, but the manuscript does not specify how volume is estimated from the final point cloud (e.g., convex hull, occupancy grid, or another discretization). This should be clarified for reproducibility.
  4. [Sec. IV-A and Sec. III] The symbol T is used both for the task T in Section III and for the number of diffusion timesteps in Section IV-A. Disambiguating these notations (for example, using script T or a subscript) would prevent confusion.
  5. [Sec. I and Related Work] The claim that this is the first generative model for diverse multi-modal deformable-object goals should be supported by a more systematic comparison with prior goal-generation methods, including image- and language-conditioned generative goal models, rather than only a contrast with Diffusion Policy and SVBP.
  6. [Eq. (7)] The KL divergence loss is scaled by 1/T, where T is the diffusion timesteps. This scaling is not standard in VAE training and is not motivated; a brief explanation of the choice would be helpful.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular dependence found; the goal-generation model is trained and evaluated against external task metrics.

full rationale

DefFusionNet's goal generator is trained to predict terminal demonstration point clouds conditioned on the initial object and context, which is a standard conditional generative modeling task. The task-success metrics in Sec. V-B (collision avoidance rate, success percentage, coverage percentage) are defined by external geometric criteria, such as the tissue being fully retracted past the target plane without collision, not by whether the robot reaches the model's own sampled goal. The statement in Sec. I that successfully deforming the object to match the predicted goal is equivalent to task success is a framing statement; the actual evaluation does not rely on that equivalence. The frozen DeformerNet component is self-cited prior work, but it is exercised inside the paper's own simulation experiments, so the pipeline results are not imported solely by citation. DefGoalNet is used as a baseline trained on the same data, and the diffusion and VAE losses (Eqs. 6-7) are standard and not fitted to force the comparison. No equation in the paper reduces to its own input, and no fitted parameter is renamed as a prediction. The absence of quantitative physical-robot metrics is an evidence/completeness concern, not circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central contribution is empirical; the method inherits standard diffusion and VAE machinery rather than introducing new axioms. The main free parameters are architectural hyperparameters; the load-bearing domain assumptions concern the validity of scripted demonstrations as goals and the out-of-the-box transfer of DeformerNet.

free parameters (4)
  • Diffusion timesteps T = 100
    Number of forward/reverse diffusion steps, chosen from DDPM convention; no sensitivity study is provided.
  • VAE latent dimension = 256
    Dimensionality of the latent goal encoding, fixed by the PointNet encoder architecture.
  • KL loss scaling 1/T = 1/T
    Relative weight of the KL divergence term in the combined loss; chosen without ablation.
  • Goal point cloud size N = not reported
    Number of points in the generated goal clouds affects representation fidelity but is not stated in the paper.
assumptions (5)
  • standard math Diffusion probabilistic models as defined by Ho et al. and Luo and Hu correctly model the conditional goal distribution p_theta(X|c).
    The method assumes the DDPM formulation and its reverse-process sampling are valid for point clouds; no proof is given, taken from literature.
  • standard math Variational autoencoder reparameterization allows sampling z ~ q_phi(z|P_g) and optimizing the ELBO-style loss (L_noise-recon + L_KL).
    The VAE machinery is standard; the paper relies on the usual reparameterization trick and closed-form KL divergence.
  • domain assumption Demonstrations are multimodal and each terminal point cloud P_g is a valid goal; expert trajectories are all equally successful.
    Sec. III and IV-D define the dataset as containing equally effective demonstrations and extract goals from terminal point clouds; if some terminal states are not truly successful, the learned distribution is biased.
  • domain assumption DeformerNet, trained separately without fine-tuning, can reliably deform the object into any goal sampled from the learned distribution in both simulation and on physical hardware.
    Sec. IV-F explicitly uses the original DeformerNet without fine-tuning; if the controller cannot reach some sampled goals, the end-to-end success rates would be lower than reported.
  • domain assumption The contextual point cloud P_T encodes all task-relevant constraints (e.g., tool pose, container opening).
    The method conditions on P_T as the only context; if important constraints are not captured in the point cloud, goal generation may violate them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DefFusionNet: Learning Multimodal Goal Shapes for Deformable Object Manipulation via a Diffusion-based Probabilistic Model." pith.science (2026). https://pith.science/paper/2XDPG7LI

@misc{pith2026250618779,
  author       = {Pith},
  title        = {Pith review of: DefFusionNet: Learning Multimodal Goal Shapes for Deformable Object Manipulation via a Diffusion-based Probabilistic Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2XDPG7LI}},
  note         = {Machine review of arXiv:2506.18779}
}
read the original abstract

Deformable object manipulation is critical to many real-world robotic applications, ranging from surgical robotics and soft material handling in manufacturing to household tasks like laundry folding. At the core of this important robotic field is shape servoing, a task focused on controlling deformable objects into desired shapes. The shape servoing formulation requires the specification of a goal shape. However, most prior works in shape servoing rely on impractical goal shape acquisition methods, such as laborious domain-knowledge engineering or manual manipulation. DefGoalNet previously posed the current state-of-the-art solution to this problem, which learns deformable object goal shapes directly from a small number of human demonstrations. However, it significantly struggles in multi-modal settings, where multiple distinct goal shapes can all lead to successful task completion. As a deterministic model, DefGoalNet collapses these possibilities into a single averaged solution, often resulting in an unusable goal. In this paper, we address this problem by developing DefFusionNet, a novel neural network that leverages the diffusion probabilistic model to learn a distribution over all valid goal shapes rather than predicting a single deterministic outcome. This enables the generation of diverse goal shapes and avoids the averaging artifacts. We demonstrate our method's effectiveness on robotic tasks inspired by both manufacturing and surgical applications, both in simulation and on a physical robot. Our work is the first generative model capable of producing a diverse, multi-modal set of deformable object goals for real-world robotic applications.

Figures

Figures reproduced from arXiv: 2506.18779 by the authors.

Figure 1
Figure 1. Multimodal Contextual Shape Servoing: Given the task context and initial shape of the deformable object, DefFusionNet predicts multiple valid goal point clouds (red and green points) by learning a diffusion-based probabilistic model from human demonstrations. Conditioned on one of these predicted goals, DeformerNet then computes actions to manipulate the object into the desired shape and successfully accomplish the … view at source ↗
Figure 2
Figure 2. Simulated retraction results comparing predicted goals from [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. provides a comprehensive overview of the noise predictor architecture. We first transform scalar t to a 3D time￾embedding vector using its corresponding αt value. Next, we process and combine all features from z, ψc, and ψT using a series of concatenation operators and fully-connected layers, which eventually produces a 512-dimension feature vector. We concatenate this vector with the time-embedding vector to produc… view at source ↗
Figures from the paper (8 more)
Figure 7
Figure 7. Figure 7: Sample manipulation sequence on the surgical retraction task, in [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 6
Figure 6. Figure 6: Example demonstrations for the surgical retraction task. The black and [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 10
Figure 10. Figure 10: Simulated retraction results - Chamfer distance metric (lower is better) between predicted and ground truth goal point clouds, on the test set, across multiple training dataset sizes [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: Sample manipulation sequence on the object packaging task, in [PITH_FULL_IMAGE:figures/full_fig_p008_11.png]
Figure 9
Figure 9. Figure 9: Simulated retraction results - Success percentage metric. below. The first and most important class of evaluation metrics directly correlates with the task success rate. For each test configuration, we execute the DeformerNet policy conditioned on the predicted goal ge…
Figure 12
Figure 12. Figure 12: Example demonstrations for the object packaging task. For a [PITH_FULL_IMAGE:figures/full_fig_p008_12.png]
Figure 15
Figure 15. Figure 15: Simulated object packaging - Chamfer distance between predicted and ground truth goal point clouds, on the test set. again captures effectively the multimodal goal distribution, while DefGoalNet still produces physically impractical goal point clouds. DefFusionNet usi…
Figure 14
Figure 14. Figure 14: Simulated object packaging - Coverage percentage metric (higher is better). From left to right: DefGoalNet [7] (current state-of-the-art), followed by DefFusionNet (ours) trained with 500, 100, and 10 demonstrations. packaging task. Instead, the rotation angle is samp…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 33 canonical work pages

  1. [1]

    Toward Robotic Manipulation,

    M. T. Mason, “Toward Robotic Manipulation,”Annual Review of Con- trol, Robotics, and Autonomous Systems, vol. 1, no. 1, pp. 1–28, 2018

    work page 2018

  2. [2]

    Robotic Manipulation and Sensing of Deformable Objects in Domestic and Industrial Applications: A Survey,

    J. Sanchez, J. A. C. Ramon, B.-C. Bouzgarrou, and Y . Mezouar, “Robotic Manipulation and Sensing of Deformable Objects in Domestic and Industrial Applications: A Survey,”Intl. Journal of Robotics Re- search, vol. 37, no. 7, pp. 688–716, 2018

    work page 2018

  3. [3]

    DefGraspSim: Simulation-based grasping of 3D deformable objects,

    I. Huang, Y . Narang, C. Eppner, B. Sundaralingam, M. Macklin, T. Hermans, and D. Fox, “DefGraspSim: Simulation-based grasping of 3D deformable objects,” inRSS Workshop on Deformable Object Simulation in Robotics (DO-Sim), 2021

    work page 2021

  4. [4]

    Challenges and outlook in robotic manipulation of deformable objects,

    J. Zhu, A. Cherubini, C. Dune, D. Navarro-Alarcon, F. Alambeigi, D. Berenson, F. Ficuciello, K. Harada, J. Kober, X. Li,et al., “Challenges and outlook in robotic manipulation of deformable objects,”IEEE Robotics & Automation Magazine, vol. 29, no. 3, pp. 67–77, 2022. 10

    work page 2022

  5. [5]

    Learning to manipulate deformable objects without demonstrations,

    Y . Wu, W. Yan, T. Kurutach, L. Pinto, and P. Abbeel, “Learning to manipulate deformable objects without demonstrations,”Robotics: Science and Systems, 2020

    work page 2020

  6. [6]

    Deformernet: Learning bimanual manipulation of 3d deformable objects,

    B. Thach, B. Y . Cho, T. Hermans, and A. Kuntz, “Deformernet: Learning bimanual manipulation of 3d deformable objects,”ArXiv, 2023

    work page 2023

  7. [7]

    Defgoalnet: Contextual goal learning from demonstrations for deformable object manipulation,

    B. Thach, T. Watts, S.-H. Ho, T. Hermans, and A. Kuntz, “Defgoalnet: Contextual goal learning from demonstrations for deformable object manipulation,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 3145–3152

    work page 2024

  8. [8]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems, vol. 33, pp. 6840– 6851, 2020

    2020

Show all 46 references
  1. [9]

    Multi- fingered grasp planning via inference in deep neural networks: Outper- forming sampling by learning differentiable models,

    Q. Lu, M. Van der Merwe, B. Sundaralingam, and T. Hermans, “Multi- fingered grasp planning via inference in deep neural networks: Outper- forming sampling by learning differentiable models,”IEEE Robotics & Automation Magazine, vol. 27, no. 2, pp. 55–65, 2020

    2020

  2. [10]

    6-dof graspnet: Variational grasp generation for object manipulation,

    A. Mousavian, C. Eppner, and D. Fox, “6-dof graspnet: Variational grasp generation for object manipulation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 2901–2910

    2019

  3. [11]

    6-dof grasping for target-driven object manipulation in clutter,

    A. Murali, A. Mousavian, C. Eppner, C. Paxton, and D. Fox, “6-dof grasping for target-driven object manipulation in clutter,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 6232–6238

    2020

  4. [12]

    Self-supervised 6D Object Pose Estimation for Robot Manipulation,

    X. Deng, Y . Xiang, A. Mousavian, C. Eppner, T. Bretl, and D. Fox, “Self-supervised 6D Object Pose Estimation for Robot Manipulation,” IEEE Intl. Conf. on Robotics and Automation, pp. 3665–3671, 2020

    2020

  5. [13]

    Multi-fingered active grasp learning,

    Q. Lu, M. Van der Merwe, and T. Hermans, “Multi-fingered active grasp learning,” in2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 8415–8422

    2020

  6. [14]

    Learning continuous 3d reconstructions for geometrically aware grasping,

    M. Van der Merwe, Q. Lu, B. Sundaralingam, M. Matak, and T. Her- mans, “Learning continuous 3d reconstructions for geometrically aware grasping,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 11 516–11 522

    2020

  7. [15]

    Rt-1: Robotics transformer for real-world control at scale,

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, J. Dabis, C. Finn, K. Gopalakrishnan, K. Hausman, A. Herzog, J. Hsu,et al., “Rt-1: Robotics transformer for real-world control at scale,”arXiv preprint arXiv:2212.06817, 2022

    2022 arXiv

  8. [16]

    Visually servoed deformation control by robot manipulators,

    D. Navarro-Alarcon, Y . Liu, J. G. Romero, and P. Li, “Visually servoed deformation control by robot manipulators,” in2013 IEEE International Conference on Robotics and Automation. IEEE, 2013, pp. 5259–5264

    2013

  9. [17]

    Au- tonomous data-driven manipulation of unknown anisotropic deformable tissues using unmodelled continuum manipulators,

    F. Alambeigi, Z. Wang, R. Hegeman, Y .-H. Liu, and M. Armand, “Au- tonomous data-driven manipulation of unknown anisotropic deformable tissues using unmodelled continuum manipulators,”IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. 254–261, 2018

    2018

  10. [18]

    Automatic 3-d manipulation of soft objects by robotic arms with an adaptive deformation model,

    D. Navarro-Alarcon, H. M. Yip, Z. Wang, Y .-H. Liu, F. Zhong, T. Zhang, and P. Li, “Automatic 3-d manipulation of soft objects by robotic arms with an adaptive deformation model,”IEEE Transactions on Robotics, vol. 32, no. 2, pp. 429–441, 2016

    2016

  11. [19]

    Model predic- tive manipulation of compliant objects with multi-objective optimizer and adversarial network for occlusion compensation,

    J. Qi, D. Li, Y . Gao, P. Zhou, and D. Navarro-Alarcon, “Model predic- tive manipulation of compliant objects with multi-objective optimizer and adversarial network for occlusion compensation,”arXiv preprint arXiv:2205.09987, 2022

    2022 arXiv

  12. [20]

    A robust data-driven approach for online learning and manipulation of unmodeled 3-d heterogeneous compliant objects,

    F. Alambeigi, Z. Wang, R. Hegeman, Y .-H. Liu, and M. Armand, “A robust data-driven approach for online learning and manipulation of unmodeled 3-d heterogeneous compliant objects,”IEEE Robotics and Automation Letters, vol. 3, no. 4, pp. 4140–4147, 2018

    2018

  13. [21]

    Lattice- based shape tracking and servoing of elastic objects,

    M. Shetab-Bushehri, M. Aranda, Y . Mezouar, and E. Ozgur, “Lattice- based shape tracking and servoing of elastic objects,”arXiv preprint arXiv:2209.01832, 2022

    2022 arXiv

  14. [22]

    3-D Deformable Object Manipulation Using Deep Neural Networks,

    Z. Hu, T. Han, P. Sun, J. Pan, and D. Manocha, “3-D Deformable Object Manipulation Using Deep Neural Networks,”IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 4255–4261, 2019

    2019

  15. [23]

    Fast 3D recognition and pose using the Viewpoint Feature Histogram,

    R. B. Rusu, G. Bradski, R. Thibaux, and J. Hsu, “Fast 3D recognition and pose using the Viewpoint Feature Histogram,”IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, pp. 2155–2162, 2010

    2010

  16. [24]

    Learning visual shape control of novel 3d deformable objects from partial-view point clouds,

    B. Thach, B. Y . Cho, A. Kuntz, and T. Hermans, “Learning visual shape control of novel 3d deformable objects from partial-view point clouds,” inIEEE International Conference on Robotics and Automation (ICRA), 2022, pp. 8274–8281

    2022

  17. [25]

    Pointconv: Deep convolutional networks on 3d point clouds,

    W. Wu, Z. Qi, and L. Fuxin, “Pointconv: Deep convolutional networks on 3d point clouds,”ArXiv, 2018

    2018

  18. [26]

    3d point cloud generative adversarial network based on tree structured graph convolutions,

    D. W. Shu, S. W. Park, and J. Kwon, “3d point cloud generative adversarial network based on tree structured graph convolutions,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 3859–3868

    2019

  19. [27]

    Learning rep- resentations and generative models for 3d point clouds,

    P. Achlioptas, O. Diamanti, I. Mitliagkas, and L. Guibas, “Learning rep- resentations and generative models for 3d point clouds,” inInternational conference on machine learning. PMLR, 2018, pp. 40–49

    2018

  20. [28]

    A progressive conditional generative adversarial network for generating dense and colored 3d point clouds,

    M. S. Arshad and W. J. Beksi, “A progressive conditional generative adversarial network for generating dense and colored 3d point clouds,” in2020 International Conference on 3D Vision (3DV). IEEE, 2020, pp. 712–722

    2020

  21. [29]

    C-flow: Conditional generative flow models for images and 3d point clouds,

    A. Pumarola, S. Popov, F. Moreno-Noguer, and V . Ferrari, “C-flow: Conditional generative flow models for images and 3d point clouds,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 7949–7958

    2020

  22. [30]

    Large-scale scenario generation for robotic manipulation via conditioned generative models,

    S. van Waveren, C. Pek, I. Leite, J. Tumova, and D. Kragic, “Large-scale scenario generation for robotic manipulation via conditioned generative models,” 2022

    2022

  23. [31]

    Learning universal policies via text-guided video generation,

    Y . Du, S. Yang, B. Dai, H. Dai, O. Nachum, J. Tenenbaum, D. Schuur- mans, and P. Abbeel, “Learning universal policies via text-guided video generation,”Advances in neural information processing systems, vol. 36, pp. 9156–9172, 2023

    2023

  24. [32]

    What matters in language conditioned robotic imitation learning over unstructured data,

    O. Mees, L. Hermann, and W. Burgard, “What matters in language conditioned robotic imitation learning over unstructured data,”IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 11 205–11 212, 2022

    2022

  25. [33]

    Bridging language and action: A survey of language-conditioned robot manipulation,

    H. Zhou, X. Yao, O. Mees, Y . Meng, T. Xiao, Y . Bisk, J. Oh, E. Johns, M. Shridhar, D. Shah,et al., “Bridging language and action: A survey of language-conditioned robot manipulation,”arXiv e-prints, pp. arXiv– 2312, 2023

    2023

  26. [34]

    Diffusion policy: Visuomotor policy learning via ac- tion diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song, “Diffusion policy: Visuomotor policy learning via ac- tion diffusion,”The International Journal of Robotics Research, p. 02783649241273668, 2023

    2023

  27. [35]

    Stein variational belief propagation for multi-robot coordination,

    J. Pavlasek, J. J. Z. Mah, R. Xu, O. C. Jenkins, and F. Ramos, “Stein variational belief propagation for multi-robot coordination,”IEEE Robotics and Automation Letters, 2024

    2024

  28. [36]

    Droid: A large-scale in-the-wild robot manipulation dataset,

    A. Khazatsky, K. Pertsch, S. Nair, A. Balakrishna, S. Dasari, S. Karam- cheti, S. Nasiriany, M. K. Srirama, L. Y . Chen, K. Ellis,et al., “Droid: A large-scale in-the-wild robot manipulation dataset,”arXiv preprint arXiv:2403.12945, 2024

    2024 arXiv

  29. [37]

    Make-a-video: Text-to-video generation without text-video data,

    U. Singer, A. Polyak, T. Hayes, X. Yin, J. An, S. Zhang, Q. Hu, H. Yang, O. Ashual, O. Gafni,et al., “Make-a-video: Text-to-video generation without text-video data,”arXiv preprint arXiv:2209.14792, 2022

    2022 arXiv

  30. [38]

    Noise2music: Text-conditioned music generation with diffusion models,

    Q. Huang, D. S. Park, T. Wang, T. I. Denk, A. Ly, N. Chen, Z. Zhang, Z. Zhang, J. Yu, C. Frank,et al., “Noise2music: Text-conditioned music generation with diffusion models,”arXiv preprint arXiv:2302.03917, 2023

    2023 arXiv

  31. [39]

    Diffusion probabilistic models for 3d point cloud generation,

    S. Luo and W. Hu, “Diffusion probabilistic models for 3d point cloud generation,” pp. 2837–2845, 2021

    2021

  32. [40]

    Auto-encoding variational bayes,

    D. P. Kingma, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013

    2013 arXiv

  33. [41]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,”Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 652–660, 2017

    2017

  34. [42]

    Long short-term memory,

    S. Hochreiter, “Long short-term memory,”Neural Computation MIT- Press, 1997

    1997

  35. [43]

    An open-source research kit for the da Vinci® Surgical System,

    P. Kazanzides, Z. Chen, A. Deguet, G. S. Fischer, R. H. Taylor, and S. P. DiMaio, “An open-source research kit for the da Vinci® Surgical System,” inProc. IEEE Int. Conf. Robotics and Automation (ICRA), May 2014, pp. 6434–6439

    2014

  36. [44]

    GPU-Accelerated Robotic Simulation for Distributed Reinforcement Learning,

    J. Liang, V . Makoviychuk, A. Handa, N. Chentanez, M. Macklin, and D. Fox, “GPU-Accelerated Robotic Simulation for Distributed Reinforcement Learning,”arXiv:1810.05762, 2018

    2018 arXiv

  37. [45]

    Central airway obstruction,

    A. Ernst, D. Feller-Kopman, H. D. Becker, and A. C. Mehta, “Central airway obstruction,”American journal of respiratory and critical care medicine, vol. 169, no. 12, pp. 1278–1297, 2004

    2004

  38. [46]

    A point set generation network for 3d object reconstruction from a single image,

    H. Fan, H. Su, and L. J. Guibas, “A point set generation network for 3d object reconstruction from a single image,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 605–613

    2017

Pith tools

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