pith. sign in

arxiv: 2604.15013 · v1 · submitted 2026-04-16 · 💻 cs.RO

DEX-Mouse: A Low-cost Portable and Universal Interface with Force Feedback for Data Collection of Dexterous Robotic Hands

Joonho Koh , Haechan Jung , Nayoung Kim , Wook Ko , Changjoo Nam This is my paper

Pith reviewed 2026-05-10 10:31 UTC · model grok-4.3

classification 💻 cs.RO
keywords dexterous manipulationteleoperation interfaceforce feedbackdemonstration data collectionportable devicerobotic handsuser studyimitation learning
0
0 comments X

The pith

A low-cost handheld interface collects robot-aligned demonstration data for dexterous hands by mounting the target hand on the operator's forearm.

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

The paper introduces DEX-Mouse as a portable teleoperation device built from off-the-shelf parts for under 150 dollars that requires no calibration or per-operator adjustments. It supports an attached mode where the robot hand is fixed directly to the operator's forearm to generate kinematically matched data for training manipulation policies. A user study across dexterous tasks found operators reached an 86.67 percent task completion rate in this attached setup while reporting lower workload than in setups with the hand placed separately. This approach addresses the need for large volumes of physically consistent real-world demonstrations that simulation and video methods often lack due to gaps or retargeting issues.

Core claim

DEX-Mouse provides a calibration-free, portable hand-held teleoperation interface with integrated kinesthetic force feedback that supports an attached configuration in which the target robot hand mounts directly on the operator's forearm, yielding robot-aligned demonstration data and delivering an 86.67 percent task completion rate in comparative studies while reducing perceived workload relative to spatially separated teleoperation setups across all tested interfaces.

What carries the argument

The attached configuration, in which the robot hand is mounted directly on the operator's forearm to produce robot-aligned demonstration data without calibration or structural modifications.

If this is right

  • The calibration-free design allows immediate use across different operators, robot platforms, and environments without setup overhead.
  • Open-sourcing the full hardware bill of materials, CAD models, and firmware enables replication and scaling of data collection efforts.
  • Reduced operator workload in the attached mode supports collection of longer or more varied demonstration sequences for training data.
  • Force feedback provides real-time kinesthetic information that can improve the fidelity of collected trajectories compared to vision-only interfaces.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If forearm mounting truly removes kinematic mismatches, the resulting datasets could train policies that transfer more reliably to physical robots than retargeted video or glove data.
  • Portability opens the possibility of gathering demonstrations in unstructured settings rather than fixed laboratory benches.
  • Universal compatibility across platforms could support standardized benchmarks for comparing imitation learning algorithms on the same demonstration sets.

Load-bearing premise

Mounting the robot hand on the operator's forearm creates demonstration data that matches the robot's true kinematics and perception without adding new mismatches from the mounting or human movement.

What would settle it

A measurement study that records joint-angle errors or end-effector pose deviations between the operator's hand and the mounted robot hand during manipulation tasks, or a larger user study that finds no workload reduction in the attached mode.

Figures

Figures reproduced from arXiv: 2604.15013 by Changjoo Nam, Haechan Jung, Joonho Koh, Nayoung Kim, Wook Ko.

Figure 1
Figure 1. Figure 1: An overview of DEX-Mouse. The proposed USD 150, fully open-sourced teleoperation interface enables scalable collection of high-quality [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Dual tendons facilitate bidirectional finger movement. A single thread [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Attached (left column): The teleoperation interface and the target robot hand are co-located on the arm of the operator. Teleoperation (right column): The operator wears the interface to control a target robot located in a spatially separated, fixed workspace. The rows correspond to the three different interface devices evaluated. is integrated directly onto the device. The tracker employs inside-out track… view at source ↗
Figure 4
Figure 4. Figure 4: An overview of the experimental procedure and evaluated manipulation tasks. The study follows a [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Overall task success rates (%) and completion times (s) across [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Subjective operator experience results evaluating perceived workload [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Demonstration of cross-embodiment compatibility. By employing [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
read the original abstract

Data-driven dexterous hand manipulation requires large-scale, physically consistent demonstration data. Simulation and video-based methods suffer from sim-to-real gaps and retargeting problems, while MoCap glove-based teleoperation systems require per-operator calibration and lack portability, as the robot hand is typically fixed to a stationary arm. Portable alternatives improve mobility but lack cross-platform and cross-operator compatibility. We present DEX-Mouse, a portable, calibration-free hand-held teleoperation interface with integrated kinesthetic force feedback, built from commercial off-the-shelf components under USD 150. The operator-agnostic design requires no calibration or structural modification, enabling immediate deployment across diverse environments and platforms. The interface supports a configuration in which the target robot hand is mounted directly on the forearm of an operator, producing robot-aligned data. In a comparative user study across various dexterous manipulation tasks, operators using the proposed system achieved an 86.67% task completion rate under the attached configuration. Also, we found that the attached configuration reduced the perceived workload of the operators compared to spatially separated teleoperation setups across all compared interfaces. The complete hardware and software stack, including bill of materials, CAD models, and firmware, is open-sourced at https://dex-mouse.github.io/ to facilitate replication and adoption.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces DEX-Mouse, a low-cost (under USD 150), portable, calibration-free teleoperation interface for dexterous robotic hands featuring integrated kinesthetic force feedback and built from COTS components. It supports an attached configuration in which the target robot hand is mounted directly on the operator's forearm to produce robot-aligned demonstration data. A comparative user study across dexterous manipulation tasks reports an 86.67% task completion rate under the attached configuration and reduced perceived workload relative to spatially separated teleoperation setups. The complete hardware/software stack (BOM, CAD models, firmware) is open-sourced.

Significance. If the empirical claims hold under proper validation, the work provides a practical, accessible tool for collecting large-scale, physically consistent demonstration data that avoids sim-to-real gaps, retargeting issues, and per-operator calibration. The open-sourcing of all replication materials is a clear strength that supports adoption across platforms and operators. The attached configuration and force-feedback integration address longstanding portability and alignment limitations in teleoperation for data-driven dexterous manipulation.

major comments (2)
  1. [Abstract] Abstract: The headline empirical results (86.67% completion rate and workload reduction in attached mode) are stated without any information on participant count, task definitions, statistical tests, controls, or variance, leaving the central claims about the advantages of the attached configuration with insufficient verifiable support.
  2. [Attached configuration description] Attached configuration description: The assertion that forearm mounting yields truly robot-aligned data without new kinematic, force-feedback, or visual-proprioceptive mismatches lacks any quantitative validation (e.g., joint-angle correlation, end-effector pose error, or retargeting residual metrics comparing attached vs. separated modes). Observed performance differences could arise from reduced visual separation or interface familiarity rather than alignment.
minor comments (2)
  1. [Hardware figures] Figure captions and hardware diagrams should include explicit scale bars and component labels to aid replication of the low-cost assembly.
  2. [Open-source section] The open-source repository link is provided, but the manuscript should include a brief summary table of the bill of materials with current prices and sourcing links.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment point-by-point below, indicating where revisions will be made to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The headline empirical results (86.67% completion rate and workload reduction in attached mode) are stated without any information on participant count, task definitions, statistical tests, controls, or variance, leaving the central claims about the advantages of the attached configuration with insufficient verifiable support.

    Authors: We agree that the abstract would benefit from additional context on the user study to make the empirical claims more verifiable at a glance. In the revised manuscript we will expand the abstract to include the participant count, a brief description of the dexterous manipulation tasks, reference to the statistical tests performed, and mention of observed variance, while preserving overall length. The full experimental protocol, controls, and statistical details already appear in the Experiments section and will be cross-referenced. revision: yes

  2. Referee: [Attached configuration description] Attached configuration description: The assertion that forearm mounting yields truly robot-aligned data without new kinematic, force-feedback, or visual-proprioceptive mismatches lacks any quantitative validation (e.g., joint-angle correlation, end-effector pose error, or retargeting residual metrics comparing attached vs. separated modes). Observed performance differences could arise from reduced visual separation or interface familiarity rather than alignment.

    Authors: We acknowledge that the current manuscript does not supply direct quantitative metrics (joint-angle correlation, pose error, etc.) comparing kinematic or proprioceptive alignment between the attached and separated configurations. The attached setup is motivated by the design goal of co-locating the operator’s and robot’s hands to minimize spatial offsets; the observed improvements in task completion and workload are consistent with this intent. However, we agree that alternative factors such as reduced visual separation or familiarity cannot be excluded without additional data. In the revision we will add an explicit discussion of these potential confounds and will include any available pilot alignment observations. A comprehensive quantitative validation would require new experiments that are outside the scope of the present work. revision: partial

Circularity Check

0 steps flagged

No circularity; claims rest on hardware description and direct empirical user-study outcomes

full rationale

The manuscript presents a hardware interface design (DEX-Mouse) built from COTS components and reports empirical results from a comparative user study on task completion rates and perceived workload. No mathematical derivations, equations, fitted parameters, or first-principles predictions are present. Central claims about portability, calibration-free operation, and performance advantages of the attached configuration are supported by direct measurements and open-sourced artifacts rather than any reduction to self-referential inputs, self-citations, or ansatzes. The analysis is therefore self-contained with no load-bearing steps that collapse by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an applied engineering paper describing a hardware prototype and user study results. No mathematical derivations, free parameters, or postulated entities are involved in the central claims. The performance claims rest entirely on the described user study.

pith-pipeline@v0.9.0 · 5546 in / 1319 out tokens · 83561 ms · 2026-05-10T10:31:45.810541+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [1]

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

    C. Chi, Z. Xu, C. Pan, E. Cousineau, B. Burchfiel, S. Feng, R. Tedrake, and S. Song, “Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots,” inProc. of Robot. Sci. Syst. (RSS), 2024

    work page 2024

  2. [2]

    Learning fine-grained bimanual manipulation with low-cost hardware,

    T. Zhao, V . Kumar, S. Levine, and C. Finn, “Learning fine-grained bimanual manipulation with low-cost hardware,” inProc. of Robot. Sci. Syst. (RSS), 2023

    work page 2023

  3. [3]

    Bridgedata v2: A dataset for robot learning at scale,

    H. R. Walke, K. Black, T. Z. Zhao, Q. Vuong, C. Zheng, P. Hansen- Estruch, A. W. He, V . Myers, M. J. Kim, M. Du, A. Lee, K. Fang, C. Finn, and S. Levine, “Bridgedata v2: A dataset for robot learning at scale,” inProc. Conf. Robot Learn. (CoRL), 2023, pp. 1723–1736

    work page 2023

  4. [4]

    Sim-and-real co-training: A simple recipe for vision-based robotic manipulation,

    A. Maddukuri, Z. Jiang, L. Y . Chen, S. Nasiriany, Y . Xie, Y . Fang, W. Huang, Z. Wang, Z. Xu, N. Chernyadev, S. Reed, K. Goldberg, A. Mandlekar, L. Fan, and Y . Zhu, “Sim-and-real co-training: A simple recipe for vision-based robotic manipulation,” inProc. of Robot. Sci. Syst. (RSS), 2025

    work page 2025

  5. [5]

    Video prediction policy: A generalist robot policy with predictive visual representations,

    Y . Hu, Y . Guo, P. Wang, X. Chen, Y .-J. Wang, J. Zhang, K. Sreenath, C. Lu, and J. Chen, “Video prediction policy: A generalist robot policy with predictive visual representations,” inProc. of Int. Conf. Mach. Learn. (ICML), 2025, pp. 24 328–24 346

    work page 2025

  6. [6]

    Video2policy: Scaling up manipulation tasks in simulation through internet videos

    W. Ye, F. Liu, Z. Ding, Y . Gao, O. Rybkin, and P. Abbeel, “Video2policy: Scaling up manipulation tasks in simulation through internet videos,”arXiv preprint arXiv:2502.09886, 2025

    work page arXiv 2025

  7. [7]

    Doglove: Dexterous manip- ulation with a low-cost open-source haptic force feedback glove,

    H. Zhang, S. Hu, Z. Yuan, and H. Xu, “Doglove: Dexterous manip- ulation with a low-cost open-source haptic force feedback glove,” in Proc. of Robot. Sci. Syst. (RSS), 2025

    work page 2025

  8. [8]

    Bimanual dexterity for complex tasks,

    K. Shaw, Y . Li, J. Yang, M. K. Srirama, R. Liu, H. Xiong, R. Men- donca, and D. Pathak, “Bimanual dexterity for complex tasks,” inProc. Conf. Robot Learn. (CoRL), 2025, pp. 5166–5183

    work page 2025

  9. [9]

    Dexumi: Using human hand as the universal manipulation interface for dexterous manipulation,

    M. Xu, H. Zhang, Y . Hou, Z. Xu, L. Fan, M. Veloso, and S. Song, “Dexumi: Using human hand as the universal manipulation interface for dexterous manipulation,” inProc. Conf. Robot Learn. (CoRL), 2025, pp. 437–459

    work page 2025

  10. [10]

    Fang et al.,DEXOP: A device for robotic transfer of dexterous human manipulation, 2025

    H.-S. Fang, B. Romero, Y . Xie, A. Hu, B.-R. Huang, J. Alvarez, M. Kim, G. Margolis, K. Anbarasu, M. Tomizukaet al., “Dexop: A device for robotic transfer of dexterous human manipulation,”arXiv preprint arXiv:2509.04441, 2025

    work page arXiv 2025

  11. [11]

    R3m: A universal visual representation for robot manipulation,

    S. Nair, A. Rajeswaran, V . Kumar, C. Finn, and A. Gupta, “R3m: A universal visual representation for robot manipulation,” inProc. Conf. Robot Learn. (CoRL), 2023, pp. 892–909

    work page 2023

  12. [12]

    VIP: Towards universal visual reward and representation via value-implicit pre-training,

    Y . J. Ma, S. Sodhani, D. Jayaraman, O. Bastani, V . Kumar, and A. Zhang, “VIP: Towards universal visual reward and representation via value-implicit pre-training,” inProc. of Int. Conf. Learn. Represent. (ICLR), 2023

    work page 2023

  13. [13]

    Real-world robot learning with masked visual pre-training,

    I. Radosavovic, T. Xiao, S. James, P. Abbeel, J. Malik, and T. Darrell, “Real-world robot learning with masked visual pre-training,” inProc. Conf. Robot Learn. (CoRL), 2023, pp. 416–426

    work page 2023

  14. [14]

    Dexpilot: Vision-based teleop- eration of dexterous robotic hand-arm system,

    A. Handa, K. Van Wyk, W. Yang, J. Liang, Y .-W. Chao, Q. Wan, S. Birchfield, N. Ratliff, and D. Fox, “Dexpilot: Vision-based teleop- eration of dexterous robotic hand-arm system,” inProc. of IEEE Int. Conf. Robot. Automat. (ICRA), 2020, pp. 9164–9170

    work page 2020

  15. [15]

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

    Y . Qin, W. Yang, B. Huang, K. Van Wyk, H. Su, X. Wang, Y .-W. Chao, and D. Fox, “Anyteleop: A general vision-based dexterous robot arm- hand teleoperation system,” inProc. of Robot. Sci. Syst. (RSS), 2023

    work page 2023

  16. [16]

    Ace: A cross-platform and visual-exoskeletons system for low-cost dexterous teleoperation,

    S. Yang, M. Liu, Y . Qin, R. Ding, J. Li, X. Cheng, R. Yang, S. Yi, and X. Wang, “Ace: A cross-platform and visual-exoskeletons system for low-cost dexterous teleoperation,” inProc. Conf. Robot Learn. (CoRL), 2025, pp. 4895–4911

    work page 2025

  17. [17]

    Open-television: Teleoperation with immersive active visual feedback,

    X. Cheng, J. Li, S. Yang, G. Yang, and X. Wang, “Open-television: Teleoperation with immersive active visual feedback,” inProc. Conf. Robot Learn. (CoRL), 2025, pp. 2729–2749

    work page 2025

  18. [18]

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

    R. Ding, Y . Qin, J. Zhu, C. Jia, S. Yang, R. Yang, X. Qi, and X. Wang, “Bunny-visionpro: Real-time bimanual dexterous teleoperation for imitation learning,” inProc. of IEEE/RSJ Int. Conf. Intell. Robot. Syst. (IROS), 2025, pp. 12 248–12 255

    work page 2025

  19. [19]

    Open teach: A versatile teleoperation system for robotic manipulation,

    A. Iyer, Z. Peng, Y . Dai, I. Guzey, S. Haldar, S. Chintala, and L. Pinto, “Open teach: A versatile teleoperation system for robotic manipulation,” inProc. Conf. Robot Learn. (CoRL), 2025, pp. 2372– 2395

    work page 2025

  20. [20]

    Dex- cap: Scalable and portable mocap data collection system for dexterous manipulation,

    C. Wang, H. Shi, W. Wang, R. Zhang, L. Fei-Fei, and C. K. Liu, “Dexcap: Scalable and portable mocap data collection system for dexterous manipulation,”arXiv preprint arXiv:2403.07788, 2024

    work page arXiv 2024

  21. [21]

    Snu-avatar haptic glove: Novel modularized haptic glove via trigonometric series elastic actuators,

    E. Sung, S. You, S. Moon, J. Kim, and J. Park, “Snu-avatar haptic glove: Novel modularized haptic glove via trigonometric series elastic actuators,” inProc. of IEEE/RSJ Int. Conf. Intell. Robot. Syst. (IROS), 2024, pp. 3573–3580

    work page 2024

  22. [22]

    2012 anthropometric survey of U.S. army personnel: Methods and summary statistics,

    C. C. Gordon, C. L. Blackwell, B. Bradtmiller, J. L. Parham, P. Bar- rientos, S. P. Paquette, B. D. Corner, J. M. Carson, J. C. Venezia, B. M. Rockwellet al., “2012 anthropometric survey of U.S. army personnel: Methods and summary statistics,” U.S. Army Natick Soldier Research, Development and Engineering Center, Natick, MA, Tech. Rep. NATICK/TR-15/007, 2014

    work page 2012

  23. [23]

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

    M. R. Cutkosky, “On grasp choice, grasp models, and the design of hands for manufacturing tasks,”IEEE Trans. on Robot. and Automat., vol. 5, no. 3, pp. 269–279, 1989

    work page 1989

  24. [24]

    The grasp taxonomy: Orthogonality status in human grasp types,

    T. Feix, J. Romero, H.-B. Schmiedmayer, A. M. Dollar, and D. Kragic, “The grasp taxonomy: Orthogonality status in human grasp types,” IEEE Trans. on Human-Machine Syst., vol. 46, no. 1, pp. 66–77, 2015

    work page 2015

  25. [25]

    VIVE ultimate tracker,

    HTC Corporation, “VIVE ultimate tracker,” [Online]. Available: https: //www.vive.com/kr/accessory/vive-ultimate-tracker/. Accessed: Apr. 4, 2026

    work page 2026

  26. [26]

    Quantum metagloves,

    Manus Meta, “Quantum metagloves,” [Online]. Available: https:// www.manus-meta.com/products/quantum-metagloves. Accessed: Feb. 12, 2025

    work page 2025

  27. [27]

    Blue Robin dexterous robot hand,

    Blue Robin, “Blue Robin dexterous robot hand,” [Online]. Available: https://www.bluerobin.co.kr/EN/include/pages/products/robot_hand. php. Accessed: Feb. 12, 2025

    work page 2025

  28. [28]

    The Pinocchio C++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives,

    J. Carpentier, G. Saurel, G. Buondonno, J. Mirabel, F. Lamiraux, O. Stasse, and N. Mansard, “The Pinocchio C++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives,” inProc. IEEE/SICE Int. Symp. Syst. Integr. (SII), Paris, France, Jan. 2019

    work page 2019

  29. [29]

    Nasa-task load index (NASA-TLX); 20 years later,

    S. G. Hart, “Nasa-task load index (NASA-TLX); 20 years later,” in Proc. of Hum. Factors Ergonom. Soc. Annu. Meet. (HFES), vol. 50, no. 9, 2006, pp. 904–908

    work page 2006

  30. [30]

    The sense of agency scale: A measure of consciously perceived control over one’s mind, body, and the immediate environment,

    A. Tapal, E. Oren, R. Dar, and B. Eitam, “The sense of agency scale: A measure of consciously perceived control over one’s mind, body, and the immediate environment,”Frontiers in Psychology, vol. 8, p. 1552, 2017

    work page 2017

  31. [31]

    Learning complex dexterous manipulation with deep reinforcement learning and demonstrations,

    A. Rajeswaran, V . Kumar, A. Gupta, G. Vezzani, J. Schulman, E. Todorov, and S. Levine, “Learning complex dexterous manipulation with deep reinforcement learning and demonstrations,” inProc. of Robot. Sci. Syst. (RSS), 2018

    work page 2018

  32. [32]

    IGRIS-C: Compact artificial intelligence humanoid robot,

    ROBROS, “IGRIS-C: Compact artificial intelligence humanoid robot,” [Online]. Available: https://robros.co.kr/igrisc/. Accessed: Apr. 4, 2026

    work page 2026

  33. [33]

    FR5 collaborative robot,

    FAIR Innovation, “FR5 collaborative robot,” [Online]. Available: https: //www.frtech.fr/FR/4.html. Accessed: Apr. 14, 2026

    work page 2026