Pith. sign in

REVIEW 6 major objections 5 minor 5 cited by

DOGlove: Dexterous Manipulation with a Low-Cost Open-Source Haptic Force Feedback Glove

T0 review · 6 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read DOGlove is a sub-600-dollar, open-source haptic glove that pairs 21-DoF hand tracking with force and vibration feedback, and the paper argues this makes low-cost teleoperation of dexterous robotic hands practical.

desk verdict Genuinely low-cost haptic glove with real demos, but the precision claim rests on unvalidated calibration and thresholds tuned on the same tasks. read the letter →

arxiv 2502.07730 v1 pith:E2M35FRT submitted 2025-02-11 cs.RO

classification cs.RO
keywords hapticforcefeedbackglovedexterousteleoperationmotioncaptureretargetingimitationlearninglow-costrobothardwarecontact-richmanipulationcable-drivenactuation
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

DOGlove is a haptic force feedback glove that can be assembled from 3D-printed parts and off-the-shelf components in about six hours for under 600 US dollars, and the paper argues this is enough to teleoperate a dexterous robotic hand precisely through contact-rich tasks. The glove combines 21-degree-of-freedom motion capture with bidirectional cable-driven force feedback on the fingers and vibration at the fingertips, so the operator can feel contacts the robot hand makes. The paper demonstrates that with this feedback, operators can slip a bottle a controlled distance, rotate a carton in hand, and even identify objects by touch alone, and that demonstrations collected with the glove can train imitation learning policies. The central claim is that haptic force feedback, not just motion capture, is what makes low-cost teleoperation of dexterous hands practical.

What carries the argument

The load-bearing object is the glove's kinematic-and-force feedback loop: a customized joint structure with rotary encoders on each joint and five servo-actuated cable-driven joints produces 21 degrees of freedom, while a pulley system with a stainless steel cable transmits torque bidirectionally so the servos can resist or assist finger motion. Fingertip vibration comes from linear resonant actuators driven by a haptic waveform library. The retargeting layer maps glove joint angles through built-in forward kinematics to fingertip positions and uses a differential inverse-kinematics solver to command the target robotic hand, with a scaling factor for hand-size differences. The haptic force retargeting layer thresholds the robot fingertip force readings at 10 grams, 50 grams, and 100 grams, deciding when to activate vibration, force, both, or neither.

What would settle it

Measure the actual positions of the operator's fingertips with an independent motion-capture system while the operator wears DOGlove and moves through a range of grasps, then compare those positions with the fingertip positions that the glove's forward kinematics reports from its encoders. If the mismatch is substantially larger than the paper's accuracy claims, or grows after repeated donning, the retargeting precision claim fails.

Watch

Extended reading notes

Core claim

The paper claims that a glove built at low cost can deliver both precise hand-pose capture and force and haptic feedback sufficient for immersive dexterous teleoperation. DOGlove tracks 21 degrees of freedom through custom joint structures with compact rotary encoders, and it applies force to the fingers through a cable-driven pulley system actuated by five servos, one per finger, while each fingertip carries a linear resonant actuator for vibration. Using forward kinematics of the glove and inverse kinematics of the target robotic hand, fingertip positions are retargeted to robot hand commands, and force readings at the robot fingertips are thresholded and mapped to a combination of force and haptic feedback. In experiments with a low-cost anthropomorphic robot hand mounted on a robot arm, operators achieved high success rates in contact-rich tasks, and the paper reports that removing feedback degrades performance, especially without visual feedback.

Load-bearing premise

The load-bearing premise is that the glove's joint encoders, after calibration, report joint angles accurately enough and that the glove stays rigidly coupled to the operator's hand, so fingertip positions computed by the glove's forward kinematics are accurate enough to transfer to the robot hand.

Editorial extensions

If this is right

  • Low-cost glove-based teleoperation becomes a realistic data-collection tool for robot learning, since the system runs at 30 Hz and produces demonstrations that train imitation policies.
  • Force-sensitive teleoperation no longer requires five-figure commercial gloves; a 3D-printed glove with servo-driven cables is enough for contact-rich tasks.
  • The reported combination thresholds for haptic and force feedback give other designers a concrete starting point for rendering contact information.
  • In-hand manipulation that relies on adjusting friction, such as rotating a carton, becomes feasible because the operator can feel the contact state rather than inferring it visually.
  • The same glove can serve as both the human-robot interface and the demonstration collector for imitation learning, reducing the hardware gap between teleoperation and policy training.

Reading between the lines

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

  • The 600 US dollar cost figure appears to cover the glove itself; adding wrist tracking, the robotic hand, and the robot arm would raise the total system cost, so direct cost comparisons with commercial teleoperation systems should use the full setup.
  • Because the retargeting and feedback success depends on rigid coupling between glove and hand, the same design may need per-user linkage sizes or a quick calibration routine before each session to keep fingertip positions accurate across different hands.
  • The robot-side fingertip force sensors could be logged during demonstration collection, producing force-labeled demonstrations for learning policies that need contact information; the paper trains policies on pose data only.
  • A blindfolded object-perception test with more object pairs and repeated sessions would turn the promising user study into a quantitative benchmark for haptic rendering quality.
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

6 major / 5 minor

Summary. The paper introduces DOGlove, a low-cost open-source haptic force feedback glove for dexterous teleoperation. The hardware combines 21-DoF joint-angle capture using rotary encoders, a 5-DoF cable-driven force feedback mechanism, and fingertip linear resonant actuators, with an action retargeting framework based on forward kinematics and inverse kinematics (Mink). The authors report three categories of evaluation: a user study on object perception without visual feedback, quantitative teleoperation tasks (bottle-slipping and in-hand rotation of a carton), and imitation learning experiments using DOGlove demonstrations to train 3D Diffusion Policy (DP3). The central claim is that DOGlove provides precise, immersive, low-cost teleoperation and that haptic/force feedback materially improves task performance.

Significance. If the claims are borne out, DOGlove would be a valuable community resource: a sub-600 USD, open-source glove with both motion capture and multi-modal feedback, demonstrated across multiple robotic hands and integrated with a modern imitation learning pipeline. The authors commit to releasing mechanical designs, circuit designs, embedded code, URDF models, retargeting code, and simulation environments, which is a notable strength for reproducibility. The paper also reports concrete successes in contact-rich tasks and in collecting demonstrations for DP3. However, several load-bearing validation items are missing or underpowered: the claimed ±1° calibration accuracy is not measured, the haptic/force threshold strategy is derived from and evaluated on the same tasks, and most quantitative comparisons rest on 10 or fewer trials with no statistical analysis. These gaps currently prevent the paper from fully supporting its precision and feedback-benefit claims.

major comments (6)
  1. [§IV-B1] The calibration claim is load-bearing but unverified. The text states that encoder linearity error is ±7.2° (Eq. 1) and that a correction table 'can be reduced to within ±1°', but no calibration protocol, residual distribution, sample size, or repeatability study is reported. Section V-A then relies on the glove acting as a rigid body with accurate joint angles to compute fingertip positions, and all downstream teleoperation and imitation results inherit this assumption. Please provide a calibration experiment with ground-truth joint angles (e.g., a high-precision encoder or optical tracker), report per-encoder residuals, and include a slip/play test across operators or hand sizes.
  2. [§V-B] The force/haptic feedback thresholds in Table I (10 g, 50 g, 100 g) are presented as a 'combination strategy' that was identified during the user study (Section VI-A) and the bottle-slipping experiment (Section VI-B), and the same experimental settings are then used to claim that Haptic+Force improves performance. This is an in-sample tuning procedure: the strategy and the evaluation are not independent, so Table III and Figure 9 do not by themselves demonstrate a generalizable benefit. Please either show that the thresholds are robust (e.g., sensitivity analysis over threshold values) or evaluate the strategy on held-out tasks and operators, and clarify the temporal ordering of threshold selection versus the reported experiments.
  3. [§VI-A, Table II] The user study is too small to support the qualitative conclusions drawn from it. Five operators perform what appears to be a single trial per condition per pair, yielding success counts such as 5/5 vs 4/5 vs 3/5; there are no error bars, no repeated trials, and no statistical tests. The claims that 'force feedback is more sensitive to softness' and 'haptic feedback is more sensitive to shape' are not supported by these counts. Please increase the number of subjects and trials, report per-condition and per-subject results, and apply an appropriate paired significance test or confidence interval.
  4. [§VI-B and §VI-C] The quantitative teleoperation comparisons are underpowered: bottle-slipping experiments and Table III report 10 trials per condition with no variance, confidence intervals, or significance testing. Observed differences such as 9/10 vs 10/10 success or average completion times of 18.92 s vs 19.89 s are not distinguishable from noise. Please provide per-trial data, standard errors or confidence intervals, and paired tests across operators, or explicitly frame these as pilot demonstrations rather than quantitative evidence.
  5. [§VI-C] The AnyTeleop baseline comparison appears potentially inequitable. The paper states that 'due to differences in retargeting strategies, even a slight change in human finger position can lead to a significant deviation', but it does not describe how AnyTeleop was configured (e.g., whether it used the same wrist tracker, the same object positions, and the same operator instructions). A vision-based baseline may be disadvantaged by different calibration or control gains rather than by the glove itself. Please specify the exact baseline setup and, if possible, include an additional glove-based MoCap-only condition that isolates the effect of haptic feedback from the choice of retargeting method.
  6. [§VI-D] The imitation learning results, particularly the long-horizon 'Rotating and Placing the Carton' task, are reported without essential experimental detail. Training a policy on only 3 human demonstrations and achieving 9/10 success is surprising and would benefit substantially from clarification: how many policy training seeds were used, what variance is observed, whether evaluation trials use the same initial conditions as training, and whether the success criterion is identical to that in Section VI-C. Without this information, the claim that DOGlove collects 'high-quality demonstrations' for IL is not yet quantitatively supported.
minor comments (5)
  1. [Abstract] The device name is rendered inconsistently as both DOGlove and DoGlove in the abstract and elsewhere (e.g., 'DoGlove can be assembled in hours'); please standardize to one spelling throughout.
  2. [§IV-B3] The term 'linear resonate actuator' should be 'linear resonant actuator' (and the corresponding acronym LRA is already used correctly).
  3. [§IV-A] The text uses 'metaphalangeal' in the thumb description; the standard anatomical term is 'metacarpophalangeal' (MCP), which is used elsewhere in the paper.
  4. [§III-D] The latency section states maximum frequencies (120 Hz MoCap, 30 Hz haptic) and a minimum system frequency of 30 Hz, but no end-to-end latency measurement is reported; please either provide a measured latency or clarify that these are design specifications.
  5. [§VI] Figure 9a is described as showing success rates, but the plotted quantities and error bars (if any) are not defined in the caption or text; please clarify what the bars represent and whether they are across subjects or trials.

Circularity Check

1 steps flagged · score 6.0 of 10

Haptic 'combination strategy' fit: thresholds in Table I are tuned using the same user study and bottle-slipping experiments that are then reported as evidence that Haptic+Force improves performance; the glove's hardware and IL results otherwise have independent support.

  1. fitted input called prediction [Section V-B (Haptic Force Retargeting), Table I; Section VI (Experiments), Tables II-III]
    "During our quantitative experiments (Section VI), we identify a combination strategy for integrating haptic and force feedback that optimizes performance. This strategy along with the corresponding thresholds and feedback patterns is summarized in Table I. ... During a user study without visual feedback (Section VI-A), we observe that human operators are highly sensitive to force feedback. To create a more realistic experience, force feedback is applied only after the force sensor readings exceed 50 g ..."

    The 10/50/100 g thresholds defining the Haptic+Force condition are explicitly selected using observations from the user study (Section VI-A) and the bottle-slipping experiment (Section VI-B). Section V-B then asserts that 'This combination strategy ... enables human operators to distinguish object shape, size and softness without visual feedback. It also improves performance in complex, contact-rich manipulation tasks,' with Section VI presenting those same user-study and bottle-slipping results plus Table III as the evidence. The favorable Haptic+Force outcome is therefore not an independent test of the strategy; it is partly a restatement of the threshold-fitting objective.

full rationale

The only substantive circularity is in the haptic-feedback evaluation: the combination strategy in Table I (thresholds 10/50/100 g) was designed from observations in the very experiments (VI-A user study, VI-B bottle-slipping) that are used to claim that Haptic+Force improves performance. Because the feedback policy was tuned to those tasks, the reported high success of Haptic+Force is not an independent confirmation. I do not count the unvalidated ±1-degree calibration claim (Section IV-B1: 'With this calibration, the error can be reduced to within ±1°' with no calibration protocol or ground-truth data) as circularity; it is a missing-evidence weakness in the motion-capture accuracy chain, as is the untested rigid-coupling assumption in Section V-A. The cost, mechanical design, FK/IK retargeting, and imitation-learning demonstrations are self-contained and do not reduce to the fitted thresholds. Overall score 6 reflects one fitted input feeding a central haptic-effectiveness claim, while the rest of the system has independent content.

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

The central claim rests on unvalidated encoder calibration, a rigid hand-glove coupling assumption, and fitted feedback thresholds. These are the main items the reader must accept on trust, and they are not backed by shipped calibration data or code in the preprint.

free parameters (4)
  • Force and haptic feedback thresholds = 10g, 50g, 100g
    Selected based on observations from the same user study and bottle-slipping experiments (Section V-B), then used in the evaluation of Haptic+Force, so they are tuned to the tasks.
  • Encoder calibration correction tables = not provided
    Each of the 16 encoders is mapped from voltage to true angle via an external encoder, but the correction data are not shipped; the paper asserts error falls to within ±1 degree without reporting calibration results (Section IV-B1).
  • IK retargeting scaling factor = not provided
    A scaling factor maps human hand size to different robot hands (Section V-A); its value is not given, and per-hand tuning is required.
  • Force-to-KP gain mapping = not specified
    Force readings in [0,3000]g are linearly mapped to Dynamixel KP gain (Section V-B); the slope and offset are not reported, so the mapping is not reproducible.
assumptions (5)
  • domain assumption The human hand kinematics can be represented by the simplified chain from Cerulo et al.: DIP/PIP are 1-DoF hinges, MCP is a 2-DoF ball joint, and the thumb includes a TM ball joint plus wrist pronation/supination.
    This kinematic model underlies the glove's joint structure (Section IV-A) and the forward-kinematic fingertip calculation; if the model is inaccurate for a user's hand, pose capture and retargeting degrade.
  • domain assumption The glove acts as a rigid body rigidly coupled to the human hand, so fingertip positions relative to the glove origin are fully determined by the joint encoders.
    Action retargeting computes fingertip positions by forward kinematics of the glove (Section V-A); this assumes no slip between glove and hand and no soft-tissue deformation, which is not validated.
  • domain assumption Joint angle is a linear function of ADC voltage, with 0 V at 0 degrees and VCC at 360 degrees (Eq. 1).
    The encoder is a variable resistor with ±2 percent linearity error; the linear mapping is assumed after calibration but not validated in the paper.
  • domain assumption The 1D force sensors on the LEAP Hand (3 kg range, 1 g precision) provide sufficient and accurate contact force readings to drive haptic feedback.
    Used as the basis of all force and haptic feedback (Section V-B); force sensor accuracy is taken on trust and not characterized.
  • standard math Mink IK and MuJoCo provide feasible and smooth retargeting for the robot hands.
    The action retargeting relies on Mink (Section V-A) and is not verified for all robot hands beyond the qualitative Figure 7.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DOGlove: Dexterous Manipulation with a Low-Cost Open-Source Haptic Force Feedback Glove." pith.science (2026). https://pith.science/paper/E2M35FRT

@misc{pith2026250207730,
  author       = {Pith},
  title        = {Pith review of: DOGlove: Dexterous Manipulation with a Low-Cost Open-Source Haptic Force Feedback Glove},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E2M35FRT}},
  note         = {Machine review of arXiv:2502.07730}
}
read the original abstract

Dexterous hand teleoperation plays a pivotal role in enabling robots to achieve human-level manipulation dexterity. However, current teleoperation systems often rely on expensive equipment and lack multi-modal sensory feedback, restricting human operators' ability to perceive object properties and perform complex manipulation tasks. To address these limitations, we present DOGlove, a low-cost, precise, and haptic force feedback glove system for teleoperation and manipulation. DoGlove can be assembled in hours at a cost under 600 USD. It features a customized joint structure for 21-DoF motion capture, a compact cable-driven torque transmission mechanism for 5-DoF multidirectional force feedback, and a linear resonate actuator for 5-DoF fingertip haptic feedback. Leveraging action and haptic force retargeting, DOGlove enables precise and immersive teleoperation of dexterous robotic hands, achieving high success rates in complex, contact-rich tasks. We further evaluate DOGlove in scenarios without visual feedback, demonstrating the critical role of haptic force feedback in task performance. In addition, we utilize the collected demonstrations to train imitation learning policies, highlighting the potential and effectiveness of DOGlove. DOGlove's hardware and software system will be fully open-sourced at https://do-glove.github.io/.

Figures

Figures reproduced from arXiv: 2502.07730 by the authors.

Figure 1
Figure 1. DOGlove, a haptic force feedback glove designed for precise teleoperation and dexterous manipulation. It features 21-DoF motion capture and 5-DoF haptic force feedback. By leveraging action and force retargeting, it enables the teleoperation of dexterous hands for complex, contact-rich tasks, including: a) without visual feedback, adjusting contact force with a bottle during teleoperation, b) regulating the flow of … view at source ↗
Figure 2
Figure 2. Teleoperation demos. a) While squeezing condensed milk, the operator regulates the flow using haptic force feedback from DOGlove. b) The operator grasps a slipping bottle without visual feedback. c) The user identifies object pairs solely through haptic force feedback. Vision-based methods are primarily used for tracking the human hand in dexterous hand teleoperation. A simple ap￾proach involves using RGB cameras [2… view at source ↗
Figure 3
Figure 3. The kinematic structure of DOGlove, designed to replicate the kinematics of a human hand. The MCP (B+S) and TM (B+S) joints are modeled as ball joints using a combination of two rotary joints. The right figure from [3] illustrates the simplified human hand kinematics. with 1 DoF, while the additional trapeziometacarpal (TM) joint is a ball joint that supports both flexion-extension and adduction-abduction movements.… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Improper linkage lengths can cause collisions between the human finger (link f, g) and the glove (link m), restricting finger movements and leading to discomfort and poor MoCap performance. The exploded view of a single finger is illustrated in [PITH_FULL_IMAGE:figure…
Figure 5
Figure 5. Figure 5: Exploded view of the finger assembly, with the highlighted area indicating the basic components of a rotary joint. M4×15 shoulder screw to connect the finger linkages, ball bearing, and joint encoder, secured with an M3 locknut. This design ensures smooth and reliable …
Figure 6
Figure 6. Figure 6: Pulley system of the cable-driven mechanism. To address these challenges, DOGlove utilizes a pulley system to provide the bi-directional force feedback, as shown in [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Action retargeting results: Teleoperating the LEAP Hand to grasp a toy in the real world and teleoperating the Shadow Hand, Inspire Hand, and Allegro Hand in simulation. hand’s gestures. For instance, when the human operator opens their hand, the robotic hand open prop…
Figure 8
Figure 8. Figure 8: User Study. (a) Experiment setup: Users wear an eyemask and headphones to eliminate visual and auditory feedback. (b)–(f) Object pairs tested in the study. Task: Five untrained human operators participate in this user study. During the experiment, they are required to …
Figure 9
Figure 9. Figure 9: Teleoperation experiments and quantitative results. a) Without visual feedback, force feedback significantly improves the task success rate. With visual feedback, it enhances precise control. b) In in-hand rotation, the challenge is to slightly release the fingers, all…
Figure 10
Figure 10. Figure 10: The imitation learning experiment. (a) The robot must first locate the correct position of the box and then apply adequate force to press it. Excessive force prevents movement, while insufficient force causes the fingers to slip. (b) The robot must first locate the be…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

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

  1. LocoTouch: Learning Dynamic Quadrupedal Transport with Tactile Sensing

    cs.RO 2025-05 conditional novelty 7.0 of 10

    LocoTouch trains a quadrupedal policy that uses a 221-taxel tactile back to balance and transport unsecured cylindrical objects, transferring zero-shot to a real Unitree Go1.

  2. Towards Human-level Dexterous Teleoperation

    cs.RO 2026-07 conditional novelty 6.0 of 10

    A single-stage RL co-tracking controller trained on consecutive human-derived hand–object subgoals achieves ~75% real-robot success on long-horizon dexterous teleoperation where baselines fail.

  3. BEAVR: Bimanual, multi-Embodiment, Accessible, Virtual Reality Teleoperation System for Robots

    cs.RO 2025-08 conditional novelty 6.0 of 10

    BEAVR provides an open-source, low-cost VR teleoperation pipeline for multiple robot embodiments, with LeRobot-format data recording and compatibility with ACT, Diffusion Policy, and SmolVLA.

  4. DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation

    cs.RO 2025-05 conditional novelty 6.0 of 10

    DexMachina uses decaying virtual object controllers as a curriculum to train bimanual dexterous policies that track demonstrated object states, and reports large gains over baselines on a new six-hand benchmark.

  5. Human-Exoskeleton Kinematic Calibration to Improve Hand Tracking for Dexterous Teleoperation

    cs.RO 2025-07 conditional novelty 5.0 of 10

    A subject-specific calibration framework for the MAESTRO hand exoskeleton reduces joint-angle and fingertip tracking errors relative to uncalibrated tracking, using two reference poses and data-tuned weights.

Reference graph

Works this paper leans on

49 extracted references · 37 canonical work pages · cited by 5 Pith papers

  1. [1]

    Aloha 2: An enhanced low-cost hardware for bimanual teleoperation

    Jorge Aldaco, Travis Armstrong, Robert Baruch, Jeff Bingham, Sanky Chan, Kenneth Draper, De- bidatta Dwibedi, Chelsea Finn, Pete Florence, Spencer Goodrich, et al. Aloha 2: An enhanced low-cost hardware for bimanual teleoperation. arXiv preprint arXiv:2405.02292, 2024

  2. [2]

    Towards generalizable zero-shot manipulation via translating human interaction plans

    Homanga Bharadhwaj, Abhinav Gupta, Vikash Kumar, and Shubham Tulsiani. Towards generalizable zero-shot manipulation via translating human interaction plans. In 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages 6904–6911. IEEE, 2024

  3. [3]

    Teleoperation of the schunk s5fh under- actuated anthropomorphic hand using human hand mo- tion tracking

    Ilaria Cerulo, Fanny Ficuciello, Vincenzo Lippiello, and Bruno Siciliano. Teleoperation of the schunk s5fh under- actuated anthropomorphic hand using human hand mo- tion tracking. Robotics and Autonomous Systems , 89: 75–84, 2017

  4. [4]

    Finger kinematic modeling and real-time hand motion estima- tion

    Pietro Cerveri, Elena De Momi, N Lopomo, Gabriel Baud-Bovy, RML Barros, and Giancarlo Ferrigno. Finger kinematic modeling and real-time hand motion estima- tion. Annals of biomedical engineering , 35:1989–2002, 2007

  5. [5]

    Dexforce: Extract- ing force-informed actions from kinesthetic demon- strations for dexterous manipulation

    Claire Chen, Zhongchun Yu, Hojung Choi, Mark Cutkosky, and Jeannette Bohg. Dexforce: Extract- ing force-informed actions from kinesthetic demon- strations for dexterous manipulation. arXiv preprint arXiv:2501.10356, 2025

  6. [6]

    Open-television: Teleoperation with im- mersive active visual feedback

    Xuxin Cheng, Jialong Li, Shiqi Yang, Ge Yang, and Xiaolong Wang. Open-television: Teleoperation with im- mersive active visual feedback. In 8th Annual Conference on Robot Learning , 2024

  7. [7]

    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

  8. [8]

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

    Cheng Chi, Zhenjia Xu, Chuer Pan, Eric Cousineau, Benjamin Burchfiel, Siyuan Feng, Russ Tedrake, and Shuran Song. Universal manipulation interface: In- the-wild robot teaching without in-the-wild robots. In Proceedings of Robotics: Science and Systems (RSS) , 2024

Show all 49 references
  1. [9]

    Dexta robotics official website

    Dexta Robotics. Dexta robotics official website. https: //www.dextarobotics.com/, 2025

  2. [10]

    Bunny-visionpro: Real-time bimanual dexterous teleoperation for imitation learning

    Runyu Ding, Yuzhe Qin, Jiyue Zhu, Chengzhe Jia, Shiqi Yang, Ruihan Yang, Xiaojuan Qi, and Xiaolong Wang. Bunny-visionpro: Real-time bimanual dexterous teleoperation for imitation learning. arXiv preprint arXiv:2407.03162, 2024

  3. [11]

    Arctic: A dataset for dexterous bimanual hand-object manipulation

    Zicong Fan, Omid Taheri, Dimitrios Tzionas, Muhammed Kocabas, Manuel Kaufmann, Michael J Black, and Otmar Hilliges. Arctic: A dataset for dexterous bimanual hand-object manipulation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1...

  4. [12]

    Airexo: Low-cost exoskeletons for learning whole- arm manipulation in the wild

    Hongjie Fang, Hao-Shu Fang, Yiming Wang, Jieji Ren, Jingjing Chen, Ruo Zhang, Weiming Wang, and Cewu Lu. Airexo: Low-cost exoskeletons for learning whole- arm manipulation in the wild. In 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages 15031–15038....

  5. [13]

    Zhao, and Chelsea Finn

    Zipeng Fu, Tony Z. Zhao, and Chelsea Finn. Mobile aloha: Learning bimanual mobile manipulation with low- cost whole-body teleoperation. In Conference on Robot Learning (CoRL) , 2024

  6. [14]

    Haptx official website

    HaptX. Haptx official website. https://haptx.com/, 2025

  7. [15]

    OPEN TEACH: A versatile teleoperation system for robotic manipulation

    Aadhithya Iyer, Zhuoran Peng, Yinlong Dai, Irmak Guzey, Siddhant Haldar, Soumith Chintala, and Lerrel Pinto. OPEN TEACH: A versatile teleoperation system for robotic manipulation. In 8th Annual Conference on Robot Learning , 2024

  8. [16]

    Tactile sensing in dexterous robot hands

    Zhanat Kappassov, Juan-Antonio Corrales, and V´eronique Perdereau. Tactile sensing in dexterous robot hands. Robotics and Autonomous Systems , 74: 195–220, 2015

  9. [17]

    Six-axis force/torque finger- tip sensor for an anthropomorphic robot hand

    Uikyum Kim, Heeyeon Jeong, Hyunmin Do, Jongwoo Park, and Chanhun Park. Six-axis force/torque finger- tip sensor for an anthropomorphic robot hand. IEEE Robotics and Automation Letters , 5(4):5566–5572, 2020

  10. [18]

    9dtact: A compact vision-based tactile sensor for accurate 3d shape reconstruction and generalizable 6d force estimation

    Changyi Lin, Han Zhang, Jikai Xu, Lei Wu, and Huazhe Xu. 9dtact: A compact vision-based tactile sensor for accurate 3d shape reconstruction and generalizable 6d force estimation. IEEE Robotics and Automation Letters , 2023

  11. [19]

    A glove-based system for study- ing hand-object manipulation via joint pose and force sensing

    Hangxin Liu, Xu Xie, Matt Millar, Mark Edmonds, Feng Gao, Yixin Zhu, Veronica J Santos, Brandon Rothrock, and Song-Chun Zhu. A glove-based system for study- ing hand-object manipulation via joint pose and force sensing. In 2017 IEEE/RSJ International Conference on Intelligent ...

  12. [20]

    High- fidelity grasping in virtual reality using a glove-based system

    Hangxin Liu, Zhenliang Zhang, Xu Xie, Yixin Zhu, Yue Liu, Yongtian Wang, and Song-Chun Zhu. High- fidelity grasping in virtual reality using a glove-based system. In 2019 international conference on robotics and automation (icra) , pages 5180–5186. IEEE, 2019

  13. [21]

    Manus meta robotics official website

    Manus Meta. Manus meta robotics official website. https: //www.manus-meta.com/robotics, 2025

  14. [22]

    Accel- erating interactive human-like manipulation learning with gpu-based simulation and high-quality demonstrations

    Malte Mosbach, Kara Moraw, and Sven Behnke. Accel- erating interactive human-like manipulation learning with gpu-based simulation and high-quality demonstrations. In 2022 IEEE-RAS 21st International Conference on Humanoid Robots (Humanoids) , pages 435–441. IEEE, 2022

  15. [23]

    OptiTrack Motion Capture Systems

    OptiTrack. OptiTrack Motion Capture Systems. https://www.optitrack.com/, 2025

  16. [24]

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

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE conference on computer vision and pattern recog- nition, pages 652–660, 2017

  17. [25]

    Anyteleop: A general vision-based dexterous robot arm- hand teleoperation system

    Yuzhe Qin, Wei Yang, Binghao Huang, Karl Van Wyk, Hao Su, Xiaolong Wang, Yu-Wei Chao, and Dieter Fox. Anyteleop: A general vision-based dexterous robot arm- hand teleoperation system. In Proceedings of Robotics: Science and Systems (RSS) , 2023

  18. [26]

    Scalable

    Felipe Sanches, Geng Gao, Nathan Elangovan, Ricardo V Godoy, Jayden Chapman, Ke Wang, Patrick Jarvis, and Minas Liarokapis. Scalable. intuitive human to robot skill transfer with wearable human machine interfaces: On complex, dexterous tasks. In 2023 IEEE/RSJ Inter- national C...

  19. [27]

    Low-cost, fast and accurate reconstruction of robotic and human postures via imu measurements

    Gaspare Santaera, Emanuele Luberto, Alessandro Serio, Marco Gabiccini, and Antonio Bicchi. Low-cost, fast and accurate reconstruction of robotic and human postures via imu measurements. In 2015 IEEE International Conference on Robotics and Automation (ICRA) , pages 2728–2735. ...

  20. [28]

    Nimbro avatar: Interactive immersive telepresence with force-feedback telemanipu- lation

    Max Schwarz, Christian Lenz, Andre Rochow, Michael Schreiber, and Sven Behnke. Nimbro avatar: Interactive immersive telepresence with force-feedback telemanipu- lation. In 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5312–5319. IEEE, 2021

  21. [29]

    Senseglove official website

    SenseGlove. Senseglove official website. https://www. senseglove.com/, 2025

  22. [30]

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

    Kenneth Shaw, Ananye Agarwal, and Deepak Pathak. Leap hand: Low-cost, efficient, and anthropomorphic hand for robot learning. Robotics: Science and Systems (RSS), 2023

  23. [31]

    Bimanual dexterity for complex tasks

    Kenneth Shaw, Yulong Li, Jiahui Yang, Mohan Kumar Srirama, Ray Liu, Haoyu Xiong, Russell Mendonca, and Deepak Pathak. Bimanual dexterity for complex tasks. In 8th Annual Conference on Robot Learning , 2024

  24. [32]

    Robotic telekinesis: Learning a robotic hand imitator by watching humans on youtube

    Aravind Sivakumar, Kenneth Shaw, and Deepak Pathak. Robotic telekinesis: Learning a robotic hand imitator by watching humans on youtube. In Robotics: Science and Systems, 2022

  25. [33]

    Grasping in the wild: Learning 6dof closed- loop grasping from low-cost demonstrations

    Shuran Song, Andy Zeng, Johnny Lee, and Thomas Funkhouser. Grasping in the wild: Learning 6dof closed- loop grasping from low-cost demonstrations. IEEE Robotics and Automation Letters , 5(3):4978–4985, 2020

  26. [34]

    Grab: A dataset of whole-body human grasping of objects

    Omid Taheri, Nima Ghorbani, Michael J Black, and Dimitrios Tzionas. Grab: A dataset of whole-body human grasping of objects. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16 , pages 581–600. Springer, 2020

  27. [35]

    Ultraleap official website

    Ultraleap. Ultraleap official website. https://www. ultraleap.com/, 2025

  28. [36]

    Vicon Motion Capture Systems

    Vicon Motion Systems. Vicon Motion Capture Systems. https://www.vicon.com/, 2025

  29. [37]

    Mimicplay: Long-horizon imitation learning by watching human play

    Chen Wang, Linxi Fan, Jiankai Sun, Ruohan Zhang, Li Fei-Fei, Danfei Xu, Yuke Zhu, and Anima Anand- kumar. Mimicplay: Long-horizon imitation learning by watching human play. In 7th Annual Conference on Robot Learning , 2023

  30. [38]

    Dexcap: Scalable and portable mocap data collection system for dexterous manipulation

    Chen Wang, Haochen Shi, Weizhuo Wang, Ruohan Zhang, Li Fei-Fei, and C Karen Liu. Dexcap: Scalable and portable mocap data collection system for dexterous manipulation. In Proceedings of Robotics: Science and Systems (RSS) , 2024

  31. [39]

    An immersive virtual reality bimanual telerobotic system with haptic feedback

    Han Xu, Mingqi Chen, Gaofeng Li, Lei Wei, Shichi Peng, Haoliang Xu, and Qiang Li. An immersive virtual reality bimanual telerobotic system with haptic feedback. arXiv preprint arXiv:2501.00822 , 2025

  32. [40]

    XSkill: Cross embodiment skill discovery

    Mengda Xu, Zhenjia Xu, Cheng Chi, Manuela Veloso, and Shuran Song. XSkill: Cross embodiment skill discovery. In 7th Annual Conference on Robot Learning , 2023

  33. [41]

    Flow as the cross-domain manipulation interface

    Mengda Xu, Zhenjia Xu, Yinghao Xu, Cheng Chi, Gordon Wetzstein, Manuela Veloso, and Shuran Song. Flow as the cross-domain manipulation interface. In 8th Annual Conference on Robot Learning , 2024

  34. [42]

    Learning periodic tasks from human demonstrations

    Jingyun Yang, Junwu Zhang, Connor Settle, Akshara Rai, Rika Antonova, and Jeannette Bohg. Learning periodic tasks from human demonstrations. In 2022 International Conference on Robotics and Automation (ICRA) , pages 8658–8665. IEEE, 2022

  35. [43]

    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

  36. [44]

    Mink: Python inverse kinematics based on MuJoCo, July 2024

    Kevin Zakka. Mink: Python inverse kinematics based on MuJoCo, July 2024. URL https://github.com/kevinzakka/ mink. Software

  37. [45]

    3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations

    Yanjie Ze, Gu Zhang, Kangning Zhang, Chenyuan Hu, Muhan Wang, and Huazhe Xu. 3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations. In Proceedings of Robotics: Science and Systems (RSS) , 2024

  38. [46]

    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. In Proceedings of Robotics: Science and Systems (RSS) , 2023

  39. [47]

    Aloha unleashed: A simple recipe for robot dexterity

    Tony Z Zhao, Jonathan Tompson, Danny Driess, Pete Florence, Seyed Kamyar Seyed Ghasemipour, Chelsea Finn, and Ayzaan Wahid. Aloha unleashed: A simple recipe for robot dexterity. In 8th Annual Conference on Robot Learning , 2024

  40. [48]

    Com- bining marker-based mocap and rgb-d camera for acquir- ing high-fidelity hand motion data

    Wenping Zhao, Jinxiang Chai, and Ying-Qing Xu. Com- bining marker-based mocap and rgb-d camera for acquir- ing high-fidelity hand motion data. In Proceedings of the ACM SIGGRAPH/eurographics symposium on computer animation, pages 33–42, 2012

  41. [49]

    Freihand: A dataset for markerless capture of hand pose and shape from single rgb images

    Christian Zimmermann, Duygu Ceylan, Jimei Yang, Bryan Russell, Max Argus, and Thomas Brox. Freihand: A dataset for markerless capture of hand pose and shape from single rgb images. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 813–822, 2019

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.