23 DoF Grasping Policies from a Raw Point Cloud

Karl Van Wyk; Martin Matak; Tucker Hermans

REVIEW 3 major objections 5 minor 29 references

23 DoF Grasping Policies from a Raw Point Cloud

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A Neural Geometric Fabric policy directly predicts 23-DoF joint accelerations from a raw point cloud and, paired with a motion planner, grasps objects not seen in policy training.

desk verdict Solid NGF-to-point-cloud extension, but the 'unseen object' claim hinges on an unnamed autoencoder object set. read the letter →

arxiv 2411.14400 v1 pith:FTIFB54C submitted 2024-11-21 cs.RO

Martin Matak , Karl Van Wyk , Tucker Hermans This is my paper · ORCID

classification cs.RO

keywords dexterousgraspinggeometricfabricsneuralpoliciesimitationlearningpointcloudencodinghigh-DoFmanipulationDAgger

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

The paper claims that a robot hand-arm system can learn to grasp objects it has never seen by directly predicting joint accelerations from a single camera's partial point cloud. At the core is a Neural Geometric Fabric policy, a second-order dynamical system whose stability comes from geometric fabric theory, trained by imitating trajectories generated by a geometric fabric planner with no human demonstrations. At test time, a motion planner brings the hand near the object, the policy executes the grasp, and the planner lifts the object. In simulation the policy outperforms an MLP baseline, and conditioning on the latent point-cloud encoding improves success over conditioning on object position alone. If correct, this points toward high-frequency, high-degree-of-freedom grasping policies that avoid per-object optimization at run time.

What carries the argument

The Neural Geometric Fabric (NGF): a geometric fabric, a stable second-order differential equation of the form $\ddot{q} = e_h - M^{-1}\partial_q\psi - B\dot{q}$, whose metric, potential, and damping are neural-network functions of joint state and a latent object encoding. The fabric's stability guarantees and path consistency let the learned accelerations behave like a controller rather than an open-loop trajectory; the frozen point-cloud autoencoder supplies the object-conditioning feature $z$. The motion-planner-in-the-loop design is the other central mechanism: it restricts learning to the challenging grasp segment while traditional planning handles approach and lift.

What would settle it

Check the four YCB objects used to train the point-cloud autoencoder in Section III-D: if 'mustard' is among them, rerun the 100-trial mustard grasp evaluation with mustard excluded from encoder training. A materially lower success rate would show the reported generalization conflates grasp-policy novelty with encoder familiarity.

Watch

Extended reading notes

Core claim

The central discovery is that a Neural Geometric Fabric policy can output 23-dimensional joint accelerations that form smooth grasping trajectories directly from the latent code of a partial point cloud, and these trajectories generalize to an object whose point cloud was not part of grasp-policy training. The paper builds the fabric by parameterizing its metric, driving force, and damping with neural networks conditioned on a frozen PointNet-based autoencoder encoding of the scene. Trained with DAgger against a surrogate PD expert built from geometric-fabric-generated trajectories, the policy is evaluated in simulation on three objects; it consistently beats an MLP baseline, and point-cloud conditioning improves success over position-only conditioning. The authors combine the learned policy with a geometric fabric motion planner in a loop, using the planner for approach and lift while the policy handles only the contact-rich grasp, and report that this setup grasps a previously unseen object.

Load-bearing premise

The result depends on the test object 'mustard' not being one of the four YCB objects used to train the frozen point-cloud encoder; the paper never lists those four objects, so the 'previously unseen object' claim cannot be fully verified from the text.

Editorial extensions

If this is right

High-DoF grasping can be cast as a learned second-order dynamical system rather than a per-scene optimization, so run-time grasp synthesis could be replaced by a single 30 Hz policy.
Point-cloud encodings carry geometric information that improves grasping over position-only inputs, supporting the idea that raw perception can directly condition contact-rich policies.
Because the fabric structure provides stability, the learned policy only needs to imitate the expert's accelerations rather than enforce constraint satisfaction from scratch.
The same training pipeline, fabric-generated trajectories plus DAgger, could in principle be applied to other high-DoF manipulation skills without human demonstrations.
Generalization over object shape and pose opens a route to multi-object grasping with a single policy, provided the encoder generalizes across the object distribution.

Reading between the lines

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

The strongest interpretation of "novel object" applies only to the grasp policy; the frozen point-cloud encoder was trained on four YCB objects, so the paper's generalization claim is only fully about novel objects if the test object was not among those four, a detail the paper does not disclose.
If the encoder had seen the test object, the result still demonstrates grasp-policy generalization but not end-to-end perception generalization; a clean test would retrain the encoder without the test object.
The approach's reliance on a dataset lookup for the pre-grasp configuration means the method does not yet solve full pick-and-place from arbitrary configurations; learning that retrieval is an obvious next step the paper acknowledges.
Because the fabric expert itself only succeeds on a fraction of its own training trajectories (59% for one object), the policy's ceiling is coupled to expert data quality; improving the expert's success rate before collecting training data should raise the learned policy's success rate.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Here's the take: this is a solid extension of Neural Geometric Fabrics to raw point clouds, and the training pipeline is clean—but the headline claim about grasping a previously unseen object is undercut by a missing disclosure.

What's actually new: the paper trains an NGF policy from geometric-fabric-generated trajectories without human demonstrations, conditions it on a frozen PointNet autoencoder latent from a single partial point cloud, and deploys it in a loop with an off-the-shelf motion planner. The DAgger setup with a surrogate PD expert is sensible, and the ablations comparing NGF vs. MLP and position vs. point-cloud encodings are informative. NGF consistently beats the MLP baseline, and the point-cloud encoding helps over position-only input. That is a credible step for high-DoF grasping.

The soft spots are real, and the biggest one is exactly what the stress-test note flags: Section III-D says the autoencoder is trained on four YCB objects but never names them. The evaluation object "mustard" is labeled "not seen" in Fig. 5, but "not seen" is ambiguous—it might mean not in the grasp-policy training set. If mustard was among the four autoencoder objects, then the perception module has already seen mustard point clouds, and the experiment only tests generalization of the grasp head, not generalization to novel object shape. That is a load-bearing omission, not a cosmetic one.

Other concerns: only three objects total, one of which is the "novel" one; the dataset itself only yields 59% successful lifts for sugarbox; no error bars or repeated-seed statistics are reported; the PD gains for the surrogate expert and other hyperparameters are not given; and there is no code or data release. These are all addressable in a revision. The math and fabric construction look fine, equations 1–2 are standard, and the citation pattern is appropriate—prior NGF work is credited.

Net: this deserves a serious referee. The core idea is plausible and the experimental setup is identifiable, but the autoencoder object set must be disclosed and the novelty claim re-scoped or re-tested. A conditional accept with requested revisions is the right outcome.

Referee Report

3 major / 5 minor

Summary. The paper proposes an imitation-learning pipeline for 23-DoF grasping from a single fixed RGBD camera. The authors first generate successful grasp configurations using a DexGrasp-1M-derived grasp set and a geometric-fabric motion planner, then reverse the resulting trajectories and train a Neural Geometric Fabric (NGF) policy, with an MLP baseline, to predict joint accelerations. Object information is supplied either by object position or by a frozen PointNet-based autoencoder latent encoding of the partial point cloud. Experiments compare NGF and MLP with position and point-cloud encodings on three objects, of which mustard is labeled as not seen.

Significance. If the empirical claims hold, the paper would support a useful design: a second-order fabric-structured policy that can be trained without human demonstrations and deployed in a loop with an off-the-shelf motion planner, with the learned component handling only the final grasping motion. The expert construction and loss in Eqs. (1)-(2) are standard, and the NGF parameterization is a coherent extension of geometric fabrics. The paper also gives credit to the geometric-fabric literature and clearly separates the learned grasping component from the motion-planner loop. However, the evidence for the headline claim of generalization to novel object geometry is incomplete: the autoencoder's training objects are not disclosed, no uncertainty or significance measures are reported, and the underlying dataset for one training object has only a 59% successful-lift rate.

major comments (3)

[Section III-D and Section IV, Fig. 5] The autoencoder is trained on point clouds of '4 YCB objects' but the objects are not named. The evaluation object 'mustard' is labeled '(not seen)' in Fig. 5, but the label only says it was not seen by the grasp-policy training; the paper never states whether mustard was among the four autoencoder training objects. If mustard was included, the 'novel object' experiment tests generalization of the grasp head but not generalization of the perception-to-grasp system to an unseen shape, which is the central contribution claimed in the abstract and introduction. Please disclose the four objects and, if necessary, retrain the autoencoder with mustard excluded before re-evaluating.
[Section IV, dataset note] The paper reports that only 59% of the generated trajectories for 'sugarbox' successfully lift the object, versus 93% for 'bleach'. Since the policies are trained to imitate these trajectories, the upper bound on achievable success for sugarbox is at most 59% under the current data-generation process, independently of policy architecture. This directly qualifies the introductory claim that the model 'reliably predicts a smooth 23-DoF grasping trajectory' and should be discussed explicitly as an upper bound and addressed, for example by filtering the training set to successful demonstrations or by reporting per-object ceiling performance.
[Section IV and Fig. 5] No error bars, confidence intervals, or statistical tests are reported for the 100-trial success rates. For a binary outcome with n=100, the 95% Wilson interval has half-width around 5-10 percentage points, so the reported differences between NGF-PCD and MLP-PCD, and between PCD and POS encodings, may not be significant. Please report exact counts, intervals, and tests (or at least multiple random seeds) for each bar.

minor comments (5)

[Section III-A] The word 'postive' in the description of the damping matrix should be 'positive'.
[Fig. 3 caption] The caption says 'course geometry' but should read 'coarse geometry'.
[Section III-B and Eq. (5)] In Eq. (5), B is introduced as a positive semi-definite damping matrix, but the definition B = beta_f M^{-1} dot-q appears to produce a vector rather than a matrix. Please clarify the intended construction.
[Section II-A] The notation K is used in |D_tau| = M K but K is not defined; it should be explicitly tied to the number of object poses |G| times the number of objects |O|.
[Section III-D] Training details of the autoencoder, such as latent dimension, number of epochs, regularization weight, and point-cloud preprocessing, are missing and are needed for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the policy is trained by imitation of fabric-generated trajectories and evaluated by physical grasp success, with no success signal used as a training input.

full rationale

The paper's derivation chain is empirical rather than formal. Grasp trajectories are generated in simulation by geometric fabrics [20], reversed, and stored; a PD controller (Eq. 1) forms the surrogate expert, and the policy is trained by DAgger to match the expert's accelerations (Eq. 2). The expert's target is a dataset trajectory, not the grasp-success label; success is measured only after execution in Section IV and is never used as a training objective. Hence no prediction reduces by construction to a fitted parameter or to the training signal. The NGF architecture is inherited from prior work [21], but this is a structural choice tested directly against an MLP baseline in Section IV and Figure 5, so the self-citation is not load-bearing. The point-cloud autoencoder in Section III-D is trained on a reconstruction task and then frozen; its latent vector is an input feature to the policy, not a prediction of grasp outcome. One disclosure gap should be flagged: the paper says the autoencoder was trained on '4 YCB [26] objects' without naming them, and Figure 5 labels 'mustard' as '(not seen)' while Section IV states the grasp policies were trained on bleach and sugarbox only. If mustard was among the four autoencoder objects, the 'previously unseen object' claim would be weakened for perception generalization. This is an experimental-support ambiguity, not a circularity by construction: no equation or fitting step equates the test output to the training data. Score 0.

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

The central result depends on standard stability results for geometric fabrics, a set of simulation fidelity assumptions, and several hand-chosen hyperparameters. The autoencoder and policy weights are learned from data, but key hyperparameters such as PD gains and latent dimension are unreported, which increases the burden on the reader for reproduction.

free parameters (6)

Additional damping scalar beta = 5
Hand-chosen in Section III-B to stabilize training; directly alters the final acceleration equation (Eq. 3) and is not derived from task requirements.
Number of grasp poses per object N = 7
Chosen in Section II-A to reduce variance in the grasp distribution; arbitrary and affects dataset diversity.
Number of trajectories per object pose M = 256
Dataset size choice in Section II-A; affects coverage of the approach region.
PD gains kp, kd in surrogate expert = not reported
Eq. 1 defines the expert with unspecified gains; the imitation signal depends critically on these values.
Autoencoder latent dimension = not reported
Section III-D does not specify the latent size, which controls how much geometric information is preserved for the policy input.
Number of random Fourier features = 1000
Used in the F_theta encoders in Section III-B; a design choice affecting function approximation capacity.

assumptions (5)

standard math Geometric fabrics are provably stable and produce path-consistent motion (citing [20])
Used in Section III-A as the foundation for the NGF policy structure and for the motion planner generating training trajectories.
domain assumption The simulator physics and fixed-camera point cloud faithfully represent the real robot environment
All training and evaluation are in simulation (Section IV); no real-robot validation is reported.
domain assumption The surrogate expert with PD control can reliably track dataset trajectories and provides a valid teaching signal for DAgger
Equation (1) in Section II-B assumes the PD controller with unspecified gains will collapse onto the dataset trajectory over successive executions.
domain assumption The point-cloud autoencoder trained with Chamfer distance preserves information sufficient for grasping
Section III-D and Fig. 3 show reconstructions are imperfect; the experiments assume the latent encoding still captures pose and gross geometry useful for policy learning.
ad hoc to paper The sampled start region above the object (based on bounding box) is sufficient for the policy to generalize
Section II-A states this region is a design choice and any other region could be used; the policy is only evaluated on starts from this region.

how reviews work

0 comments

Cite this review

Pith. "Pith review of 23 DoF Grasping Policies from a Raw Point Cloud." pith.science (2026). https://pith.science/paper/FTIFB54C

@misc{pith2026241114400,
  author       = {Pith},
  title        = {Pith review of: 23 DoF Grasping Policies from a Raw Point Cloud},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FTIFB54C}},
  note         = {Machine review of arXiv:2411.14400}
}

read the original abstract

Coordinating the motion of robots with high degrees of freedom (DoF) to grasp objects gives rise to many challenges. In this paper, we propose a novel imitation learning approach to learn a policy that directly predicts 23 DoF grasp trajectories from a partial point cloud provided by a single, fixed camera. At the core of the approach is a second-order geometric-based model of behavioral dynamics. This Neural Geometric Fabric (NGF) policy predicts accelerations directly in joint space. We show that our policy is capable of generalizing to novel objects, and combine our policy with a geometric fabric motion planner in a loop to generate stable grasping trajectories. We evaluate our approach on a set of three different objects, compare different policy structures, and run ablation studies to understand the importance of different object encodings for policy learning.

Figures

Figures reproduced from arXiv: 2411.14400 by the authors.

**Figure 1.** Our approach predicts 23 DoF trajectories based on raw point cloud and robot state (position and velocity) to grasp the object of interest. Some examples of resulting grasps are shown in top right. An example of a generated trajectory is shown in the bottom part of the figure. Abstract— Coordinating the motion of robots with high degrees of freedom (DoF) to grasp objects gives rise to many challenges. In this paper,… view at source ↗

**Figure 2.** Region which we sample from above the object is computed based on the object’s bounding box. Left: Training data is collected by generating trajectories from the grasp configuration to the sampled poses and then reversing those trajectories. Right: At evaluation time, we sample a target pose above the object and use a motion planner to get there. Then, the policy predicts a trajectory to grasp the object. B. Surroga… view at source ↗

**Figure 3.** Example point cloud reconstruction; input in blue, output in red. While imperfect (left), it preserves information about object pose and course geometry (right). C. Acceleration MLP We implement an MLP-based acceleration policy as a baseline. Input and output of the policy is the same as for NGF, i.e. input joint position, velocity, and object encoding and output joint acceleration. The input layer is followed by tw… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Top: The red lines show the camera location and workspace boundaries. The objects are (left to right): ‘mustard’, ‘sugarbox’, ‘bleach’. Bottom: Our manipulation pipeline from left to right: a motion planner moves the robot close to the object, our policy grasps the obj…

**Figure 5.** Figure 5: Grasp success rates for different grasp-trajectory predicting models and the dataset we use for training. Not all samples in the dataset are successful grasps. Input modalities: POS-position only, PCD - pointcloud. the partial view point cloud produces useful features …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 21 canonical work pages

[1]

Hand posture subspaces for dexterous robotic grasping,

M. T. Ciocarlie and P. K. Allen, “Hand posture subspaces for dexterous robotic grasping,” Intl. Journal of Robotics Research , vol. 28, no. 7, 2009

work page 2009
[2]

A probabilistic framework for uncertainty-aware high-accuracy precision grasping of unknown objects,

D. Chen, V . Dietrich, Z. Liu, and G. von Wichert, “A probabilistic framework for uncertainty-aware high-accuracy precision grasping of unknown objects,” Journal of Intelligent & Robotic Systems , vol. 90, no. 1, 2018

work page 2018
[3]

Examples of 3d grasp quality computations,

A. Miller and P. Allen, “Examples of 3d grasp quality computations,” in IEEE Intl. Conf. on Robotics and Automation , vol. 2, 1999, pp. 1240–1246 vol.2

work page 1999
[4]

Computation of independent contact regions for grasping 3-d objects,

M. A. Roa and R. Suarez, “Computation of independent contact regions for grasping 3-d objects,” IEEE Trans. on Robotics , vol. 25, no. 4, 2009

work page 2009
[5]

Synthesizing grasp configurations with specified contact regions,

C. Rosales, L. Ros, J. M. Porta, and R. Su ´arez, “Synthesizing grasp configurations with specified contact regions,”Intl. Journal of Robotics Research, vol. 30, no. 4, 2011

work page 2011
[6]

Coping with the grasping uncertainties in force-closure analysis,

Y . Zheng and W.-H. Qian, “Coping with the grasping uncertainties in force-closure analysis,” Intl. Journal of Robotics Research , vol. 24, no. 4, 2005

work page 2005
[7]

Hierarchical fingertip space: A unified framework for grasp planning and in-hand grasp adaptation,

K. Hang, M. Li, J. A. Stork, Y . Bekiroglu, F. T. Pokorny, A. Billard, and D. Kragic, “Hierarchical fingertip space: A unified framework for grasp planning and in-hand grasp adaptation,” IEEE Trans. on Robotics, vol. 32, no. 4, 2016

work page 2016
[8]

Grasp stability prediction for a dexterous robotic hand combining depth vision and haptic bayesian exploration,

M. S. Siddiqui, C. Coppola, G. Solak, and L. Jamone, “Grasp stability prediction for a dexterous robotic hand combining depth vision and haptic bayesian exploration,” Frontiers in Robotics and AI , 2021

work page 2021

Show all 29 references

[9]

Planning Multi-Fingered Grasps as Probabilistic Inference in a Learned Deep Network,

Q. Lu, K. Chenna, B. Sundaralingam, and T. Hermans, “Planning Multi-Fingered Grasps as Probabilistic Inference in a Learned Deep Network,” in Intl. Symposium on Robotics Research , 2017

2017
[10]

Modeling Grasp Type Improves Learning- Based Grasp Planning,

Q. Lu and T. Hermans, “Modeling Grasp Type Improves Learning- Based Grasp Planning,” IEEE Robotics and Automation Letters , 2019

2019
[11]

Multi- fingered grasp planning via inference in deep neural networks,

Q. Lu, M. V . der Merwe, B. Sundaralingam, and T. Hermans, “Multi- fingered grasp planning via inference in deep neural networks,” IEEE Robotics & Automation Magazine , 2020

2020
[12]

Multi-Fingered Active Grasp Learning,

Q. Lu, M. V . der Merwe, and T. Hermans, “Multi-Fingered Active Grasp Learning,” in IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems, 2020

2020
[13]

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,” in IEEE Intl. Conf. on Robotics and Automation , 2020

2020
[14]

Planning visual-tactile precision grasps via complementary use of vision and touch,

M. Matak and T. Hermans, “Planning visual-tactile precision grasps via complementary use of vision and touch,” IEEE Robotics and Automation Letters, vol. 8, no. 2, pp. 768–775, 2023

2023
[15]

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,” in Intl. Conf. on Computer Vision, 2019

2019
[16]

Dynamical movement primitives: learning attractor models for motor behaviors,

A. J. Ijspeert, J. Nakanishi, H. Hoffmann, P. Pastor, and S. Schaal, “Dynamical movement primitives: learning attractor models for motor behaviors,” Neural computation, vol. 25, no. 2, pp. 328–373, 2013

2013
[17]

Rie- mannian motion policies,

N. D. Ratliff, J. Issac, D. Kappler, S. Birchfield, and D. Fox, “Rie- mannian motion policies,” arXiv preprint arXiv:1801.02854 , 2018

2018 arXiv
[18]

Rmp flow: A computational graph for automatic mo- tion policy generation,

C.-A. Cheng, M. Mukadam, J. Issac, S. Birchfield, D. Fox, B. Boots, and N. Ratliff, “Rmp flow: A computational graph for automatic mo- tion policy generation,” in Algorithmic Foundations of Robotics XIII: Proceedings of the 13th Workshop on the Algorithmic Foundations of Robot...

2020
[19]

A unified approach for motion and force control of robot manipulators: The operational space formulation,

O. Khatib, “A unified approach for motion and force control of robot manipulators: The operational space formulation,” IEEE Journal on Robotics and Automation , vol. 3, no. 1, pp. 43–53, 1987

1987
[20]

Geometric fabrics: Generalizing classical mechanics to capture the physics of behavior,

K. Van Wyk, M. Xie, A. Li, M. A. Rana, B. Babich, B. Peele, Q. Wan, I. Akinola, B. Sundaralingam, D. Fox, B. Boots, and N. D. Ratliff, “Geometric fabrics: Generalizing classical mechanics to capture the physics of behavior,” IEEE Robotics and Automation Letters , vol. 7, no. 2...

2022
[21]

Neural geometric fabrics: Efficiently learning high-dimensional policies from demonstration,

M. Xie, A. Handa, S. Tyree, D. Fox, H. Ravichandar, N. D. Ratliff, and K. V . Wyk, “Neural geometric fabrics: Efficiently learning high-dimensional policies from demonstration,” in 6th Annual Conference on Robot Learning , 2022. [Online]. Available: https://openreview.net/foru...

2022
[22]

Learning robust real-world dexterous grasping policies via implicit shape augmentation,

Q. Chen, K. V . Wyk, Y .-W. Chao, W. Yang, A. Mousavian, A. Gupta, and D. Fox, “Learning robust real-world dexterous grasping policies via implicit shape augmentation,” in 6th Annual Conference on Robot Learning , 2022. [Online]. Available: https: //openreview.net/forum?id=bUUf1CT1sNu

2022
[23]

A reduction of imitation learning and structured prediction to no-regret online learning,

S. Ross, G. Gordon, and D. Bagnell, “A reduction of imitation learning and structured prediction to no-regret online learning,” in Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics , ser. Proceedings of Machine Learning Research, ...

2011
[24]

Dexgrasp-1m: Dexterous multi-finger grasp generation through differentiable simulation,

D. Turpin, T. Zhong, S. Zhang, G. Zhu, E. Heiden, M. Macklin, S. Tsogkas, S. Dickinson, and A. Garg, “Dexgrasp-1m: Dexterous multi-finger grasp generation through differentiable simulation,” in IEEE Intl. Conf. on Robotics and Automation , 2023

2023
[25]

Generalized nonlinear and finsler geometry for robotics,

N. D. Ratliff, K. Van Wyk, M. Xie, A. Li, and M. A. Rana, “Generalized nonlinear and finsler geometry for robotics,” in 2021 IEEE International Conference on Robotics and Automation (ICRA) , 2021, pp. 10 206–10 212

2021
[26]

The ycb object and model set: Towards common benchmarks for manipulation research,

B. Calli, A. Singh, A. Walsman, S. Srinivasa, P. Abbeel, and A. M. Dollar, “The ycb object and model set: Towards common benchmarks for manipulation research,” in 2015 International Conference on Advanced Robotics (ICAR) , 2015, pp. 510–517

2015
[27]

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

R. Q. Charles, H. Su, M. Kaichun, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 77–85

2017
[28]

Fast graph representation learning with PyTorch Geometric,

M. Fey and J. E. Lenssen, “Fast graph representation learning with PyTorch Geometric,” in ICLR Workshop on Representation Learning on Graphs and Manifolds , 2019

2019
[29]

Deconstructing the inductive biases of hamiltonian neural networks,

N. Gruver, M. Finzi, S. Stanton, and A. G. Wilson, “Deconstructing the inductive biases of hamiltonian neural networks,” arXiv preprint arXiv:2202.04836, 2022

2022 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Hand posture subspaces for dexterous robotic grasping,

M. T. Ciocarlie and P. K. Allen, “Hand posture subspaces for dexterous robotic grasping,” Intl. Journal of Robotics Research , vol. 28, no. 7, 2009

work page 2009

[2] [2]

A probabilistic framework for uncertainty-aware high-accuracy precision grasping of unknown objects,

D. Chen, V . Dietrich, Z. Liu, and G. von Wichert, “A probabilistic framework for uncertainty-aware high-accuracy precision grasping of unknown objects,” Journal of Intelligent & Robotic Systems , vol. 90, no. 1, 2018

work page 2018

[3] [3]

Examples of 3d grasp quality computations,

A. Miller and P. Allen, “Examples of 3d grasp quality computations,” in IEEE Intl. Conf. on Robotics and Automation , vol. 2, 1999, pp. 1240–1246 vol.2

work page 1999

[4] [4]

Computation of independent contact regions for grasping 3-d objects,

M. A. Roa and R. Suarez, “Computation of independent contact regions for grasping 3-d objects,” IEEE Trans. on Robotics , vol. 25, no. 4, 2009

work page 2009

[5] [5]

Synthesizing grasp configurations with specified contact regions,

C. Rosales, L. Ros, J. M. Porta, and R. Su ´arez, “Synthesizing grasp configurations with specified contact regions,”Intl. Journal of Robotics Research, vol. 30, no. 4, 2011

work page 2011

[6] [6]

Coping with the grasping uncertainties in force-closure analysis,

Y . Zheng and W.-H. Qian, “Coping with the grasping uncertainties in force-closure analysis,” Intl. Journal of Robotics Research , vol. 24, no. 4, 2005

work page 2005

[7] [7]

Hierarchical fingertip space: A unified framework for grasp planning and in-hand grasp adaptation,

K. Hang, M. Li, J. A. Stork, Y . Bekiroglu, F. T. Pokorny, A. Billard, and D. Kragic, “Hierarchical fingertip space: A unified framework for grasp planning and in-hand grasp adaptation,” IEEE Trans. on Robotics, vol. 32, no. 4, 2016

work page 2016

[8] [8]

Grasp stability prediction for a dexterous robotic hand combining depth vision and haptic bayesian exploration,

M. S. Siddiqui, C. Coppola, G. Solak, and L. Jamone, “Grasp stability prediction for a dexterous robotic hand combining depth vision and haptic bayesian exploration,” Frontiers in Robotics and AI , 2021

work page 2021