pith. machine review for the scientific record. sign in

arxiv: 2602.22088 · v2 · submitted 2026-02-25 · 💻 cs.RO

Recognition: 2 theorem links

· Lean Theorem

Force Policy: Learning Hybrid Force-Position Control Policy under Interaction Frame for Contact-Rich Manipulation

Hongjie Fang , Shirun Tang , Mingyu Mei , Haoxiang Qin , Zihao He , Jingjing Chen , Ying Feng , Chenxi Wang
show 4 more authors
Wanxi Liu Zaixing He Cewu Lu Shiquan Wang
Authors on Pith no claims yet

Pith reviewed 2026-05-15 19:33 UTC · model grok-4.3

classification 💻 cs.RO
keywords contact-rich manipulationhybrid force-position controlinteraction framevision-force policyglobal-local policyforce regulationrobot learning from demonstration
0
0 comments X

The pith

Recovering an interaction frame from demonstrations lets a global vision policy hand off to a local high-frequency hybrid force-position controller for stable contact.

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

The paper aims to show that contact-rich manipulation improves when a robot separates long-range vision-guided motion from short-range force-regulated contact using a recovered interaction frame. Existing methods either mix these roles in one network or assume fixed task structure, leading to poor generalization or unstable touching. By recovering the frame from demos, the approach lets a global policy handle free-space actions while a local policy takes over on contact to run hybrid control at high frequency with force feedback. Real-world tests on varied tasks demonstrate steadier contact, tighter force tracking, and better performance on unseen objects. A reader would care because this structure addresses the core tension between broad planning and precise physical interaction without needing full dynamics models.

Core claim

We formalize a physically grounded interaction frame as an instantaneous local basis recovered from demonstrations that decouples force regulation from motion execution. Using this, Force Policy combines a global vision-based policy for free-space actions with a high-frequency local policy that estimates the frame on contact and executes hybrid force-position control, yielding more robust contact establishment, accurate force regulation, and generalization to novel objects across diverse contact-rich tasks.

What carries the argument

The interaction frame: an instantaneous local basis recovered from demonstrations that decouples force regulation from motion execution and enables the switch to hybrid force-position control.

If this is right

  • The method produces more robust contact establishment than monolithic or parameter-only baselines across real-world tasks.
  • Force regulation becomes more accurate because the local policy operates directly in the recovered frame.
  • Generalization to objects with new geometries and physical properties holds without retraining the full policy.
  • Both contact stability and overall task execution quality improve as a direct result of the global-local split.

Where Pith is reading between the lines

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

  • The frame recovery step could be extended to online refinement using current force measurements if initial demo-based estimates drift.
  • Similar global-local splits might apply to other hybrid control domains such as tool use where vision sets approach and force refines insertion.
  • If frame estimation proves reliable, the need for expensive high-precision force-torque sensors at every joint might decrease for many contact tasks.

Load-bearing premise

The interaction frame recovered from demonstrations remains accurate enough and the high-frequency local policy stabilizes contact without explicit dynamics models or extra sensing.

What would settle it

A trial in which the local policy fails to maintain stable contact forces when the recovered frame orientation deviates more than 15 degrees from the true surface normal on a novel object geometry would falsify the claim.

Figures

Figures reproduced from arXiv: 2602.22088 by Cewu Lu, Chenxi Wang, Haoxiang Qin, Hongjie Fang, Jingjing Chen, Mingyu Mei, Shiquan Wang, Shirun Tang, Wanxi Liu, Ying Feng, Zaixing He, Zihao He.

Figure 1
Figure 1. Figure 1: A Global-Local Vision-Force Policy Inspired by Human Interaction. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Interaction Frame for Example Contact-Rich Tasks. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Force Policy and Dual-Policy Asynchronous Scheduler. (Left) Force Policy consists of a global vision policy and a local force policy. The global policy provides task-level visual context and global actions, while the local policy predicts interaction structure and local actions to realize hybrid force-position control during contact. (Right) The dual-policy asynchronous scheduler switches between the two p… view at source ↗
Figure 4
Figure 4. Figure 4: Tasks. We design three tasks spanning two categories (polishing and insertion) to evaluate different policies for contact-rich manipulation. The descriptions on the right highlight the key challenges of each task compared to similar tasks in prior literature. All tasks require highly accurate force regulation to be successfully completed. We randomize object placement within the workspace area during both … view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of Effective Forces during Deployment and [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Asynchronous Scheduler Evaluation on the [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: IF Recovery Evaluation on the Scrape off Sticker Task. Angular error is computed between the recovered and ground-truth force control axes; a failure is counted if it exceeds 20◦ . ground-truth IF aligns with the world vertical axis in this task, so we measure angular error and count failures above 20◦ . The results are shown in [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Sensor Signal Ambiguity. Similar sensor signals (wrench and twist) can lead to different interaction types and structures [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of the Interaction Frame and the Task Mode on the [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Visualization of the Interaction Frame and the Task Mode on the [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Visualization of the Interaction Frame and the Task Mode on the  #) % #)  ."#  #) [PITH_FULL_IMAGE:figures/full_fig_p016_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: Evaluation Metrics Explanation of Partial Success. [PITH_FULL_IMAGE:figures/full_fig_p019_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Failure Analyses on the Push and Flip Task. Most failures arise from attempting to flip the object without first establishing contact with the wall. Force-aware policies using position control often exhibit unstable or unintended contacts. ForceVLA TA -VLA No Contact with Surface Unable to Make Contact FoAR Lose Contact TA -VLA Applying Excessive Force Applied Force Not Enough [PITH_FULL_IMAGE:figures/fu… view at source ↗
Figure 15
Figure 15. Figure 15: Failure Analyses on the [PITH_FULL_IMAGE:figures/full_fig_p020_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Objects in Generalization Evaluation. Unseen objects of different colors, geometries, and stiffnesses are selected to evaluate the generalization ability of the policies. • Geometric Variation. The objects feature distinct geo￾metric profiles, ranging from standard cuboids to irregular shapes and cylinders (e.g., cylindrical cup). Variations in aspect ratio and edge curvature challenge the policy’s abilit… view at source ↗
Figure 17
Figure 17. Figure 17: Contact Robustness of Force Policy under Disturbances in the Push and Flip Task. Force Policy is robust under human disturbances and can perform autonomous recovery during the contact phase. fully generalizes across these categories, attributing its success to the explicit modeling of the interaction frame and the adap￾tive fusion of proprioceptive feedback, which compensates for visual ambiguities and ge… view at source ↗
Figure 18
Figure 18. Figure 18: Demonstration Overview for Each Task. We visualize the initial configurations from 50 demonstrations for each task [PITH_FULL_IMAGE:figures/full_fig_p022_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Evaluation Configurations for Each Task. [PITH_FULL_IMAGE:figures/full_fig_p023_19.png] view at source ↗
read the original abstract

Contact-rich manipulation demands human-like integration of perception and force feedback: vision should guide task progress, while high-frequency interaction control must stabilize contact under uncertainty. Existing learning-based policies often entangle these roles in a monolithic network, trading off global generalization against stable local refinement, while control-centric approaches typically assume a known task structure or learn only controller parameters rather than the structure itself. In this paper, we formalize a physically grounded interaction frame, an instantaneous local basis that decouples force regulation from motion execution, and propose a method to recover it from demonstrations. Based on this, we address both issues by proposing Force Policy, a global-local vision-force policy in which a global policy guides free-space actions using vision, and upon contact, a high-frequency local policy with force feedback estimates the interaction frame and executes hybrid force-position control for stable interaction. Real-world experiments across diverse contact-rich tasks show consistent gains over strong baselines, with more robust contact establishment, more accurate force regulation, and reliable generalization to novel objects with varied geometries and physical properties, ultimately improving both contact stability and execution quality. Project page: https://force-policy.github.io/

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

1 major / 1 minor

Summary. The paper introduces Force Policy, a global-local vision-force policy for contact-rich manipulation. It formalizes a physically grounded interaction frame recovered from demonstrations to decouple force regulation from motion execution. A global policy uses vision for free-space actions, while a high-frequency local policy employs force feedback for hybrid force-position control upon contact. Real-world experiments on diverse tasks demonstrate consistent improvements over baselines in contact establishment, force regulation, and generalization to novel objects with different geometries and properties.

Significance. If the results hold, this work could significantly advance learning-based approaches to hybrid control in robotics by providing a structured way to integrate perception and force feedback without requiring explicit dynamics models or additional sensing. The real-world validation across multiple tasks and generalization claims are notable strengths, though the absence of detailed quantitative metrics, baselines, and ablations in the abstract makes it difficult to fully assess the magnitude of the contribution.

major comments (1)
  1. [Abstract] Abstract: The central claim of reliable generalization to novel objects with varied geometries and physical properties depends on the interaction frame recovered from demonstrations remaining valid under unmodeled contact variations (e.g., friction, compliance). No derivation or ablation is provided showing how frame estimation tolerates these variations, which is load-bearing for the assertion that the local policy stabilizes contact without explicit dynamics modeling.
minor comments (1)
  1. [Abstract] Abstract: Claims of 'consistent gains over strong baselines' and 'more robust contact establishment' are stated without any quantitative metrics, specific baseline comparisons, or ablation results, which weakens the ability to evaluate the strength of evidence.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive comments. We address the concern regarding the interaction frame's robustness to unmodeled variations below and commit to strengthening the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of reliable generalization to novel objects with varied geometries and physical properties depends on the interaction frame recovered from demonstrations remaining valid under unmodeled contact variations (e.g., friction, compliance). No derivation or ablation is provided showing how frame estimation tolerates these variations, which is load-bearing for the assertion that the local policy stabilizes contact without explicit dynamics modeling.

    Authors: We appreciate the referee highlighting this load-bearing aspect of our claims. The interaction frame is recovered directly from force-torque measurements in the demonstrations by computing the principal axes of the observed force and velocity vectors at contact; this procedure is physically grounded in the fact that contact forces align with the surface normal while tangential components reflect motion along the surface. Because the recovery uses real sensor data rather than a pre-specified model, it adapts to the instantaneous geometry and force distribution, providing inherent tolerance to moderate variations in friction and compliance. Our real-world results across objects with differing shapes, stiffnesses, and surface properties support this empirically. To address the absence of explicit analysis, we will add to the revised manuscript: (i) a short derivation showing that bounded perturbations in friction produce bounded errors in the recovered tangent directions that are corrected by the high-frequency force feedback loop, and (ii) a new ablation that injects controlled variations in friction coefficient and object compliance (both in simulation and on hardware) and reports frame estimation error together with downstream force-regulation and task-success metrics. These additions will quantify the claimed robustness. revision: yes

Circularity Check

0 steps flagged

No circularity detected; derivation relies on external demonstrations and hardware validation

full rationale

The paper formalizes an interaction frame recovered from demonstrations and deploys a global-local policy evaluated on physical hardware across novel objects. No equations or steps reduce by construction to fitted inputs, self-citations, or renamed ansatzes; the interaction frame is treated as an externally recovered quantity rather than defined in terms of the policy output. Central claims rest on empirical gains over baselines, not on internal reparameterization of the same data.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim depends on the recoverability of the interaction frame from demonstrations and the stability of the local hybrid controller under uncertainty; no explicit free parameters are named in the abstract.

axioms (1)
  • domain assumption An instantaneous local interaction frame exists and can be recovered from force demonstrations to decouple force regulation from motion execution.
    This premise is invoked to justify the hybrid control structure and is central to the method's claimed advantage.
invented entities (1)
  • Interaction frame no independent evidence
    purpose: Local basis that decouples force regulation from motion execution
    New coordinate system introduced to structure the policy; no independent falsifiable evidence provided beyond the learning procedure itself.

pith-pipeline@v0.9.0 · 5541 in / 1242 out tokens · 45603 ms · 2026-05-15T19:33:16.961989+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

93 extracted references · 93 canonical work pages · 1 internal anchor

  1. [1]

    Learning Diffusion Policies from Demonstrations for Compliant Contact-Rich Manipulation

    Malek Aburub et al. “Learning Diffusion Policies from Demonstrations for Compliant Contact-Rich Manipulation”. In:arXiv preprint arXiv:2410.19235(2024)

    work page arXiv 2024

  2. [2]

    SAIL: Faster-than- Demonstration Execution of Imitation Learning Policies

    Nadun Ranawaka Arachchige et al. “SAIL: Faster-than- Demonstration Execution of Imitation Learning Policies”. In:Conference on Robot Learning. V ol. 305. PMLR, 2025, pp. 721–749

    work page 2025

  3. [3]

    On the Analysis of Move- ment Smoothness

    Sivakumar Balasubramanian et al. “On the Analysis of Move- ment Smoothness”. In:Journal of Neuroengineering and Re- habilitation12.1 (2015), p. 112

    work page 2015

  4. [4]

    Cambridge university press, 1998

    Robert Stawell Ball.A Treatise on the Theory of Screws. Cambridge university press, 1998

    work page 1998

  5. [5]

    Echo: An Open-Source, Low-Cost Teleoperation System with Force Feedback for Dataset Collec- tion in Robot Learning

    Artem Bazhenov et al. “Echo: An Open-Source, Low-Cost Teleoperation System with Force Feedback for Dataset Collec- tion in Robot Learning”. In:arXiv preprint arXiv:2504.07939 (2025)

    work page arXiv 2025

  6. [6]

    GR00T N1: An Open Foundation Model for Generalist Humanoid Robots

    Johan Bjorck et al. “GR00T N1: An Open Foundation Model for Generalist Humanoid Robots”. In:arXiv preprint arXiv:2503.14734(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  7. [7]

    Real- Time Execution of Action Chunking Flow Policies

    Kevin Black, Manuel Y Galliker, and Sergey Levine. “Real- Time Execution of Action Chunking Flow Policies”. In:Ad- vances in Neural Information Processing Systems. 2025

    work page 2025

  8. [8]

    𝜋 0.5: a Vision-Language-Action Model with Open-World Generalization

    Kevin Black et al. “𝜋 0.5: a Vision-Language-Action Model with Open-World Generalization”. In:Conference on Robot Learning. V ol. 305. PMLR, 2025, pp. 17–40

    work page 2025

  9. [9]

    𝜋 0: A Vision-Language-Action Flow Model for General Robot Control

    Kevin Black et al. “𝜋 0: A Vision-Language-Action Flow Model for General Robot Control”. In:Robotics: Science and Systems. 2025

    work page 2025

  10. [10]

    Training-time action conditioning for efficient real-time chunking.arXiv preprint arXiv:2512.05964, 2025

    Kevin Black et al. “Training-Time Action Conditioning for Efficient Real-Time Chunking”. In:arXiv preprint arXiv:2512.05964(2025)

    work page arXiv 2025

  11. [11]

    RT-1: Robotics Transformer for Real- World Control at Scale

    Anthony Brohan et al. “RT-1: Robotics Transformer for Real- World Control at Scale”. In:Robotics: Science and Systems. 2023

    work page 2023

  12. [12]

    Specification of Force-Controlled Actions in the

    Herman Bruyninckx and Joris De Schutter. “Specification of Force-Controlled Actions in the "Task Frame Formalism" - a Synthesis”. In:IEEE Transactions on Robotics and Automation 12.4 (1996), pp. 581–589

    work page 1996

  13. [13]

    Kinematic Models for Model- Based Compliant Motion in the Presence of Uncertainty

    Herman Bruyninckx et al. “Kinematic Models for Model- Based Compliant Motion in the Presence of Uncertainty”. In:International Journal of Robotics Research14.5 (1995), pp. 465–482

    work page 1995

  14. [14]

    Bi-ACT: Bilateral Control-Based Imitation Learning via Action Chunking with Transformer

    Thanpimon Buamanee et al. “Bi-ACT: Bilateral Control-Based Imitation Learning via Action Chunking with Transformer”. In:IEEE International Conference on Advanced Intelligent Mechatronics. IEEE. 2024, pp. 410–415

    work page 2024

  15. [15]

    Learning Variable Impedance Control

    Jonas Buchli et al. “Learning Variable Impedance Control”. In:International Journal of Robotics Research30.7 (2011), pp. 820–833

    work page 2011

  16. [16]

    Integration of Haptics and Vision in Human Multisensory Grasping

    Ivan Camponogara and Robert V olcic. “Integration of Haptics and Vision in Human Multisensory Grasping”. In:Cortex135 (2021), pp. 173–185

    work page 2021

  17. [17]

    Vegetable Peeling: A Case Study in Constrained Dexterous Manipulation

    Tao Chen et al. “Vegetable Peeling: A Case Study in Constrained Dexterous Manipulation”. In:arXiv preprint arXiv:2407.07884(2024)

    work page arXiv 2024

  18. [18]

    Diffusion Policy: Visuomotor Policy Learn- ing via Action Diffusion

    Cheng Chi et al. “Diffusion Policy: Visuomotor Policy Learn- ing via Action Diffusion”. In:Robotics: Science and Systems. 2023

    work page 2023

  19. [19]

    Universal Manipulation Interface: In- The-Wild Robot Teaching Without In-The-Wild Robots

    Cheng Chi et al. “Universal Manipulation Interface: In- The-Wild Robot Teaching Without In-The-Wild Robots”. In: Robotics: Science and Systems. 2024

    work page 2024

  20. [20]

    The Parallel Ap- proach to Force/Position Control of Robotic Manipulators

    Stefano Chiaverini and Lorenzo Sciavicco. “The Parallel Ap- proach to Force/Position Control of Robotic Manipulators”. In: IEEE Transactions on Robotics and Automation9.4 (2002), pp. 361–373

    work page 2002

  21. [21]

    Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation

    Kyunghyun Cho et al. “Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation”. In:Empirical Methods in Natural Language Processing. ACL, 2014, pp. 1724–1734

    work page 2014

  22. [22]

    In-the-wild compliant manipulation with umi-ft.arXiv preprint arXiv:2601.09988, 2026

    Hojung Choi et al. “In-the-Wild Compliant Manipulation with UMI-FT”. In:arXiv preprint arXiv:2601.09988(2026)

    work page arXiv 2026

  23. [23]

    Sim-to-Real Transfer in Robotics: Addressing the Gap Between Simulation and Real-World Performance

    Naomi Chukwurah, Abiodun Sunday Adebayo, and Olanre- waju Oluwaseun Ajayi. “Sim-to-Real Transfer in Robotics: Addressing the Gap Between Simulation and Real-World Performance”. In:International Journal of Robotics and Sim- ulation6.1 (2024), pp. 89–102

    work page 2024

  24. [24]

    Open X- Embodiment: Robotic Learning Datasets and RT-X Models

    Open X-Embodiment Collaboration et al. “Open X- Embodiment: Robotic Learning Datasets and RT-X Models”. In:IEEE International Conference on Robotics and Automa- tion. 2024, pp. 6892–6903

    work page 2024

  25. [25]

    Learning Task Con- straints from Demonstration for Hybrid Force/Position Con- trol

    Adam Conkey and Tucker Hermans. “Learning Task Con- straints from Demonstration for Hybrid Force/Position Con- trol”. In:IEEE-RAS International Conference on Humanoid Robots. IEEE. 2019, pp. 162–169

    work page 2019

  26. [26]

    Compliant Robot Motion I. A Formalism for Specifying Compliant Motion Tasks

    Joris De Schutter and Hendrik Van Brussel. “Compliant Robot Motion I. A Formalism for Specifying Compliant Motion Tasks”. In:International Journal of Robotics Research7.4 (1988), pp. 3–17

    work page 1988

  27. [27]

    Accessed January 2026

    Google DeepMind.Gemini 3 Pro. Accessed January 2026

    work page 2026

  28. [28]

    RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot

    Hao-Shu Fang et al. “RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot”. In:IEEE International Conference on Robotics and Automation. 2024, pp. 653–660

    work page 2024

  29. [29]

    DEXOP: A Device for Robotic Trans- fer of Dexterous Human Manipulation

    Hao-Shu Fang et al. “DEXOP: A Device for Robotic Trans- fer of Dexterous Human Manipulation”. In:arXiv preprint arXiv:2509.04441(2025)

    work page arXiv 2025

  30. [30]

    AirExo: Low-Cost Exoskeletons for Learning Whole-Arm Manipulation in the Wild

    Hongjie Fang et al. “AirExo: Low-Cost Exoskeletons for Learning Whole-Arm Manipulation in the Wild”. In:IEEE International Conference on Robotics and Automation. IEEE. 2024, pp. 15031–15038

    work page 2024

  31. [31]

    AirExo-2: Scaling up Generalizable Robotic Imitation Learning with Low-Cost Exoskeletons

    Hongjie Fang et al. “AirExo-2: Scaling up Generalizable Robotic Imitation Learning with Low-Cost Exoskeletons”. In:Conference on Robot Learning. V ol. 305. PMLR, 2025, pp. 198–220

    work page 2025

  32. [32]

    Play to the Score: Stage-Guided Dy- namic Multi-Sensory Fusion for Robotic Manipulation

    Ruoxuan Feng et al. “Play to the Score: Stage-Guided Dy- namic Multi-Sensory Fusion for Robotic Manipulation”. In: Conference on Robot Learning. 2024

    work page 2024

  33. [33]

    Cortical Connections of Functional Zones in Posterior Pari- etal Cortex and Frontal Cortex Motor Regions in New World Monkeys

    Omar A Gharbawie, Iwona Stepniewska, and Jon H Kaas. “Cortical Connections of Functional Zones in Posterior Pari- etal Cortex and Frontal Cortex Motor Regions in New World Monkeys”. In:Cerebral Cortex21.9 (2011), pp. 1981–2002

    work page 2011

  34. [34]

    Functional Characterization of the Fronto-Parietal Reaching and Grasping Network: Reversible Deactivation of M1 and Areas 2, 5, and 7b in Awake Behav- ing Monkeys

    Adam B Goldring et al. “Functional Characterization of the Fronto-Parietal Reaching and Grasping Network: Reversible Deactivation of M1 and Areas 2, 5, and 7b in Awake Behav- ing Monkeys”. In:Journal of Neurophysiology127.5 (2022), pp. 1363–1387

    work page 2022

  35. [35]

    Deep Residual Learning for Image Recog- nition

    Kaiming He et al. “Deep Residual Learning for Image Recog- nition”. In:IEEE Conference on Computer Vision and Pattern Recognition. 2016, pp. 770–778

    work page 2016

  36. [36]

    FoAR: Force-Aware Reactive Policy for Contact-Rich Robotic Manipulation

    Zihao He et al. “FoAR: Force-Aware Reactive Policy for Contact-Rich Robotic Manipulation”. In:IEEE Robotics and Automation Letters(2025)

    work page 2025

  37. [37]

    Impedance Control: An Approach to Manip- ulation

    Neville Hogan. “Impedance Control: An Approach to Manip- ulation”. In:Journal of Dynamic Systems, Measurement, and Control107 (1985), pp. 1–24

    work page 1985

  38. [38]

    Adaptive Compliance Policy: Learning Approximate Compliance for Diffusion Guided Control

    Yifan Hou et al. “Adaptive Compliance Policy: Learning Approximate Compliance for Diffusion Guided Control”. In: IEEE International Conference on Robotics and Automation. 2025, pp. 4829–4836

    work page 2025

  39. [39]

    3d-vitac: Learning fine-grained ma- nipulation with visuo-tactile sensing.arXiv preprint arXiv:2410.24091, 2024

    Binghao Huang et al. “3D-ViTac: Learning Fine-Grained Manipulation with Visuo-Tactile Sensing”. In:arXiv preprint arXiv:2410.24091(2024)

    work page arXiv 2024

  40. [40]

    DTW-Align: Bridging the Modal- ity Gap in End-to-End Speech Translation with Dynamic Time Warping Alignment

    Abderrahmane Issam et al. “DTW-Align: Bridging the Modal- ity Gap in End-to-End Speech Translation with Dynamic Time Warping Alignment”. In:Proceedings of the Tenth Conference on Machine Translation. 2025, pp. 191–199

    work page 2025

  41. [41]

    Adaptive mixtures of local experts

    Robert A Jacobs et al. “Adaptive mixtures of local experts”. In:Neural computation3.1 (1991), pp. 79–87

    work page 1991

  42. [42]

    A Survey of Automated Threaded Fastening

    Zhenzhong Jia et al. “A Survey of Automated Threaded Fastening”. In:IEEE Transactions on Automation Science and Engineering16.1 (2018), pp. 298–310

    work page 2018

  43. [43]

    Learning Force Control Poli- cies for Compliant Manipulation

    Mrinal Kalakrishnan et al. “Learning Force Control Poli- cies for Compliant Manipulation”. In:IEEE/RSJ International Conference on Intelligent Robots and Systems. 2011, pp. 4639– 4644

    work page 2011

  44. [44]

    Learning Variable Compliance Control from a Few Demonstrations for Bimanual Robot with Haptic Feed- back Teleoperation System

    Tatsuya Kamijo, Cristian C Beltran-Hernandez, and Masashi Hamaya. “Learning Variable Compliance Control from a Few Demonstrations for Bimanual Robot with Haptic Feed- back Teleoperation System”. In:IEEE/RSJ International Con- ference on Intelligent Robots and Systems. IEEE. 2024, pp. 12663–12670

    work page 2024

  45. [45]

    Admittance control for physical human–robot inter- action

    Arvid QL Keemink, Herman Van der Kooij, and Arno HA Stienen. “Admittance control for physical human–robot inter- action”. In:The International Journal of Robotics Research 37.11 (2018), pp. 1421–1444

    work page 2018

  46. [46]

    A Unified Approach for Motion and Force Control of Robot Manipulators: The Operational Space For- mulation

    Oussama Khatib. “A Unified Approach for Motion and Force Control of Robot Manipulators: The Operational Space For- mulation”. In:IEEE Journal on Robotics and Automation3.1 (2003), pp. 43–53

    work page 2003

  47. [47]

    Learning Movement Primitives for Force Interaction Tasks

    Jens Kober, Michael Gienger, and Jochen J Steil. “Learning Movement Primitives for Force Interaction Tasks”. In:IEEE International Conference on Robotics and Automation. IEEE. 2015, pp. 3192–3199

    work page 2015

  48. [48]

    Stability Considerations for Variable Impedance Control

    Klas Kronander and Aude Billard. “Stability Considerations for Variable Impedance Control”. In:IEEE Transactions on Robotics32.5 (2016), pp. 1298–1305

    work page 2016

  49. [49]

    Learn- ing Contact-Rich Manipulation Skills With Guided Policy Search

    Sergey Levine, Nolan Wagener, and Pieter Abbeel. “Learn- ing Contact-Rich Manipulation Skills With Guided Policy Search”. In:IEEE International Conference on Robotics and Automation. 2015, pp. 156–163

    work page 2015

  50. [50]

    See, Hear, and Feel: Smart Sensory Fusion for Robotic Manipulation

    Hao Li et al. “See, Hear, and Feel: Smart Sensory Fusion for Robotic Manipulation”. In:Conference on Robot Learning. 2022, pp. 1368–1378

    work page 2022

  51. [51]

    Augmentation Enables One-Shot Generalization in Learning from Demon- stration for Contact-Rich Manipulation

    Xing Li, Manuel Baum, and Oliver Brock. “Augmentation Enables One-Shot Generalization in Learning from Demon- stration for Contact-Rich Manipulation”. In:IEEE/RSJ Inter- national Conference on Intelligent Robots and Systems. IEEE. 2023, pp. 3656–3663

    work page 2023

  52. [52]

    Learning from Demonstration Based on Environmental Constraints

    Xing Li and Oliver Brock. “Learning from Demonstration Based on Environmental Constraints”. In:IEEE Robotics and Automation Letters7.4 (2022), pp. 10938–10945

    work page 2022

  53. [53]

    ForceMimic: Force-Centric Imitation Learning With Force-Motion Capture System for Contact-Rich Manipulation

    Wenhai Liu et al. “ForceMimic: Force-Centric Imitation Learning With Force-Motion Capture System for Contact-Rich Manipulation”. In:ICRA. IEEE. 2025, pp. 1105–1112

    work page 2025

  54. [54]

    ImMimic: Cross-Domain Imitation from Human Videos via Mapping and Interpolation

    Yangcen Liu et al. “ImMimic: Cross-Domain Imitation from Human Videos via Mapping and Interpolation”. In:arXiv preprint arXiv:2509.10952(2025)

    work page arXiv 2025

  55. [55]

    ManiW A V: Learning Robot Manipulation from In-the-Wild Audio-Visual Data

    Zeyi Liu et al. “ManiW A V: Learning Robot Manipulation from In-the-Wild Audio-Visual Data”. In:Conference on Robot Learning. 2024

    work page 2024

  56. [56]

    Accessed January 2026

    Flexiv Ltd.Flexiv Teleoperation Development Kit (TDK). Accessed January 2026

    work page 2026

  57. [57]

    Compliance and Force Control for Com- puter Controlled Manipulators

    Matthew T Mason. “Compliance and Force Control for Com- puter Controlled Manipulators”. In:IEEE Transactions on Systems, Man, and Cybernetics11.6 (2007), pp. 418–432

    work page 2007

  58. [58]

    A Generic Task Model and Control Strategy to Support Learning, Robust Control, and Generalization of Contact-Rich Manipulation Tasks

    Ali Mousavi Mohammadi et al. “A Generic Task Model and Control Strategy to Support Learning, Robust Control, and Generalization of Contact-Rich Manipulation Tasks”. In: Robotics and Autonomous Systems197 (2026), p. 105270

    work page 2026

  59. [59]

    Dynamic Time Warping

    Meinard Müller. “Dynamic Time Warping”. In:Information Retrieval for Music and Motion. Springer, 2007, pp. 69–84

    work page 2007

  60. [60]

    Performance Trade- offs in Dynamic Time Warping Algorithms for Isolated Word Recognition

    C. Myers, L. Rabiner, and A. Rosenberg. “Performance Trade- offs in Dynamic Time Warping Algorithms for Isolated Word Recognition”. In:IEEE Transactions on Acoustics, Speech, and Signal Processing28.6 (1980), pp. 623–635

    work page 1980

  61. [61]

    FORGE: Force-Guided Explo- ration for Robust Contact-Rich Manipulation Under Uncer- tainty

    Michael Noseworthy et al. “FORGE: Force-Guided Explo- ration for Robust Contact-Rich Manipulation Under Uncer- tainty”. In:IEEE Robotics and Automation Letters(2025)

    work page 2025

  62. [62]

    Identifying Physical In- teractions in Contact-Based Robot Manipulation for Learning from Demonstration

    Alex Harm Gert-Jan Overbeek et al. “Identifying Physical In- teractions in Contact-Based Robot Manipulation for Learning from Demonstration”. In:Advanced Robotics Research(2025), e202500109

    work page 2025

  63. [63]

    Much ado about noising: Dispelling the myths of generative robotic control.arXiv preprint arXiv:2512.01809,

    Chaoyi Pan et al. “Much Ado About Noising: Dispelling the Myths of Generative Robotic Control”. In:arXiv preprint arXiv:2512.01809(2025)

    work page arXiv 2025

  64. [64]

    FiLM: Visual Reasoning with a General Conditioning Layer

    Ethan Perez et al. “FiLM: Visual Reasoning with a General Conditioning Layer”. In:AAAI conference on Artificial Intel- ligence. 2018, pp. 3942–3951

    work page 2018

  65. [65]

    C-Learn: Learn- ing Geometric Constraints from Demonstrations for Multi- Step Manipulation in Shared Autonomy

    Claudia Pérez-D’Arpino and Julie A Shah. “C-Learn: Learn- ing Geometric Constraints from Demonstrations for Multi- Step Manipulation in Shared Autonomy”. In:IEEE Interna- tional Conference on Robotics and Automation. IEEE. 2017, pp. 4058–4065

    work page 2017

  66. [66]

    Hybrid Position/Force Control of Manipulators

    Marc H Raibert and John J Craig. “Hybrid Position/Force Control of Manipulators”. In:Journal of dynamic systems, measurement, and control103.2 (1981), pp. 126–133

    work page 1981

  67. [67]

    Eyesight Hand: Design of a Fully- Actuated Dexterous Robot Hand with Integrated Vision-based Tactile Sensors and Compliant Actuation

    Branden Romero et al. “Eyesight Hand: Design of a Fully- Actuated Dexterous Robot Hand with Integrated Vision-based Tactile Sensors and Compliant Actuation”. In:arXiv preprint arXiv:2408.06265(2024)

    work page arXiv 2024

  68. [68]

    Dynamic Programming Algorithm Optimization for Spoken Word Recognition

    H. Sakoe and S. Chiba. “Dynamic Programming Algorithm Optimization for Spoken Word Recognition”. In:IEEE Trans- actions on Acoustics, Speech, and Signal Processing26.1 (1978), pp. 43–49

    work page 1978

  69. [69]

    Prometheus: Universal, Open-Source Mocap-Based Teleoperation System with Force Feedback for Dataset Collection in Robot Learning

    Sergei Satsevich et al. “Prometheus: Universal, Open-Source Mocap-Based Teleoperation System with Force Feedback for Dataset Collection in Robot Learning”. In:arXiv preprint arXiv:2510.01023(2025)

    work page arXiv 2025

  70. [70]

    EquiContact: A Hierarchical SE(3) Vision-To-Force Equivariant Policy for Spatially Generalizable Contact-Rich Tasks

    Joohwan Seo et al. “EquiContact: A Hierarchical SE(3) Vision-To-Force Equivariant Policy for Spatially Generalizable Contact-Rich Tasks”. In:arXiv preprint arXiv:2507.10961 (2025)

    work page arXiv 2025

  71. [71]

    Adaptive Admittance Control: An Ap- proach to Explicit Force Control in Compliant Motion

    Homayoun Seraji. “Adaptive Admittance Control: An Ap- proach to Explicit Force Control in Compliant Motion”. In: IEEE International Conference on Robotics and Automation. IEEE. 1994, pp. 2705–2712

    work page 1994

  72. [72]

    In- ferring Geometric Constraints in Human Demonstrations

    Guru Subramani, Michael Zinn, and Michael Gleicher. “In- ferring Geometric Constraints in Human Demonstrations”. In: Conference on Robot Learning. PMLR. 2018, pp. 223–236

    work page 2018

  73. [73]

    Imitation Learning-Based Framework for Learning 6-D Lin- ear Compliant Motions

    Markku Suomalainen, Fares J Abu-Dakka, and Ville Kyrki. “Imitation Learning-Based Framework for Learning 6-D Lin- ear Compliant Motions”. In:Autonomous Robots45.3 (2021), pp. 389–405

    work page 2021

  74. [74]

    A Survey of Robot Manipulation in Contact

    Markku Suomalainen, Yiannis Karayiannidis, and Ville Kyrki. “A Survey of Robot Manipulation in Contact”. In:Robotics and Autonomous Systems156 (2022), p. 104224

    work page 2022

  75. [75]

    Learning Compliant Assembly Motions from Demonstration

    Markku Suomalainen and Ville Kyrki. “Learning Compliant Assembly Motions from Demonstration”. In:IEEE/RSJ Inter- national Conference on Intelligent Robots and Systems. IEEE. 2016, pp. 871–876

    work page 2016

  76. [76]

    A Geometric Ap- proach for Learning Compliant Motions from Demonstra- tion

    Markku Suomalainen and Ville Kyrki. “A Geometric Ap- proach for Learning Compliant Motions from Demonstra- tion”. In:IEEE-RAS International Conference on Humanoid Robotics. IEEE. 2017, pp. 783–790

    work page 2017

  77. [77]

    Learning from Demonstration for Hydraulic Manipulators

    Markku Suomalainen et al. “Learning from Demonstration for Hydraulic Manipulators”. In:IEEE/RSJ International Confer- ence on Intelligent Robots and Systems. IEEE. 2018, pp. 3579– 3586

    work page 2018

  78. [78]

    Partially Decoupled Impedance Motion Force Control Using Prioritized Inertia Shaping

    Wenbo Tang, Weiming Wang, and Shiquan Wang. “Partially Decoupled Impedance Motion Force Control Using Prioritized Inertia Shaping”. In:IEEE Robotics and Automation Letters (2024)

    work page 2024

  79. [79]

    Task Parameterization Us- ing Continuous Constraints Extracted from Human Demon- strations

    Ana Lucia Pais Ureche et al. “Task Parameterization Us- ing Continuous Constraints Extracted from Human Demon- strations”. In:IEEE Transactions on Robotics31.6 (2015), pp. 1458–1471

    work page 2015

  80. [80]

    Neural Control of Motion-to-Force Transitions With the Fin- gertip

    Madhusudhan Venkadesan and Francisco J Valero-Cuevas. “Neural Control of Motion-to-Force Transitions With the Fin- gertip”. In:Journal of Neuroscience28.6 (2008), pp. 1366– 1373

    work page 2008

Showing first 80 references.