Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Dexterous Cable Manipulation: Taxonomy, Multi-Fingered Hand Design, and Long-Horizon Manipulation

T0 review · 4 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A taxonomy-driven hand with two symmetric thumbs and rotatable fingertips can replay one cable demonstration per primitive with 88% success on same-material cables and over 75% on very different cables.

desk verdict A worthwhile taxonomy and hand-design paper whose headline success rates don't survive contact with its own Table II. read the letter →

arxiv 2502.00396 v2 pith:WTONHTWG submitted 2025-02-01 cs.RO

Sun Zhaole , Xiao Gao , Xiaofeng Mao , Jihong Zhu , Aude Billard , Robert B. Fisher This is my paper · ORCID
classification cs.RO
keywords dexterouscablemanipulationtaxonomymulti-fingeredhanddesigndual-thumbrotatablefingertipsdemonstrationreplayfinitestatemachinesdeformablelinearobjects
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 dexterous cable manipulation—pulling, bending, hooking, rotating, and inserting cables in one hand—has been under-served by two-fingered grippers and by anthropomorphic hands built with a single thumb. The authors organize one-handed cable manipulation into a taxonomy, Cable Dexonomy, whose key observation is that the thumb–index combination is the workhorse of almost every primitive. That observation drives a hardware proposal: a five-fingered, 25-degree-of-freedom hand with two symmetric thumbs and a rotatable joint on every fingertip. With only one human-dragged demonstration per primitive, replayed open-loop, the hand succeeds on 88% of attempts on same-material cables and over 75% on substantially different cables across eight primitives, and composes those primitives with finite state machines to reach 64% on four long-horizon tasks. The paper also offers a human baseline using a gloved, non-dominant hand with eyes closed, against which the robot performs comparably on primitives.

What carries the argument

Three mechanisms carry the argument. Cable Dexonomy is the first: a taxonomy for one-handed dexterous cable manipulation, adapted from a hand-centric classification scheme, whose six criteria sort primitives by prehensility, motion, in-hand versus out-of-hand, external support, which fingers are used, and the cable's goal configuration. Its central units are the thumb–index combination (TIC), the workhorse of nearly all primitives, and the virtual middle finger (VMF), a functional group of the middle, ring, and little fingers. The second mechanism is the hand itself: a 25-DoF five-fingered hand built from existing anthropomorphic finger designs, with two symmetric thumb–index pairs and an extra rotatable joint on every fingertip, which lets the fingers perform pincer grasps and Z-axis cable rotation without sliding. The third mechanism is the data pipeline: humans drag the fingers to record joint-angle trajectories under low joint stiffness, then the hand replays one successful trajectory per primitive under PID position control, and finite state machines sequence primitives into long-horizon tasks.

What would settle it

Run the same demonstration collection and PID replay pipeline on a one-thumb anthropomorphic hand with the same fingers, sponge, and cable initialization; if it matches the 88% and 75% primitive success rates, the causal claim for the second thumb fails. Separately, lock the rotatable fingertip joints and repeat Z-axis orientation control in the air on Cable A; the design claim predicts a sharp drop from roughly 0.6–0.8 success.

Watch

Extended reading notes

Core claim

The paper's central claim is that cable manipulation dexterity is not primarily a control problem: it is a design problem, and the right design is discoverable from a task taxonomy. Analyzing one-handed cable manipulation with six criteria—prehensile versus non-prehensile, motion or none, in-hand or out-of-hand, with or without external support, fingers used, and goal configuration—shows that the thumb–index combination dominates almost every primitive and that the tasks are largely symmetric about the hand's Y-axis. The paper then claims that a hand with two symmetric thumb–index combinations and an extra rotation at each fingertip can exploit this structure: given one dragged demonstration per primitive, replayed without modification, the hand achieves 88% success on same-material cables, over 75% on very different cables, and, by composing primitives with finite state machines, 64% success on long-horizon tasks for same-material cables. The robot's primitive-level success is comparable to an intentionally degraded human baseline (gloved, non-dominant hand, eyes closed). The paper explicitly does not claim an autonomous system: long-horizon execution relies on human-guided primitive switching.

Load-bearing premise

The load-bearing premise is that the two hardware changes—the second thumb and the rotatable fingertip joints—are what produce the reported dexterity, rather than the overall five-fingered configuration, the sponge padding, or the specific demonstration and replay protocol.

Editorial extensions

If this is right

  • One human-dragged demonstration per primitive, replayed open-loop under PID position control, transfers to same-material cables of other diameters with 88% success and to cables of very different material, stiffness, and diameter with over 75% success across eight primitives.
  • A long-horizon task can be decomposed by the taxonomy into primitives and executed as a finite state machine without ever collecting a demonstration of the complete trajectory; four such tasks reached 64% success on same-material cables.
  • The symmetric dual-thumb layout means a single demonstration covers both left-to-right and right-to-left pulling, because the two thumb–index combinations are mirror images.
  • A hand with no tactile sensors can pull, hook, bend, and reorient cables in-hand using only finger motion, something prior tactile or gripper-based sliding methods could not do from a fixed hand base.
  • The recorded joint-angle trajectories are reusable training data for imitation learning methods such as ACT and Diffusion Policy, which the paper identifies as natural next steps.

Reading between the lines

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

  • If the causal role of the second thumb is confirmed by ablation, the paper's design principle—mirror the dominant finger pair for task families with symmetry—could transfer to other deformable-object manipulation and to non-cable in-hand tasks.
  • The sharp drop in long-horizon success from 64% on same-material cables to 10% on very different cables suggests that the bottleneck is not the primitives but the absence of feedback during FSM transitions; closed-loop state estimation is the obvious extension.
  • Because the reported rates aggregate five trials per primitive per cable, they do not distinguish grasp-slippage failures from orientation-control failures; a fine-grained error taxonomy would make the comparison to the human baseline more informative.
  • Reading the human baseline correctly matters: it is an intentionally degraded lower bound (gloved, non-dominant hand, eyes closed), so the robot's comparable primitive scores imply rough parity with a restricted human, not with ordinary unaided dexterity.
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

4 major / 4 minor

Summary. This paper addresses dexterous manipulation of cables with a multi-fingered hand. It makes three contributions: (1) a taxonomy of one-handed dexterous cable manipulation primitives called Cable Dexonomy, (2) a five-fingered non-anthropomorphic hand with two symmetric thumbs and rotatable fingertips, and (3) a demonstration-collection pipeline in which humans physically drag the robot's fingers, followed by open-loop replay and finite-state-machine composition into long-horizon tasks. Experiments on eight short-term primitives and four long-horizon tasks across six cables report an 88% replay success rate on cables of the same material and over 75% on cables of very different materials, with a human lower-bound baseline wearing a ski glove. The central claim is that one demonstration per primitive, replayed open-loop, transfers across cables and enables long-horizon manipulation.

Significance. If the reported results hold, the taxonomy and hand design would be a useful step beyond anthropomorphic grippers for deformable-object manipulation, and the demonstration-collection pipeline is a practical way to obtain training data for a hand with non-human kinematics. The candid limitations section, the explicit decomposition of long-horizon tasks, and the real-world evaluation with a human lower-bound baseline are strengths. However, the quantitative support is currently weakened by an internal arithmetic inconsistency in Table II and by the absence of ablations that would isolate the contributions of the two advertised hardware innovations. The paper does not provide machine-checked proofs or code; its main artifacts are the physical hand, the taxonomy, and the experimental video.

major comments (4)
  1. [Section VI-B, Table II] The advertised success rates do not match the per-cell entries in Table II. For the eight short-term primitives on Cables D, E, and F, summing the robot rates (five trials per cell) gives 5.4 + 4.4 + 5.6 = 15.4 successes over 24 cells, i.e., 64.2%, not the printed 75%. Separately, for the four long-horizon tasks on Cables A, B, and C, the per-cell rates sum to 45 successes out of 60 trials (75%), not the printed 64%. These inconsistencies affect the paper's headline generalization claims and the comparison with the human lower bound (71% on hard cables). The authors should correct either the aggregates or the per-cell entries and provide the raw per-trial logs so the numbers can be verified.
  2. [Section IV-A/B and Section VI-B] The paper attributes the reported dexterity to the two symmetric thumbs and the rotatable fingertips, but the experiments evaluate only the full system (Table II). There is no ablation or comparison with a one-thumb Leap hand, with the rotatable fingertip joint locked, or with an otherwise identical hand lacking these features. Since the hand design is one of the three stated contributions, the absence of any control condition leaves the causal role of the two features unestablished; the qualitative arguments in Section IV do not substitute for a quantitative comparison.
  3. [Section VI-B and Abstract] The claim that the robot achieved performance comparable to human baseline dexterity in both primitive actions and long-horizon tasks is not supported for the hard cables: Table II shows a long-horizon robot success rate of 10% versus 45% for the human lower bound. The paper acknowledges this drop later in the section, but the abstract and the use of 'comparable' overstate the result. The comparability claim should be restricted to cables of the same material as the demonstration cable, or the claim should be removed.
  4. [Section VI-B] The paper omits a taxonomy primitive, Y-axis orientation control, because no successful demonstration could be generated: the text states 'We did not add Y-axis orientation control because we could not generate a successful demonstration.' Since the taxonomy is presented as a comprehensive organization of dexterous cable manipulation and the hand is claimed to enable it, this omitted primitive weakens the scope of the contribution. The abstract and conclusion should explicitly state the demonstrated subset of the taxonomy rather than implying full coverage.
minor comments (4)
  1. [Throughout] There are multiple typos, e.g., 'multi-fingerd' in the Table I caption, 'cable waiving' for 'waving', 'showned' in Section VI-B, and 'Zhaole et. al' in Section IV-A. The paper should be proofread.
  2. [Section VI-A] Each cell in Table II is based on only five trials, and no confidence intervals are reported; because per-cell outcomes are binomially distributed as k/5, many apparent differences in the table are within sampling error. Reporting raw counts and exact binomial confidence intervals would substantially improve the quantitative claims.
  3. [Section V-A] The demonstration collection procedure states that 'the successful demonstrations are used for the later replay' and that only the first successful demonstration was recorded. The number of failed human attempts before a successful demonstration was obtained is not reported, which is relevant to assessing the one-shot nature of the pipeline.
  4. [Figure 6] The caption says 'The red dot indicates the joint positions on each identical finger. The last blue link is the rotatable fingertip.' Since the rotatable fingertip is a central design contribution, the figure should label the rotatable fingertip joint itself rather than only the link, to make the mechanism clear.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's quantitative claims are empirical replay results, not outputs of the taxonomy or of a fitted model.

full rationale

The paper's central claims are empirical hardware and pipeline results. The Cable Dexonomy (Section III) is a classification scheme; it motivates the two-thumb and rotatable-fingertip design in Section IV qualitatively, but it does not generate the success rates in Table II. Those rates come from physical trials (5 per cell) of demonstration replay, and the replayed trajectories are recorded human joint-angle traces, not parameters fitted to the test cables. The only self-citation, reference [66] in Section IV-A, reports a prior simulation observation about direction-dependent control policies and is not invoked as a forced uniqueness theorem; the dual-thumb design is justified by taxonomy evidence and by the symmetric-manipulation example, and its causal role would need an ablation, which is a correctness or completeness concern rather than a circularity. The reviewer-noted inconsistency between the claimed 'over 75%' hard-cable success rate and the 64.2% sum of the printed per-cell values in Table II is a real internal-consistency problem, but it is a data-accounting issue, not a reduction of a predicted quantity to its own input. No fitted parameter is renamed as a prediction, and no known result is merely relabeled. The derivation chain is therefore self-contained with respect to circularity.

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

The central claims rest on domain assumptions about task scope and the validity of the demonstration protocol, not on fitted numerical parameters. No physical entities are postulated; the dual-thumb and rotatable fingertip structure is a hardware construction, and the taxonomy terms (TIC, VMF) are borrowed or adapted from existing virtual-finger concepts.

assumptions (4)
  • domain assumption Cable manipulation relevant to the taxonomy is quasi-static; dynamic effects such as waving are excluded.
    Appendix A states that dynamic cable manipulation cannot be easily achieved with pure finger motions and is therefore not considered in Cable Dexonomy.
  • domain assumption A single successful demonstration per primitive, recorded on one cable, can support replay on unseen cables.
    Section V-B records only the first successful demonstration and Section VI-B evaluates replay on five other cables.
  • domain assumption The fixed hand base, tilted palm-down workspace, and cable initialization bounding box define a valid scope for dexterous cable manipulation.
    Section VI-A restricts the workspace and initialization; tasks requiring arm motion, such as overhand knotting, are deferred to future work in Section VIII.
  • domain assumption The skiing-glove, non-dominant hand, eyes-closed protocol provides a meaningful lower bound on human dexterity.
    Section VI-A introduces this baseline and Section VI-B uses it for comparison.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dexterous Cable Manipulation: Taxonomy, Multi-Fingered Hand Design, and Long-Horizon Manipulation." pith.science (2026). https://pith.science/paper/WTONHTWG

@misc{pith2026250200396,
  author       = {Pith},
  title        = {Pith review of: Dexterous Cable Manipulation: Taxonomy, Multi-Fingered Hand Design, and Long-Horizon Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WTONHTWG}},
  note         = {Machine review of arXiv:2502.00396}
}
read the original abstract

Existing research that addressed cable manipulation relied on two-fingered grippers, which make it difficult to perform similar cable manipulation tasks that humans perform. However, unlike dexterous manipulation of rigid objects, the development of dexterous cable manipulation skills in robotics remains underexplored due to the unique challenges posed by a cable's deformability and inherent uncertainty. In addition, using a dexterous hand introduces specific difficulties in tasks, such as cable grasping, pulling, and in-hand bending, for which no dedicated task definitions, benchmarks, or evaluation metrics exist. Furthermore, we observed that most existing dexterous hands are designed with structures identical to humans', typically featuring only one thumb, which often limits their effectiveness during dexterous cable manipulation. Lastly, existing non-task-specific methods did not have enough generalization ability to solve these cable manipulation tasks or are unsuitable due to the designed hardware. We have three contributions in real-world dexterous cable manipulation in the following steps: (1) We first defined and organized a set of dexterous cable manipulation tasks into a comprehensive taxonomy, covering most short-horizon action primitives and long-horizon tasks for one-handed cable manipulation. This taxonomy revealed that coordination between the thumb and the index finger is critical for cable manipulation, which decomposes long-horizon tasks into simpler primitives. (2) We designed a novel five-fingered hand with 25 degrees of freedom (DoF), featuring two symmetric thumb-index configurations and a rotatable joint on each fingertip, which enables dexterous cable manipulation. (3) We developed a demonstration collection pipeline for this non-anthropomorphic hand, which is difficult to operate by previous motion capture methods.

Figures

Figures reproduced from arXiv: 2502.00396 by the authors.

Figure 1
Figure 1. Our designed hand pulling the cable from right to left. (a)-(c): The hand performs a pre-grasp motion to position the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Human DCM with Prehensile (or not), Motion between fingers and the cable (or not), In-hand or out-of-hand, and Support from the external contact (or not). external forces like gravity or ground support [8]. 2. Motion indicates active movement exists between the hand and the cable. 3. In-hand means the manipulation mostly happens inside the hand space between fingers and the cable, and its opposite, out-of-hand, refe… view at source ↗
Figure 4
Figure 4. The thumb-index combination (TIC) refers to the [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (12 more)
Figure 3
Figure 3. Figure 3: Four types of goal configurations. The coordinate system shown at the lower right is the same for the human hand, [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: Three types of combinations of fingers. (a) TIC: the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Demonstrations of cable pulling. See the appendix for [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: A top view from the hand palm side and a side view [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Two manipulations to display advantages of our de [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Cable Dexonomy Finite State Machine for Cable Pulling. two index fingers and the second and the third joints of two thumbs. We only recorded the first successful demonstration without evaluating its robustness in other scenarios or on other cables, and only the actual …
Figure 10
Figure 10. Figure 10: Experiment workspace setup. (a) The hand is mounted [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: (a)-(b) Human non-dominant hand wearing a skiing glove as our baseline. (c)-(f) Short-term primitives of different [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: Human-guided primitive switching during one long-horizon manipulation according to the finite state machine with external intervention. (a) The grasped Cable A is dropped by a human and then the robot hand regrasps it. (b) The Cable A is replaced with Cable F, and aft…
Figure 15
Figure 15. Figure 15: The top row shows some minor design details of the [PITH_FULL_IMAGE:figures/full_fig_p015_15.png]
Figure 13
Figure 13. Figure 13: Demonstrations of three long-horizon manipulations. We ignore the pre-grasp which is always the first primitive to [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 14
Figure 14. Figure 14: Cable Dexonomy Finite State Machines of three long-horizon manipulations [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. The MOTIF Hand: A Robotic Hand for Multimodal Observations with Thermal, Inertial, and Force Sensors

    cs.RO 2025-06 conditional novelty 5.0 of 10

    The MOTIF hand adds thermal, inertial, and force sensing to a LEAP hand and demonstrates temperature-aware grasping and mass discrimination from fingertip flicks.

Reference graph

Works this paper leans on

71 extracted references · 53 canonical work pages · cited by 1 Pith paper

  1. [1]

    Robel: Robotics benchmarks for learning with low- cost robots

    Michael Ahn, Henry Zhu, Kristian Hartikainen, Hugo Ponte, Abhishek Gupta, Sergey Levine, and Vikash Ku- mar. Robel: Robotics benchmarks for learning with low- cost robots. In Conference on robot learning , pages 1300–1313. PMLR, 2020

    work page 2020

  2. [2]

    Solving rubik’s cube with a robot hand

    Ilge Akkaya, Marcin Andrychowicz, Maciek Chociej, Mateusz Litwin, Bob McGrew, Arthur Petron, Alex Paino, Matthias Plappert, Glenn Powell, Raphael Ribas, et al. Solving rubik’s cube with a robot hand. arXiv preprint arXiv:1910.07113, 2019

    arXiv 1910

  3. [3]

    Learning dexterous in-hand manipula- tion

    OpenAI: Marcin Andrychowicz, Bowen Baker, Maciek Chociej, Rafal Jozefowicz, Bob McGrew, Jakub Pa- chocki, Arthur Petron, Matthias Plappert, Glenn Powell, Alex Ray, et al. Learning dexterous in-hand manipula- tion. The International Journal of Robotics Research , 39 (1):3–20, 2020

    2020

  4. [4]

    Understanding human manipulation with the environment: A novel taxonomy for video labelling

    Visar Arapi, Cosimo Della Santina, Giuseppe Averta, An- tonio Bicchi, and Matteo Bianchi. Understanding human manipulation with the environment: A novel taxonomy for video labelling. IEEE Robotics and Automation Letters, 6(4):6537–6544, 2021

    work page 2021

  5. [5]

    Dexterous imitation made easy: A learning-based framework for efficient dexterous manip- ulation

    Sridhar Pandian Arunachalam, Sneha Silwal, Ben Evans, and Lerrel Pinto. Dexterous imitation made easy: A learning-based framework for efficient dexterous manip- ulation. In 2023 ieee international conference on robotics and automation (icra) , pages 5954–5961. IEEE, 2023

    work page 2023

  6. [6]

    Hands for dexterous manipulation and robust grasping: A difficult road toward simplicity

    Antonio Bicchi. Hands for dexterous manipulation and robust grasping: A difficult road toward simplicity. IEEE Transactions on robotics and automation , 16(6):652– 662, 2000

    work page 2000

  7. [7]

    Classifying human manipulation behavior

    Ian M Bullock and Aaron M Dollar. Classifying human manipulation behavior. In 2011 IEEE international conference on rehabilitation robotics , pages 1–6. IEEE, 2011

    work page 2011

  8. [8]

    A hand-centric classification of human and robot dexterous manipulation

    Ian M Bullock, Raymond R Ma, and Aaron M Dollar. A hand-centric classification of human and robot dexterous manipulation. IEEE transactions on Haptics , 6(2):129– 144, 2012

    work page 2012

Show all 71 references
  1. [9]

    A system for general in-hand object re-orientation

    Tao Chen, Jie Xu, and Pulkit Agrawal. A system for general in-hand object re-orientation. In Conference on Robot Learning , pages 297–307. PMLR, 2022

    2022

  2. [10]

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

    Tao Chen, Megha Tippur, Siyang Wu, Vikash Kumar, Ed- ward Adelson, and Pulkit Agrawal. Visual dexterity: In- hand reorientation of novel and complex object shapes. Science Robotics , 8(84):eadc9244, 2023

    2023

  3. [11]

    Iterative residual policy: for goal-conditioned dynamic manipulation of deformable objects

    Cheng Chi, Benjamin Burchfiel, Eric Cousineau, Siyuan Feng, and Shuran Song. Iterative residual policy: for goal-conditioned dynamic manipulation of deformable objects. arXiv preprint arXiv:2203.00663 , 2022

    2022 arXiv

  4. [12]

    Diffusion policy: Visuomotor policy learning via action diffusion

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research, page 02783649241273668, 2023

    2023

  5. [13]

    Evaluating initial usability of a hand augmenta- tion device across a large and diverse sample

    Dani Clode, Lucy Dowdall, Edmund da Silva, Klara Sel´en, Dorothy Cowie, Giulia Dominijanni, and Tamar R Makin. Evaluating initial usability of a hand augmenta- tion device across a large and diverse sample. Science Robotics, 9(90):eadk5183, 2024

    2024

  6. [14]

    On grasp choice, grasp models, and the design of hands for manufacturing tasks

    Mark R Cutkosky et al. On grasp choice, grasp models, and the design of hands for manufacturing tasks. IEEE Transactions on robotics and automation , 5(3):269–279, 1989

    1989

  7. [15]

    Extrinsic dexterity: In-hand manip- ulation with external forces

    Nikhil Chavan Dafle, Alberto Rodriguez, Robert Paolini, Bowei Tang, Siddhartha S Srinivasa, Michael Erdmann, Matthew T Mason, Ivan Lundberg, Harald Staab, and Thomas Fuhlbrigge. Extrinsic dexterity: In-hand manip- ulation with external forces. In 2014 IEEE International Confer...

    2014

  8. [16]

    A tale of two explanations: Enhancing human trust by explaining robot behavior

    Mark Edmonds, Feng Gao, Hangxin Liu, Xu Xie, Siyuan Qi, Brandon Rothrock, Yixin Zhu, Ying Nian Wu, Hongjing Lu, and Song-Chun Zhu. A tale of two explanations: Enhancing human trust by explaining robot behavior. Science Robotics , 4(37):eaay4663, 2019

    2019

  9. [17]

    The grasp taxonomy of human grasp types

    Thomas Feix, Javier Romero, Heinz-Bodo Schmied- mayer, Aaron M Dollar, and Danica Kragic. The grasp taxonomy of human grasp types. IEEE Transactions on human-machine systems , 46(1):66–77, 2015

    2015

  10. [18]

    Enhancing dexterity in confined spaces: Real- time motion planning for multifingered in-hand manipu- lation

    Xiao Gao, Kunpeng Yao, Farshad Khadivar, and Aude Billard. Enhancing dexterity in confined spaces: Real- time motion planning for multifingered in-hand manipu- lation. IEEE Robotics & Automation Magazine , 2024

    2024

  11. [19]

    Integrated task and motion plan- ning

    Caelan Reed Garrett, Rohan Chitnis, Rachel Holladay, Beomjoon Kim, Tom Silver, Leslie Pack Kaelbling, and Tom´as Lozano-P ´erez. Integrated task and motion plan- ning. Annual review of control, robotics, and autonomous systems, 4(1):265–293, 2021

    2021

  12. [20]

    Untangling dense knots by learning task-relevant keypoints

    Jennifer Grannen, Priya Sundaresan, Brijen Thanan- jeyan, Jeffrey Ichnowski, Ashwin Balakrishna, Minho Hwang, Vainavi Viswanath, Michael Laskey, Joseph E Gonzalez, and Ken Goldberg. Untangling dense knots by learning task-relevant keypoints. arXiv preprint arXiv:2011.04999, 2020

    2011 arXiv

  13. [21]

    See to touch: Learning tactile dexterity through visual incentives

    Irmak Guzey, Yinlong Dai, Ben Evans, Soumith Chintala, and Lerrel Pinto. See to touch: Learning tactile dexterity through visual incentives. In 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages 13825–13832. IEEE, 2024

    2024

  14. [22]

    Dextreme: Transfer of agile in-hand manipulation from simulation to reality

    Ankur Handa, Arthur Allshire, Viktor Makoviychuk, Aleksei Petrenko, Ritvik Singh, Jingzhou Liu, Denys Makoviichuk, Karl Van Wyk, Alexander Zhurkevich, Balakumar Sundaralingam, et al. Dextreme: Transfer of agile in-hand manipulation from simulation to reality. arXiv preprint ar...

    arXiv

  15. [23]

    Tactile dexterity: Manipulation prim- itives with tactile feedback

    Francois R Hogan, Jose Ballester, Siyuan Dong, and Alberto Rodriguez. Tactile dexterity: Manipulation prim- itives with tactile feedback. In 2020 IEEE international conference on robotics and automation (ICRA) , pages 8863–8869. IEEE, 2020

    2020

  16. [24]

    Dynamic handover: Throw and catch with bi- manual hands

    Binghao Huang, Yuanpei Chen, Tianyu Wang, Yuzhe Qin, Yaodong Yang, Nikolay Atanasov, and Xiaolong Wang. Dynamic handover: Throw and catch with bi- manual hands. arXiv preprint arXiv:2309.05655 , 2023

    2023 arXiv

  17. [25]

    Robotic cable routing with spatial representation

    Shiyu Jin, Wenzhao Lian, Changhao Wang, Masayoshi Tomizuka, and Stefan Schaal. Robotic cable routing with spatial representation. IEEE Robotics and Automation Letters, 7(2):5687–5694, 2022

    2022

  18. [26]

    Online active and dynamic object shape explo- ration with a multi-fingered robotic hand

    Farshad Khadivar, Kunpeng Yao, Xiao Gao, and Aude Billard. Online active and dynamic object shape explo- ration with a multi-fingered robotic hand. Robotics and Autonomous Systems , 166:104461, 2023

    2023

  19. [27]

    Catch- ing objects in flight

    Seungsu Kim, Ashwini Shukla, and Aude Billard. Catch- ing objects in flight. IEEE Transactions on Robotics , 30 (5):1049–1065, 2014

    2014

  20. [28]

    A bimanual manipu- lation taxonomy

    Franziska Krebs and Tamim Asfour. A bimanual manipu- lation taxonomy. IEEE Robotics and Automation Letters , 7(4):11031–11038, 2022

    2022

  21. [29]

    In-air knotting of rope by a dual-arm multi-finger robot

    Shunsuke Kudoh, Tomoyuki Gomi, Ryota Katano, Tetsuo Tomizawa, and Takashi Suehiro. In-air knotting of rope by a dual-arm multi-finger robot. In 2015 IEEE/RSJ In- ternational Conference on Intelligent Robots and Systems (IROS), pages 6202–6207. IEEE, 2015

    2015

  22. [30]

    Robotic manipulation of deformable rope-like objects using differentiable compliant position- based dynamics

    Fei Liu, Entong Su, Jingpei Lu, Mingen Li, and Michael C Yip. Robotic manipulation of deformable rope-like objects using differentiable compliant position- based dynamics. IEEE Robotics and Automation Letters , 8(7):3964–3971, 2023

    2023

  23. [31]

    Multi-stage cable routing through hierarchical imitation learning

    Jianlan Luo, Charles Xu, Xinyang Geng, Gilbert Feng, Kuan Fang, Liam Tan, Stefan Schaal, and Sergey Levine. Multi-stage cable routing through hierarchical imitation learning. IEEE Transactions on Robotics , 2024

    2024

  24. [32]

    Dynamic modeling and control of deformable linear objects for single-arm and dual-arm robot manipulations

    Naijing Lv, Jianhua Liu, and Yunyi Jia. Dynamic modeling and control of deformable linear objects for single-arm and dual-arm robot manipulations. IEEE Transactions on Robotics , 38(4):2341–2353, 2022

    2022

  25. [33]

    Action plan- ning for packing long linear elastic objects into compact boxes with bimanual robotic manipulation

    Wanyu Ma, Bin Zhang, Lijun Han, Shengzeng Huo, Hes- heng Wang, and David Navarro-Alarcon. Action plan- ning for packing long linear elastic objects into compact boxes with bimanual robotic manipulation. IEEE/ASME Transactions on Mechatronics , 28(3):1718–1729, 2022

    2022

  26. [34]

    Robotic perception-motion synergy for novel rope wrapping tasks

    Zhaoyuan Ma and Jing Xiao. Robotic perception-motion synergy for novel rope wrapping tasks. IEEE Robotics and Automation Letters , 8(7):4131–4138, 2023

    2023

  27. [35]

    Learning reusable ma- nipulation strategies

    Jiayuan Mao, Tom ´as Lozano-P ´erez, Joshua B Tenen- baum, and Leslie Pack Kaelbling. Learning reusable ma- nipulation strategies. In Conference on Robot Learning , pages 1467–1483. PMLR, 2023

    2023

  28. [36]

    Dexskills: Skill segmentation using haptic data for learning autonomous long-horizon robotic manipula- tion tasks

    Xiaofeng Mao, Gabriele Giudici, Claudio Coppola, Kas- par Althoefer, Ildar Farkhatdinov, Zhibin Li, and Lorenzo Jamone. Dexskills: Skill segmentation using haptic data for learning autonomous long-horizon robotic manipula- tion tasks. arXiv preprint arXiv:2405.03476 , 2024

    2024 arXiv

  29. [37]

    Toward robotic manipulation

    Matthew T Mason. Toward robotic manipulation. Annual Review of Control, Robotics, and Autonomous Systems , 1:1–28, 2018

    2018

  30. [38]

    Path planning for de- formable linear objects

    Mark Moll and Lydia E Kavraki. Path planning for de- formable linear objects. IEEE Transactions on Robotics , 22(4):625–636, 2006

    2006

  31. [39]

    Contact-invariant optimization for hand manipulation

    Igor Mordatch, Zoran Popovi ´c, and Emanuel Todorov. Contact-invariant optimization for hand manipulation. In Proceedings of the ACM SIGGRAPH/Eurographics sym- posium on computer animation , pages 137–144, 2012

    2012

  32. [40]

    Deep dynamics models for learning dex- terous manipulation

    Anusha Nagabandi, Kurt Konolige, Sergey Levine, and Vikash Kumar. Deep dynamics models for learning dex- terous manipulation. In Conference on Robot Learning , pages 1101–1112. PMLR, 2020

    2020

  33. [41]

    Dexpbt: Scaling up dexterous manipulation for hand-arm systems with pop- ulation based training

    Aleksei Petrenko, Arthur Allshire, Gavriel State, Ankur Handa, and Viktor Makoviychuk. Dexpbt: Scaling up dexterous manipulation for hand-arm systems with pop- ulation based training. arXiv preprint arXiv:2305.12127 , 2023

    2023 arXiv

  34. [42]

    Dexmv: Im- itation learning for dexterous manipulation from human videos

    Yuzhe Qin, Yueh-Hua Wu, Shaowei Liu, Hanwen Jiang, Ruihan Yang, Yang Fu, and Xiaolong Wang. Dexmv: Im- itation learning for dexterous manipulation from human videos. In European Conference on Computer Vision , pages 570–587. Springer, 2022

    2022

  35. [43]

    Robotic manipulation and sensing of deformable objects in domestic and indus- trial applications: a survey

    Jose Sanchez, Juan-Antonio Corrales, Belhassen-Chedli Bouzgarrou, and Youcef Mezouar. Robotic manipulation and sensing of deformable objects in domestic and indus- trial applications: a survey. The International Journal of Robotics Research, 37(7):688–716, 2018

    2018

  36. [44]

    Leap hand: Low-cost, efficient, and anthropomor- phic hand for robot learning

    Kenneth Shaw, Ananye Agarwal, and Deepak Pathak. Leap hand: Low-cost, efficient, and anthropomor- phic hand for robot learning. arXiv preprint arXiv:2309.06440, 2023

    2023 arXiv

  37. [45]

    Cable ma- nipulation with a tactile-reactive gripper

    Yu She, Shaoxiong Wang, Siyuan Dong, Neha Sunil, Alberto Rodriguez, and Edward Adelson. Cable ma- nipulation with a tactile-reactive gripper. The Interna- tional Journal of Robotics Research , 40(12-14):1385– 1401, 2021

    2021

  38. [46]

    Learning pregrasp manipulation of objects from ungraspable poses

    Zhaole Sun, Kai Yuan, Wenbin Hu, Chuanyu Yang, and Zhibin Li. Learning pregrasp manipulation of objects from ungraspable poses. In 2020 IEEE International Conference on Robotics and Automation (ICRA) , pages 9917–9923. IEEE, 2020

    2020

  39. [47]

    Neural feels with neural fields: Visuo-tactile perception for in-hand manipulation

    Sudharshan Suresh, Haozhi Qi, Tingfan Wu, Taosha Fan, Luis Pineda, Mike Lambeta, Jitendra Malik, Mri- nal Kalakrishnan, Roberto Calandra, Michael Kaess, et al. Neural feels with neural fields: Visuo-tactile perception for in-hand manipulation. arXiv preprint arXiv:2312.13469, 2023

    2023 arXiv

  40. [48]

    Representation for knot- tying tasks

    Jun Takamatsu, Takuma Morita, Koichi Ogawara, Hiroshi Kimura, and Katsushi Ikeuchi. Representation for knot- tying tasks. IEEE Transactions on Robotics, 22(1):65–78, 2006

    2006

  41. [49]

    Hybrid hierarchical learning for solving complex sequential tasks using the robotic ma- nipulation network roman

    Eleftherios Triantafyllidis, Fernando Acero, Zhaocheng Liu, and Zhibin Li. Hybrid hierarchical learning for solving complex sequential tasks using the robotic ma- nipulation network roman. Nature Machine Intelligence , 5(9):991–1005, 2023

    2023

  42. [50]

    Autonomously untangling long cables

    Vainavi Viswanath, Kaushik Shivakumar, Justin Kerr, Brijen Thananjeyan, Ellen Novoseller, Jeffrey Ichnowski, Alejandro Escontrela, Michael Laskey, Joseph E Gonza- lez, and Ken Goldberg. Autonomously untangling long cables. arXiv preprint arXiv:2207.07813 , 2022

    2022 arXiv

  43. [51]

    Self-supervised learning of dynamic planar manipulation of free-end cables

    Jonathan Wang, Huang Huang, Vincent Lim, Harry Zhang, Jeffrey Ichnowski, Daniel Seita, Yunliang Chen, and Ken Goldberg. Self-supervised learning of dynamic planar manipulation of free-end cables. arXiv preprint arXiv:2405.09581, 2024

    2024 arXiv

  44. [52]

    An online method for tight-tolerance insertion tasks for string and rope

    Weifu Wang, Dmitry Berenson, and Devin Balkcom. An online method for tight-tolerance insertion tasks for string and rope. In 2015 IEEE International Conference on Robotics and Automation (ICRA) , pages 2488–2495. IEEE, 2015

    2015

  45. [53]

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

    Yinzhen Xu, Weikang Wan, Jialiang Zhang, Haoran Liu, Zikang Shan, Hao Shen, Ruicheng Wang, Haoran Geng, Yijia Weng, Jiayi Chen, et al. Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy. In Proceedings of the ...

    2023

  46. [54]

    Self-supervised learning of state estimation for manipulating deformable linear objects

    Mengyuan Yan, Yilin Zhu, Ning Jin, and Jeannette Bohg. Self-supervised learning of state estimation for manipulating deformable linear objects. IEEE robotics and automation letters , 5(2):2372–2379, 2020

    2020

  47. [55]

    Learning predictive representations for deformable objects using contrastive estimation

    Wilson Yan, Ashwin Vangipuram, Pieter Abbeel, and Lerrel Pinto. Learning predictive representations for deformable objects using contrastive estimation. In Conference on Robot Learning , pages 564–574. PMLR, 2021

    2021

  48. [56]

    Multi-expert learning of adaptive legged locomotion

    Chuanyu Yang, Kai Yuan, Qiuguo Zhu, Wanming Yu, and Zhibin Li. Multi-expert learning of adaptive legged locomotion. Science Robotics , 5(49):eabb2174, 2020

    2020

  49. [57]

    Exploiting kinematic redundancy for robotic grasping of multiple objects

    Kunpeng Yao and Aude Billard. Exploiting kinematic redundancy for robotic grasping of multiple objects. IEEE Transactions on Robotics , 2023

    2023

  50. [58]

    Mod- eling, learning, perception, and control methods for de- formable object manipulation

    Hang Yin, Anastasia Varava, and Danica Kragic. Mod- eling, learning, perception, and control methods for de- formable object manipulation. Science Robotics , 6(54): eabd8803, 2021

    2021

  51. [59]

    Shape con- trol of deformable linear objects with offline and online learning of local linear deformation models

    Mingrui Yu, Hanzhong Zhong, and Xiang Li. Shape con- trol of deformable linear objects with offline and online learning of local linear deformation models. In 2022 International Conference on Robotics and Automation (ICRA), pages 1337–1343. IEEE, 2022

    2022

  52. [60]

    In-hand following of deformable linear objects using dexterous fingers with tactile sensing

    Mingrui Yu, Boyuan Liang, Xiang Zhang, Xinghao Zhu, Xiang Li, and Masayoshi Tomizuka. In-hand following of deformable linear objects using dexterous fingers with tactile sensing. arXiv preprint arXiv:2403.12676 , 2024

    2024 arXiv

  53. [61]

    Precise robotic needle-threading with tactile perception and reinforcement learning

    Zhenjun Yu, Wenqiang Xu, Siqiong Yao, Jieji Ren, Tutian Tang, Yutong Li, Guoying Gu, and Cewu Lu. Precise robotic needle-threading with tactile perception and reinforcement learning. In Conference on Robot Learning, pages 3266–3276. PMLR, 2023

    2023

  54. [62]

    Gelsight: High-resolution robot tactile sensors for esti- mating geometry and force

    Wenzhen Yuan, Siyuan Dong, and Edward H Adelson. Gelsight: High-resolution robot tactile sensors for esti- mating geometry and force. Sensors, 17(12):2762, 2017

    2017

  55. [63]

    Robots of the lost arc: Self-supervised learning to dynamically manip- ulate fixed-endpoint cables

    Harry Zhang, Jeffrey Ichnowski, Daniel Seita, Jonathan Wang, Huang Huang, and Ken Goldberg. Robots of the lost arc: Self-supervised learning to dynamically manip- ulate fixed-endpoint cables. In 2021 IEEE International Conference on Robotics and Automation (ICRA) , pages 4560–...

    2021

  56. [64]

    Learning fine-grained bimanual manipulation with low-cost hardware

    Tony Z Zhao, Vikash Kumar, Sergey Levine, and Chelsea Finn. Learning fine-grained bimanual manipulation with low-cost hardware. arXiv preprint arXiv:2304.13705 , 2023

    2023 arXiv

  57. [65]

    Aloha unleashed: A simple recipe for robot dexterity

    Tony Z Zhao, Jonathan Tompson, Danny Driess, Pete Florence, Kamyar Ghasemipour, Chelsea Finn, and Ayzaan Wahid. Aloha unleashed: A simple recipe for robot dexterity. arXiv preprint arXiv:2410.13126 , 2024

    2024 arXiv

  58. [66]

    Dexdlo: Learning goal-conditioned dexterous policy for dynamic manipulation of deformable linear objects

    Sun Zhaole, Jihong Zhu, and Robert B Fisher. Dexdlo: Learning goal-conditioned dexterous policy for dynamic manipulation of deformable linear objects. arXiv preprint arXiv:2312.15204, 2023

    2023 arXiv

  59. [67]

    A practical solution to deformable linear object manipula- tion: A case study on cable harness connection

    Hang Zhou, Shunchong Li, Qi Lu, and Jinwu Qian. A practical solution to deformable linear object manipula- tion: A case study on cable harness connection. In 2020 5th International Conference on Advanced Robotics and Mechatronics (ICARM), pages 329–333. IEEE, 2020

    2020

  60. [68]

    Challenges and outlook in robotic manipulation of deformable objects

    Jihong Zhu, Andrea Cherubini, Claire Dune, David Navarro-Alarcon, Farshid Alambeigi, Dmitry Berenson, Fanny Ficuciello, Kensuke Harada, Jens Kober, Xiang Li, et al. Challenges and outlook in robotic manipulation of deformable objects. IEEE Robotics & Automation Magazine, 29(3)...

    2022

  61. [69]

    Finally, the left TIC re-grasps the rotated end-tip to finish the whole direction flipping

    Direction flipping: With the middle finger bent first, the cable needs to be grasped by the right thumb-index combo (TIC), then the middle finger and the right index finger perform adduction gripping (exactly like grasping a pencil), and the released right thumb performs Z-axi...

  62. [70]

    U-shape bending and parallel grasping: The middle finger first bends the cable into a U-shape, then two TICs on both sides perform Z-axis orientation control to rotate the cable’s two sides, and finally two TICs perform parallel grasping

  63. [71]

    Then the right TIC which grasps the cable performs X- axis position control inserts the cable inside the tube repeatedly to reach a certain insertion depth

    In-hand insertion: The object with the hole (the hollow transparent tube) and the cable are grasped separately by two TICs. Then the right TIC which grasps the cable performs X- axis position control inserts the cable inside the tube repeatedly to reach a certain insertion dep...

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.