Dexterous Manipulation Based on Prior Dexterous Grasp Pose Knowledge

Cewu Lu; Haoshu Fang; Hengxu Yan

REVIEW 4 major objections 6 minor 44 references

Dexterous Manipulation Based on Prior Dexterous Grasp Pose Knowledge

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper argues that seeding reinforcement learning with a dexterous grasp pose on the object's functional part makes articulated-object manipulation learn up to 150 times faster and succeed 15 to 29 percentage points more often than…

desk verdict Plausible core result, but the headline speedup numbers rest on an unvalidated manual grasp mapping and a loosely reproduced baseline; still worth refereeing. read the letter →

arxiv 2412.15587 v1 pith:PPNVBHRR submitted 2024-12-20 cs.RO cs.LG

Hengxu Yan , Haoshu Fang , Cewu Lu This is my paper · ORCID

classification cs.ROcs.LG

keywords dexterousmanipulationreinforcementlearninggraspposepriorarticulatedobjectsfunctionalpartsegmentationAnygraspPPOsim-to-realtransfer

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that the slow, sample-hungry phase of learning dexterous manipulation is not the fine finger control itself but the search for where to stand and how to initially grasp the object. It proposes a two-stage pipeline: first segment the object's functional part and use a two-finger grasp generator to propose grasp poses, which are mapped to an Allegro hand configuration and filtered by collision detection to obtain an initial dexterous grasp pose; then train a PPO policy to refine that grasp. Across four articulated-object tasks (bucket, laptop, faucet, toilet) the method reports success-rate gains of 15 to 29.31 percentage points and learning-efficiency gains of 80 to 150 times over the DexArt baseline, with the gain growing as more prior knowledge (position, rotation, hand pose) is supplied. A sympathetic reader would care because it points to a cheap way to inject geometric prior knowledge into high-degree-of-freedom manipulation without demonstrations or teleoperation data.

What carries the argument

The load-bearing object is the initial dexterous grasp pose $G=[R\ t\ B]$, where $R$ and $t$ are the end-effector rotation and translation and $B\in\mathbb{R}^{16}$ are the Allegro hand joint angles. The pipeline that produces it is the mechanism: a segmentation network extracts the functional part, Anygrasp proposes two-finger grasps $\hat G=[R\ t\ w]$, and a manually constructed mapping $f$ discretizes the grasp width $w$ into $n$ levels and assigns four hand configurations per level. This mapped pose, after collision filtering, converts the RL problem from one of discovering where to grasp into one of refining a known grasp, which is what makes the learning curve steeper.

What would settle it

Replace the manually designed mapping f in Eq. (5) with random Allegro hand configurations at the same wrist pose, run the same PPO training on bucket and laptop, and compare learning curves; if random initialization matches the reported 80 to 150 times speed-up, then prior grasp knowledge is not the cause. A simpler check is to score the mapped initial grasps with a force-closure or physical grasp-success test before RL and see whether the efficiency gains track grasp quality.

Watch

Extended reading notes

Core claim

The paper's central claim is that a dexterous manipulation policy trained with reinforcement learning can be made much more sample-efficient and accurate if it is initialized at a grasp pose that targets the object's functional part, rather than starting from a fixed or task-specific posture. The authors decompose the problem into two phases: in the first phase, a partial-view point cloud is segmented to isolate the functional part; Anygrasp generates two-finger grasp proposals on that part; a hand-designed mapping converts each two-finger grasp's width into four candidate Allegro hand configurations; and collision detection plus a nearest-camera heuristic selects the initial dexterous grasp pose. In the second phase, PPO refines this pose using point-cloud features and a dense reward decomposed into interaction, completion, and restriction terms. The reported outcome is a 15 to 29.31 percentage-point improvement in final success rate and an 80 to 150 times reduction in training time to reach a given success level compared to the DexArt baseline across bucket, laptop, faucet, and toilet tasks.

Load-bearing premise

The hand-written mapping from two-finger grasp width to four Allegro hand configurations (Eq. 5) is assumed to produce feasible, functional initial grasps, and it is never checked with a grasp-quality metric or physical test; if those initial poses are poor, the claimed speed-up collapses.

Editorial extensions

If this is right

For articulated objects, an RL policy can skip most of the viewpoint and position search if it starts from a grasp pose on the functional part.
Adding prior knowledge in stages (position only, then rotation, then hand pose) gives steadily better learning efficiency, so partial priors are still useful.
The learned policy transfers from simulation to a real UR5 and Allegro setup for bucket lifting and laptop opening, with reported success rates of 79.08% and 74.31%.
The method is not tied to one arm: retraining with a UR5 reproduces the XArm6 results.
Success degrades gracefully when the test workspace is enlarged 16 times, but remains above the baseline, suggesting the prior helps generalization to extreme placements.

Reading between the lines

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

Extension not claimed by the paper: if the manual mapping $f$ were replaced by a learned regressor from two-finger grasp parameters to Allegro hand poses, the method would likely transfer to other hand morphologies without manual redesign.
Extension: applying the same functional-part grasp prior to rigid objects with handles would test whether the efficiency gain comes from the grasp prior itself or from the articulated-object dynamics.
Extension: the real-world failures the authors list suggest that force and torque saturation at the wrist, not the initial grasp pose, is the next bottleneck; adding tactile feedback is a natural test.

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

Quick take: the paper has a plausible core direction, and the ablations do show that giving the RL policy a good initial position, rotation, and hand configuration speeds learning. The headline numbers (80-150x speedup, 15-29 pp success) are probably inflated by a loose DexArt baseline, and the load-bearing manual mapping from two-finger width to Allegro hand pose (Eq. 5) is never validated. I'd send it to review, but with a request to fix the baseline and test the mapping.

What's actually new: a two-stage pipeline that segments the functional part, runs Anygrasp, maps two-finger grasps to a 16-DOF Allegro configuration, then seeds PPO with that grasp. Prior work has used grasp priors or affordances for dexterous manipulation (e.g., human grasp affordance), but this specific combination on DexArt articulated tasks, with partial point clouds, is new. The ablation logic is sound: MGP (position only), MGPR (position+rotation), and full method (plus hand B) show monotonic improvement, which is a clean demonstration that prior knowledge helps. The real-world transfer with UR5 and Allegro gives roughly 10 trials per task and reports 74-79% success, with the failure modes described honestly.

Soft spots, in order of severity. First, Eq. (5) is the heart of the 'prior dexterous grasp pose knowledge' claim, and it's a hand-authored lookup table with no grasp-quality check or physical validation. The ablations vary t and R but hold B fixed to the manual mapping, so they cannot rule out that the gains come from a task-specific hand pose that happens to work, rather than from 'prior knowledge' in any transferable sense. The stress-test note is right: replacing f with a learned or randomized mapping is the missing experiment. Second, the DexArt baseline is described as 'manipulation from a fixed position' but the original DexArt uses a specific pipeline; the paper's reproduction is underspecified. Third, no error bars despite three seeds; the aggregate numbers in Table I are point estimates. Fourth, the real-world sample is tiny, and the paper says 'approximately 10 times'—that should be exact. None of these are fatal; the central direction is credible. But the magnitude of the claimed gains is not established.

I'd send this to a serious venue: the experiments are reproducible in principle, the ablations are informative, and the question (does grasp-pose initialization pay off?) is worth an answer. The review should demand error bars, a precise baseline definition, a validation or ablation of the mapping f, and ideally code release.

Referee Report

4 major / 6 minor

Summary. The paper proposes a two-stage framework for dexterous manipulation of articulated objects. In the first stage, a segmentation network identifies the object's functional part from a partial point cloud, Anygrasp generates two-finger grasp poses on that part, and a manually designed mapping f converts each two-finger grasp width into four Allegro hand configurations; after collision filtering, the pose closest to the camera is selected as the initial dexterous grasp pose. In the second stage, PPO refines this initial grasp using point-cloud observations and a dense reward with interaction, completion, and restriction terms. Experiments on four PartNet-Mobility tasks report higher success rates and 80--150x faster learning than a DexArt baseline, with ablations (MGP, MGPR) showing that position and rotation priors each contribute. Real-world tests on bucket-lifting and laptop-opening report success rates around 74--79%. The central claim is that initializing RL with a prior dexterous grasp pose on the functional part improves both learning speed and final success.

Significance. If substantiated, the paper's direction is valuable: it combines grasp-pose priors with RL to avoid the costly early exploration of approach position and viewpoint, and it provides a concrete recipe (segmentation + Anygrasp + mapping + PPO) that is evaluated on a standard articulated-object benchmark. The paper's strengths include the two-stage decomposition, the MGP/MGPR ablations that isolate the contributions of position and rotation, the use of a pretrained point-cloud feature extractor, and the attempt to demonstrate real-world transfer. However, the central mechanism for injecting 'dexterous grasp pose knowledge' is the manually authored mapping f in Eq. (5), which is never validated or ablated; the baseline is loosely specified; and the reported quantitative gains lack variance information and precise definitions. These issues currently prevent the work from conclusively supporting the claimed 15--29.31% success improvement and 80--150x efficiency factor.

major comments (4)

[Section III.C, Eq. (5)] The mapping f from discretized two-finger widths wi to four Allegro hand configurations {Bi1, Bi2, Bi3, Bi4} is described as 'manually designed' and is never validated with a grasp-quality metric, a force-closure check, a physical feasibility test, or an ablation that varies B. Since this lookup table fully determines the initial dexterous grasp pose that seeds PPO, the paper's central claim that 'prior dexterous grasp pose knowledge' drives the efficiency gain rests on an unvalidated, hand-authored component. The existing MGP/MGPR ablations isolate position and rotation but do not vary B or f, so they cannot show that the specific hand configuration is responsible for the reported gains. Please add (i) a validation of f (e.g., simulated or physical grasp success rates on the functional part for each width bin), and (ii) an ablation that replaces the hand-authored B with a random or learned initialization while keeping t and R fixed; without such evidence, the improvement could be attributable to task-specific tuning of f rather than to generalizable grasp knowledge.
[Section IV.A.2, baseline and efficiency claim] The baseline 'Manipulation from a Fixed Position' is described only as 'equivalent to the DexArt method,' with no specification of its initial pose, observation space, reward function, or hyperparameters. The headline numbers 'enhances the success rate by 15% to 29.31% and improves learning efficiency by a factor of 80 to 150' cannot be verified without a precise baseline definition and a formal measure of learning efficiency (e.g., training timesteps to reach a given success threshold, with the threshold specified). In addition, the stated success-rate range is not consistent with Table I: for faucet the improvement is 73.33% - 60% = 13.33 percentage points, which lies outside the claimed 15--29.31% range. Please state the exact baseline configuration, define the efficiency metric, and correct or qualify the claimed range.
[Section IV.A.3, Table I and Figures 5-7] The paper states that evaluation uses 'three random seeds' and reports average success rates, but no variance, standard deviation, or per-seed curves are shown. Figure 5, 6, and 7 each display single curves for 'Ours' and the baselines. Without error bars or individual seed trajectories, it is impossible to judge whether the differences between Ours, DexArt, MGP, and MGPR are meaningful or within seed-to-seed noise. Please report standard deviations (e.g., shaded regions in the learning curves) or per-seed results for all reported numbers.
[Section IV.B, Table II] The real-world evaluation is described as 'tested approximately 10 times' per task, but the exact number of trials per object and per difficulty category is not given, and no confidence intervals are provided. With roughly 10 trials, the reported success rates (79.08% for bucket, 74.31% for laptop) have very wide confidence intervals; for example, a 79% rate with 10 trials has a 95% binomial confidence interval of roughly 49% to 95%. Please report the trial counts and confidence intervals, and avoid drawing strong quantitative conclusions from such a small sample.

minor comments (6)

[Abstract] The sentence 'Unlike previous work, they always make the robotic hand go with a fixed dexterous grasp pose' is grammatically unclear and should be rewritten (e.g., 'Unlike prior work that always uses a fixed dexterous grasp pose...').
[Section III.C, Eq. (3)] The notation R ∈ R3×1 for the rotation component is nonstandard; rotations are not generally vectors in R3. Please clarify whether R is a rotation vector (axis-angle), a quaternion, or a rotation matrix, and specify the exact dimension of G.
[Section III.D.2] The action space description says the 6-dimensional vector 'captures the angular and linear velocities' but then states that 'we compute the translation and rotation of the robotic arm's end-effector.' Please clarify how the velocities are integrated and how the resulting end-effector pose is converted to joint angles via IKFast.
[Section III.D.3, Eq. (6)-(13)] The reward balance parameters α, β, η and the regularization terms λ, γ, δ are never given numeric values. Since the reward is a central design component, please report the values used in all experiments.
[Section IV.A.1 and Fig. 4] The 'red ball' in Fig. 4 is not defined in the main text; please explain in the caption or text that it marks the target point in the functional part used by the interaction reward.
[Section IV.B] The statement 'the process is tested approximately 10 times' is vague; replace 'approximately' with the exact number of trials per condition.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the grasp-pose prior is a modular external input and the reported gains are empirical comparisons against a baseline, not quantities fitted by construction.

full rationale

The derivation chain uses Anygrasp [14] to produce two-finger grasp proposals from the segmented functional part (Eq. 2), maps widths to Allegro hand configurations through a manually designed lookup table (Eq. 5), and then trains PPO from the resulting initial pose. The central claim, that this initialization improves learning speed and success, is established by comparing against DexArt and by ablations MGP/MGPR. None of the reported success rates or efficiency factors are fitted parameters of the method; the reward weights alpha/beta/eta and the hand mapping f are design choices, not quantities solved for from the target results. The one same-group citation, Anygrasp, is a published grasp-perception system trained on external grasp data and evaluated on its own benchmark; the present paper uses it as a modular input generator rather than as evidence for the RL improvement. Eq. (5) is a hand-authored ansatz and is a legitimate correctness risk (the mapping is not validated with grasp-quality metrics), but it is not circular: the contribution of the hand configuration B is tested empirically by comparing Ours with MGPR, and the outcome could in principle have gone the other way. No equation or fitted value in the paper is, by construction, identical to a predicted quantity. Therefore no circular step is present.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the quality of the hand-designed mapping from two-finger widths to Allegro hand configurations, the transferability of Anygrasp proposals to a dexterous hand, and the definition of the functional part; none of these are externally benchmarked. The reward hyperparameters and the camera-closest pose selection are additional tunable choices.

free parameters (3)

manual mapping f (discretized widths to hand configurations)
Section III.C Eq. (5): the authors discretize the two-finger grasp width w into n widths and manually design four Allegro hand configurations {Bi1..Bi4} for each. No numerical values or validation are given; this hand-authored mapping fully determines the quality of the initial dexterous grasp pose.
reward balance coefficients alpha, beta, eta (and lambda, gamma, delta)
Section III.D.3 Eq. (6)-(10): the dense reward is a weighted sum with balance parameters alpha, beta, eta, plus regularization terms lambda, gamma and completion bonus delta. Values are not reported; the training outcome depends on them.
pose selection criterion (closest to camera)
Section III.A: among collision-free mapped grasp poses, the pose closest to the camera is selected. This heuristic is a design choice that affects the initialization and is not justified.

assumptions (4)

domain assumption Two-finger grasp poses from Anygrasp on the segmented functional part are a sufficient prior for dexterous manipulation.
The entire first stage relies on Anygrasp's proposals being task-relevant and transferable to a dexterous hand after the manual mapping.
ad hoc to paper The manual mapping f produces feasible, functional Allegro hand grasps.
Section III.C: no grasp success metric or physical validation is provided for the mapped configurations {Bi1..Bi4}.
domain assumption The functional part of an object (the red-ball target) is the correct region to grasp for every task.
Section III.B and Fig. 4: the segmentation network is trained to locate the functional part, and the grasp is generated on that part; task success depends on this definition.
domain assumption The SAPIEN simulator with a friction coefficient of 5 and given time steps is a faithful model for Sim2Real transfer.
Section IV.A.1: the simulation parameters are set but the real-world transfer shows drops and requires task-specific modifications (foam board, nail), indicating a domain gap.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Dexterous Manipulation Based on Prior Dexterous Grasp Pose Knowledge." pith.science (2026). https://pith.science/paper/PPNVBHRR

@misc{pith2026241215587,
  author       = {Pith},
  title        = {Pith review of: Dexterous Manipulation Based on Prior Dexterous Grasp Pose Knowledge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PPNVBHRR}},
  note         = {Machine review of arXiv:2412.15587}
}

read the original abstract

Dexterous manipulation has received considerable attention in recent research. Predominantly, existing studies have concentrated on reinforcement learning methods to address the substantial degrees of freedom in hand movements. Nonetheless, these methods typically suffer from low efficiency and accuracy. In this work, we introduce a novel reinforcement learning approach that leverages prior dexterous grasp pose knowledge to enhance both efficiency and accuracy. Unlike previous work, they always make the robotic hand go with a fixed dexterous grasp pose, We decouple the manipulation process into two distinct phases: initially, we generate a dexterous grasp pose targeting the functional part of the object; after that, we employ reinforcement learning to comprehensively explore the environment. Our findings suggest that the majority of learning time is expended in identifying the appropriate initial position and selecting the optimal manipulation viewpoint. Experimental results demonstrate significant improvements in learning efficiency and success rates across four distinct tasks.

Figures

Figures reproduced from arXiv: 2412.15587 by the authors.

**Figure 1.** For the tasks of lifting the bucket and opening the laptop, we set the initial dexterous grasp pose to facilitate successful task completion. Inspired by the learning process observed in infants, we propose a dexterous manipulation method grounded in prior dexterous grasp pose knowledge, as illustrated in Fig.1. Our approach is structured into two phases. First, we segment the functional part of the object, which is… view at source ↗

**Figure 2.** Illustration of our dexterous manipulation method. We employ PPO to teach the dexterous hand how to manipulate objects based on a dexterous grasp pose. (1) Starting with a partial-view point cloud captured by the initial camera, we use PointNet1 to segment the functional part of the object, which is then used to generate a set of two-finger grasp poses with Anygrasp. These poses are subsequently mapped to a dexterou… view at source ↗

**Figure 3.** Mapping of the coordinate system from two-finger grasp poses to four grasp types for the dexterous hand. {Bi1, Bi2, Bi3, Bi4} = f(wi). i ∈ {1, 2, · · · , n}. (5) [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: The center of the red ball indicates the position the finger should approach, as shown at the top of the figure. It will adjust as the functional part of the object changes. Below, the illustrations depict the segmentation results for the functional parts across four t…

**Figure 5.** Figure 5: Illustration of the success rate as a function of the training process using XArm6 on the simulation test dataset. Category Bucket Laptop Faucet Toilet Ours 94.17 90 73.33 87.88 Ours(big) 85.21 76.3 47.73 42.38 DexArt 79.17 66.67 60 58.57 DexArt(big) 54.58 34 17.56 34.…

**Figure 6.** Figure 6: Comparison of success rates between our method and various exploration policies using XArm6 on the simulation test dataset. To further evaluate our model, we expand the test space by a factor of 16 to challenge the model with extreme object placements. Objects are rand…

**Figure 7.** Figure 7: Comparison of success rates between our method and DexArt using XArm6 on the simulation test dataset. B. Real-world Experiments To validate our model in a real-world environment, we select two tasks: opening a laptop and lifting a bucket. The experimental setup is illu…

**Figure 8.** Figure 8: The red box illustrates the real-world setting, the blue box shows the initial dexterous grasp pose, and the green box shows the manipulation process over time in the real world. 0 10 20 30 40 Training timestep (X10 6 ) 0.0 0.2 0.4 0.6 0.8 1.0 Success rate Ours_small O…

**Figure 9.** Figure 9: Comparison of success rates between our method and DexArt using the UR5 on the simulation test dataset [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]

**Figure 10.** Figure 10: Test objects in the real world. On the left are the more difficult objects for the laptop and bucket tasks, while on the right are the easier objects. poses. Our method is rigorously evaluated through a series of simulations and real-world experiments across various t…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 32 canonical work pages

[1]

Learning reward functions for robotic manipulation by observing humans

Minttu Alakuijala et al. “Learning reward functions for robotic manipulation by observing humans”. In: 2023 IEEE International Conference on Robotics and Automation (ICRA) . IEEE. 2023, pp. 5006–5012

work page 2023
[2]

Transferring dexterous manip- ulation from gpu simulation to a remote real-world trifinger

Arthur Allshire et al. “Transferring dexterous manip- ulation from gpu simulation to a remote real-world trifinger”. In: 2022 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) . IEEE. 2022, pp. 11802–11809

work page 2022
[3]

Learning dex- terous in-hand manipulation

OpenAI: Marcin Andrychowicz et al. “Learning dex- terous in-hand manipulation”. In: The International Journal of Robotics Research 39.1 (2020), pp. 3–20

work page 2020
[4]

Dexterous im- itation made easy: A learning-based framework for efficient dexterous manipulation

Sridhar Pandian Arunachalam et al. “Dexterous im- itation made easy: A learning-based framework for efficient dexterous manipulation”. In: 2023 ieee inter- national conference on robotics and automation (icra). IEEE. 2023, pp. 5954–5961

work page 2023
[5]

Holo-dex: Teach- ing dexterity with immersive mixed reality

Sridhar Pandian Arunachalam et al. “Holo-dex: Teach- ing dexterity with immersive mixed reality”. In: 2023 IEEE International Conference on Robotics and Au- tomation (ICRA) . IEEE. 2023, pp. 5962–5969

work page 2023
[6]

Dexart: Benchmarking general- izable dexterous manipulation with articulated ob- jects

Chen Bao et al. “Dexart: Benchmarking general- izable dexterous manipulation with articulated ob- jects”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 2023, pp. 21190–21200

work page 2023
[7]

ContactGrasp: Functional Multi-finger Grasp Synthesis from Contact

Samarth Brahmbhatt et al. “ContactGrasp: Functional Multi-finger Grasp Synthesis from Contact”. In: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . 2019, pp. 2386–2393. DOI: 10.1109/IROS40897.2019.8967960

arXiv 2019
[8]

A survey of POMDP appli- cations

Anthony R Cassandra. “A survey of POMDP appli- cations”. In: Working notes of AAAI 1998 fall sym- posium on planning with partially observable Markov decision processes. V ol. 1724. 1998

work page 1998

Show all 44 references

[9]

Learning generalizable robotic reward functions from

Annie S Chen, Suraj Nair, and Chelsea Finn. “Learning generalizable robotic reward functions from” in-the-wild” human videos”. In: arXiv preprint arXiv:2103.16817 (2021)

2021 arXiv
[10]

Visual dexterity: In-hand reorienta- tion of novel and complex object shapes

Tao Chen et al. “Visual dexterity: In-hand reorienta- tion of novel and complex object shapes”. In: Science Robotics 8.84 (2023), eadc9244

2023
[11]

Guided deep reinforcement learning of control poli- cies for dexterous human-robot interaction

Sammy Christen, Stefan Stev ˇsi´c, and Otmar Hilliges. “Guided deep reinforcement learning of control poli- cies for dexterous human-robot interaction”. In: 2019 International Conference on Robotics and Automation (ICRA). IEEE. 2019, pp. 2161–2167

2019
[12]

Automated Construction of Robotic Manipulation Programs

Rosen Diankov. “Automated Construction of Robotic Manipulation Programs”. PhD thesis. Carnegie Mel- lon University, Robotics Institute, Aug. 2010. URL: http : / / www . programmingvision . com / rosen_diankov_thesis.pdf

2010
[13]

Bridge data: Boosting general- ization of robotic skills with cross-domain datasets

Frederik Ebert et al. “Bridge data: Boosting general- ization of robotic skills with cross-domain datasets”. In: arXiv preprint arXiv:2109.13396 (2021)

2021 arXiv
[14]

Anygrasp: Robust and efficient grasp perception in spatial and temporal domains

Hao-Shu Fang et al. “Anygrasp: Robust and efficient grasp perception in spatial and temporal domains”. In: IEEE Transactions on Robotics (2023)

2023
[15]

GraspNet-1Billion: A Large- Scale Benchmark for General Object Grasping

Hao-Shu Fang et al. “GraspNet-1Billion: A Large- Scale Benchmark for General Object Grasping”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 2020, pp. 11444–11453

2020
[16]

Physics-based dexterous manipulations with estimated hand poses and residual reinforcement learning

Guillermo Garcia-Hernando, Edward Johns, and Tae- Kyun Kim. “Physics-based dexterous manipulations with estimated hand poses and residual reinforcement learning”. In: 2020 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) . IEEE. 2020, pp. 9561–9568

2020
[17]

Reset-free reinforcement learn- ing via multi-task learning: Learning dexterous ma- nipulation behaviors without human intervention

Abhishek Gupta et al. “Reset-free reinforcement learn- ing via multi-task learning: Learning dexterous ma- nipulation behaviors without human intervention”. In: 2021 IEEE International Conference on Robotics and Automation (ICRA) . IEEE. 2021, pp. 6664–6671

2021
[18]

Teach a robot to fish: Versatile imitation from one minute of demonstrations

Siddhant Haldar et al. “Teach a robot to fish: Versatile imitation from one minute of demonstrations”. In: arXiv preprint arXiv:2303.01497 (2023)

2023 arXiv
[19]

Learning dexterous manipula- tion from suboptimal experts

Rae Jeong et al. “Learning dexterous manipula- tion from suboptimal experts”. In: arXiv preprint arXiv:2010.08587 (2020)

2020 arXiv
[20]

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

Edward Johns. “Coarse-to-fine imitation learning: Robot manipulation from a single demonstration”. In: 2021 IEEE international conference on robotics and automation (ICRA) . IEEE. 2021, pp. 4613–4619

2021
[21]

Sampling-based exploration for reinforcement learning of dexterous manipulation

Gagan Khandate et al. “Sampling-based exploration for reinforcement learning of dexterous manipulation”. In: arXiv preprint arXiv:2303.03486 (2023)

2023 arXiv
[22]

Learning task-oriented grasping from human activity datasets

Mia Kokic, Danica Kragic, and Jeannette Bohg. “Learning task-oriented grasping from human activity datasets”. In: IEEE Robotics and Automation Letters 5.2 (2020), pp. 3352–3359

2020
[23]

Learning ambidextrous robot grasping policies

Jeffrey Mahler et al. “Learning ambidextrous robot grasping policies”. In: Science Robotics 4.26 (2019), eaau4984

2019
[24]

Dexvip: Learning dexterous grasping with human hand pose priors from video

Priyanka Mandikal and Kristen Grauman. “Dexvip: Learning dexterous grasping with human hand pose priors from video”. In: Conference on Robot Learning. PMLR. 2022, pp. 651–661

2022
[25]

Learn- ing dexterous grasping with object-centric visual af- fordances

Priyanka Mandikal and Kristen Grauman. “Learn- ing dexterous grasping with object-centric visual af- fordances”. In: 2021 IEEE international conference on robotics and automation (ICRA) . IEEE. 2021, pp. 6169–6176

2021
[26]

Deep dynamics models for learning dexterous manipulation

Anusha Nagabandi et al. “Deep dynamics models for learning dexterous manipulation”. In: Conference on Robot Learning . PMLR. 2020, pp. 1101–1112

2020
[27]

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

Charles R Qi et al. “Pointnet: Deep learning on point sets for 3d classification and segmentation”. In: Pro- ceedings of the IEEE conference on computer vision and pattern recognition . 2017, pp. 652–660

2017
[28]

From one hand to multiple hands: Imitation learning for dexter- ous manipulation from single-camera teleoperation

Yuzhe Qin, Hao Su, and Xiaolong Wang. “From one hand to multiple hands: Imitation learning for dexter- ous manipulation from single-camera teleoperation”. In: IEEE Robotics and Automation Letters 7.4 (2022), pp. 10873–10881

2022
[29]

Dexpoint: Generalizable point cloud reinforcement learning for sim-to-real dexterous manipulation

Yuzhe Qin et al. “Dexpoint: Generalizable point cloud reinforcement learning for sim-to-real dexterous manipulation”. In: Conference on Robot Learning . PMLR. 2023, pp. 594–605

2023
[30]

Learning complex dexterous manipulation with deep reinforcement learning and demonstrations

Aravind Rajeswaran et al. “Learning complex dexterous manipulation with deep reinforcement learning and demonstrations”. In: arXiv preprint arXiv:1709.10087 (2017)

2017 arXiv
[31]

Reinforcement learning with videos: Combining offline observations with interaction

Karl Schmeckpeper et al. “Reinforcement learning with videos: Combining offline observations with interaction”. In: arXiv preprint arXiv:2011.06507 (2020)

2020 arXiv
[32]

Proximal policy optimiza- tion algorithms

John Schulman et al. “Proximal policy optimiza- tion algorithms”. In: arXiv preprint arXiv:1707.06347 (2017)

2017 arXiv
[33]

Concept2robot: Learning manipula- tion concepts from instructions and human demon- strations

Lin Shao et al. “Concept2robot: Learning manipula- tion concepts from instructions and human demon- strations”. In: The International Journal of Robotics Research 40.12-14 (2021), pp. 1419–1434

2021
[34]

Videodex: Learning dexterity from internet videos

Kenneth Shaw, Shikhar Bahl, and Deepak Pathak. “Videodex: Learning dexterity from internet videos”. In: Conference on Robot Learning . PMLR. 2023, pp. 654–665

2023
[35]

Avid: Learning multi-stage tasks via pixel-level translation of human videos

Laura Smith et al. “Avid: Learning multi-stage tasks via pixel-level translation of human videos”. In: arXiv preprint arXiv:1912.04443 (2019)

2019 arXiv
[36]

Grasp pose detection in point clouds

Andreas Ten Pas et al. “Grasp pose detection in point clouds”. In: The International Journal of Robotics Research 36.13-14 (2017), pp. 1455–1473

2017
[37]

Unidexgrasp++: Improving dex- terous grasping policy learning via geometry-aware curriculum and iterative generalist-specialist learning

Weikang Wan et al. “Unidexgrasp++: Improving dex- terous grasping policy learning via geometry-aware curriculum and iterative generalist-specialist learning”. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision . 2023, pp. 3891–3902

2023
[38]

Dexcap: Scalable and portable mocap data collection system for dexterous manipu- lation

Chen Wang et al. “Dexcap: Scalable and portable mocap data collection system for dexterous manipu- lation”. In: arXiv preprint arXiv:2403.07788 (2024)

2024 arXiv
[39]

Graspness Discovery in Clutters for Fast and Accurate Grasp Detection

Chenxi Wang et al. “Graspness Discovery in Clutters for Fast and Accurate Grasp Detection”. In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) . Oct. 2021, pp. 15964– 15973

2021
[40]

Learning generalizable dexterous manipulation from human grasp affordance

Yueh-Hua Wu, Jiashun Wang, and Xiaolong Wang. “Learning generalizable dexterous manipulation from human grasp affordance”. In: Conference on Robot Learning. PMLR. 2023, pp. 618–629

2023
[41]

Sapien: A simulated part- based interactive environment

Fanbo Xiang et al. “Sapien: A simulated part- based interactive environment”. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020, pp. 11097–11107

2020
[42]

Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal gen- eration and goal-conditioned policy

Yinzhen Xu et al. “Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal gen- eration and goal-conditioned policy”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023, pp. 4737–4746

2023
[43]

Rotating without seeing: To- wards in-hand dexterity through touch

Zhao-Heng Yin et al. “Rotating without seeing: To- wards in-hand dexterity through touch”. In: arXiv preprint arXiv:2303.10880 (2023)

2023 arXiv
[44]

Deep imitation learning for complex manipulation tasks from virtual reality tele- operation

Tianhao Zhang et al. “Deep imitation learning for complex manipulation tasks from virtual reality tele- operation”. In: 2018 IEEE international conference on robotics and automation (ICRA) . IEEE. 2018, pp. 5628–5635

2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Learning reward functions for robotic manipulation by observing humans

Minttu Alakuijala et al. “Learning reward functions for robotic manipulation by observing humans”. In: 2023 IEEE International Conference on Robotics and Automation (ICRA) . IEEE. 2023, pp. 5006–5012

work page 2023

[2] [2]

Transferring dexterous manip- ulation from gpu simulation to a remote real-world trifinger

Arthur Allshire et al. “Transferring dexterous manip- ulation from gpu simulation to a remote real-world trifinger”. In: 2022 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) . IEEE. 2022, pp. 11802–11809

work page 2022

[3] [3]

Learning dex- terous in-hand manipulation

OpenAI: Marcin Andrychowicz et al. “Learning dex- terous in-hand manipulation”. In: The International Journal of Robotics Research 39.1 (2020), pp. 3–20

work page 2020

[4] [4]

Dexterous im- itation made easy: A learning-based framework for efficient dexterous manipulation

Sridhar Pandian Arunachalam et al. “Dexterous im- itation made easy: A learning-based framework for efficient dexterous manipulation”. In: 2023 ieee inter- national conference on robotics and automation (icra). IEEE. 2023, pp. 5954–5961

work page 2023

[5] [5]

Holo-dex: Teach- ing dexterity with immersive mixed reality

Sridhar Pandian Arunachalam et al. “Holo-dex: Teach- ing dexterity with immersive mixed reality”. In: 2023 IEEE International Conference on Robotics and Au- tomation (ICRA) . IEEE. 2023, pp. 5962–5969

work page 2023

[6] [6]

Dexart: Benchmarking general- izable dexterous manipulation with articulated ob- jects

Chen Bao et al. “Dexart: Benchmarking general- izable dexterous manipulation with articulated ob- jects”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 2023, pp. 21190–21200

work page 2023

[7] [7]

ContactGrasp: Functional Multi-finger Grasp Synthesis from Contact

Samarth Brahmbhatt et al. “ContactGrasp: Functional Multi-finger Grasp Synthesis from Contact”. In: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . 2019, pp. 2386–2393. DOI: 10.1109/IROS40897.2019.8967960

arXiv 2019

[8] [8]

A survey of POMDP appli- cations

Anthony R Cassandra. “A survey of POMDP appli- cations”. In: Working notes of AAAI 1998 fall sym- posium on planning with partially observable Markov decision processes. V ol. 1724. 1998

work page 1998