Pith. sign in

REVIEW 5 major objections 5 minor 2 cited by

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

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read FoAR claims that gating force/torque feedback by a learned future-contact probability lets a vision-based imitation policy perform contact-rich manipulation with simple position control, beating vision-only and naive-fusion baselines on…

desk verdict A clean, effective recipe for force-aware contact-rich manipulation that deserves review but needs better reporting on the Chopping baselines and error bars. read the letter →

arxiv 2411.15753 v2 pith:ZHQSNJOH submitted 2024-11-24 cs.RO

Zihao He , Hongjie Fang , Jingjing Chen , Hao-Shu Fang , Cewu Lu This is my paper · ORCID
classification cs.RO
keywords force-awarepolicycontact-richmanipulationfuturecontactpredictorforce/torquesensingmultimodalfeaturefusionreactivecontrolimitationlearningrobot
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

FoAR tries to establish that the missing ingredient for contact-rich manipulation is not a more complex force controller but a way to let force/torque data speak only when contact is actually relevant. The paper combines a point-cloud visuomotor policy with high-frequency force/torque history and a learned future-contact probability that gates the fusion of the two modalities. The same probability also triggers a reactive control step at deployment, so a robot can compensate for insufficient contact force using plain position control. On wiping, wiping with arbitrary board orientation, peeling, and chopping, FoAR reports higher scores than vision-only and naive-fusion baselines, with 100% action success on the first three tasks.

What carries the argument

The load-bearing mechanism is the future contact predictor and the gated fusion it controls. The predictor is a small network that takes the current RGB image and the recent force/torque stream and outputs $\phi(t)$, trained with binary cross-entropy against contact labels extracted from demonstrations using the same static wrench thresholds used at deployment. Its output multiplies the force feature before concatenation with the scene feature, and at inference it decides whether the current time step belongs to the contact or non-contact temporal ensemble buffer. The reactive control law then closes the loop: if the policy predicts contact but the sensor reads too little force, it nudges the whole predicted action chunk a fixed distance toward its own average direction. This is the mechanism that lets a position-controlled end-effector perform sustained-contact skills.

What would settle it

Sweep $\epsilon$ from 0 to 20 mm and $\delta_f$ from 2 to 20 N on the wiping and peeling tasks, or run FoAR unchanged on objects of different stiffness and curvature using the same thresholds; if performance has no plateau, or the optimal values move with the object, then the reactive control is task-tuned rather than a general contact-rich manipulation mechanism.

Watch

Extended reading notes

Core claim

The central discovery is that contact-rich manipulation can be treated as a phase-gating problem rather than a force-control problem. FoAR represents the current scene as a sparse point-cloud feature $h^s_t$, encodes the last two seconds of 100 Hz wrench data into a force feature $h^f_t$, and learns a future contact probability $\phi(t)\in[0,1]$ from the current RGB image and wrench history. The fused conditioning feature is $h_t = [h^s_t;\, \phi(t)h^f_t + (1-\phi(t))h^*]$, where $h^*$ is a learnable neutral embedding, so wrench information is emphasized exactly when contact is coming and blended out during free motion. At deployment, when $\phi$ exceeds 0.9 but the measured wrench is below 8 N force and 5 N·m torque, the controller shifts the predicted end-effector trajectory by $\epsilon=0.006$ m along the average predicted motion direction. The paper reports scores of 0.875, 0.850, and 0.756 on Wiping, Wiping(General), and Peeling versus 0.500, 0.500, and 0.377 for the strongest vision-only baseline, and chopping results of 3.9 versus 1.8 segments with lower length dispersion.

Load-bearing premise

The load-bearing premise is that a fixed manual rule — if predicted contact exceeds 0.9 but measured wrench stays below 8 N and 5 N·m, nudge the end-effector 6 mm along the predicted motion direction — is the right corrective behavior across objects and contact geometries, rather than a heuristic calibrated to the three tested tasks.

Editorial extensions

If this is right

  • If FoAR's results hold, contact-rich imitation policies do not need compliant or hybrid force controllers; a reactive position-level nudge driven by a contact predictor is enough for the evaluated skills.
  • Gating by predicted future contact should remove the need to choose a single fusion strategy for an entire task, since force information is suppressed during phases where it is mostly noise.
  • The same predictor can serve two roles — feature weighting and control triggering — so the policy and the reactive controller are co-designed rather than tuned separately.
  • With 50 demonstrations per task, the method reaches 100% grasp and contact-action success on wiping and peeling, suggesting the force gate adds capability without requiring more data.
  • Because the system kept its score when the whiteboard was rewritten or moved mid-task, the reactive mechanism appears to absorb unexpected state changes rather than needing replanning.

Reading between the lines

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

  • The hand-set thresholds ($\delta_\phi=0.9$, 8 N, 5 N·m, $\epsilon=0.006$ m) are not learned, and the paper itself notes static thresholds may struggle in complex environments; a natural extension is to make the nudge size or wrench threshold depend on predicted contact state or object stiffness.
  • The same gated-fusion idea should transfer to other noisy-but-informative modalities such as tactile or audio, where the signal-to-noise ratio also switches with contact phase; the paper only demonstrates it for force/torque.
  • A testable extension is to measure how much of the gain comes from the 100 Hz sampling rate by retraining FoAR with downsampled wrench histories; the paper's ablation table suggests the advantage degrades sharply at lower rates.
  • Because the reactive correction only uses the average predicted direction, it should also work when the action predictor is replaced by a different one, as long as a contact probability and wrench readings are available; FoAR's own experiments do not test this portability.
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

5 major / 5 minor

Summary. FoAR is a force-aware reactive policy for contact-rich manipulation built on top of the RISE visuomotor policy. It uses a learned future contact predictor to gate the fusion of high-frequency force/torque features with visual scene features, and it adds a reactive control step at deployment: when the predicted contact probability is high but the measured wrench is low, the end-effector action is nudged by a fixed step along the predicted action direction. The paper reports real-robot experiments on Wiping, Wiping (General), Peeling, and Chopping, plus robustness evaluations under dynamic disturbances, claiming that FoAR significantly outperforms all baselines.

Significance. If the empirical claims hold, FoAR would be a practically valuable contribution: it shows that force/torque feedback can be integrated into a position-controlled imitation policy through a learned contact gate plus a simple reactive nudge, and it demonstrates large margins over the RISE baseline on several contact-rich tasks. The paper includes real-robot evaluations, ablations of the predictor and of sensing frequency, and robustness tests, which are useful for the community. However, the strength of the current claims exceeds the evidence: the Chopping task omits most baselines, no error bars or significance tests are reported for the main tables, the predictor labels and reactive thresholds are manually chosen without sensitivity analysis, and one ablation table is inconsistent with the main results. These issues need to be addressed before the paper can be recommended for publication.

major comments (5)
  1. [§IV-C, Table II] The abstract and Section IV claim that FoAR 'significantly outperforms all baselines across various challenging contact-rich tasks,' but the Chopping task evaluates only FoAR and RISE, as stated in §IV-A ('10 trials each only for FoAR and RISE on the Chopping task to conserve resources'). ACT, Diffusion Policy, and the three RISE ablation variants are missing from Table II. The universal 'all baselines' claim is therefore not supported for the instantaneous-force-impact task. Please either evaluate all baselines on Chopping or explicitly qualify the claim to the tasks where all baselines were run.
  2. [Tables I, II, and IV] The term 'significantly outperforms' is used throughout, but Tables I and IV report no error bars, confidence intervals, or significance tests, and trial counts are 20 per method (10 for Chopping). With these sample sizes and with hand-assigned scores (e.g., 1/0.5/0 for wiping), the reported margins cannot be assessed for statistical reliability. Please report per-trial results, confidence intervals, and a significance test where appropriate, or replace 'significantly' with a weaker qualitative claim.
  3. [§III-B Supervision, Appendix A] The ground-truth labels for the 'future contact predictor' are described in Appendix A as generated from force/torque data 'within the time window [t − 2s, t + 2s]', which is a symmetric window that includes past contact. This makes the predictor a smoothed contact detector rather than a strictly future predictor, and it blurs the claimed distinction from the 'contact detection' variant discussed in §IV-D and Table III. Please specify the exact temporal window and, if future prediction is intended, use labels from a future-only window; otherwise rename the quantity and adjust the interpretation of the ablation.
  4. [§IV-A Implementation, Appendix A, Algorithm 1] The reactive control parameters (δφ = 0.9, δf = 8 N, δt = 5 N·m, ε = 0.006 m) and the predictor supervision threshold are all manually chosen, and no sensitivity analysis is reported. Since the predictor labels and the reactive trigger share the same threshold family, the reported gains could depend on these hand-tuned values rather than on the learned force-awareness. This is a load-bearing generalization concern: please add a sensitivity study over thresholds and ε, or demonstrate that the method works with thresholds set by a principled, task-independent rule.
  5. [Table V vs. Table I] The Peeling ablation in Table V reports RISE score 0.293 and FoAR score 0.588, whereas Table I reports RISE 0.377 and FoAR 0.756 for the same task and protocol. The discrepancy is not explained; if these numbers come from different seeds, evaluation splits, or a different version of the method, that must be stated. As printed, the two tables are inconsistent and prevent the reader from trusting the ablation comparison.
minor comments (5)
  1. [Table III] The checkmarks in Table III are not aligned with the column headers ('w. Predictor', 'w. Reactive'), making it impossible to tell which rows include which components. Please reformat the table so each row clearly indicates the presence or absence of the predictor and reactive control.
  2. [§IV-A] The sentence 'we run 20 trials per method for the Wiping and Peeling tasks, and 10 trials each only for FoAR and RISE on the Chopping task' is unclear because 'each only for FoAR and RISE' could be read as applying to the whole sentence. Please rephrase to state explicitly which methods are evaluated on which tasks.
  3. [Abstract, §IV-A] The abstract says 'With only 50 demonstrations per task,' but the Chopping task uses 40 demonstrations. Please adjust the wording to reflect the actual per-task demonstration counts.
  4. [Table IV] In Table IV, the RISE baseline achieves a higher score under the 'Move' condition (0.600) than under 'Original' (0.500), which is counterintuitive for a disturbance evaluation. A sentence explaining this result would help the reader interpret the robustness comparison.
  5. [§III-C, Algorithm 1] The reactive control correction in Lines 12-14 adds a fixed 6 mm displacement to all predicted position actions without any validation of the step size ε or the averaging horizon Tf = 5. Please add a brief discussion or a small ablation addressing the sensitivity to these values.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FoAR's reported gains are measured against external task outcomes and are not forced by construction.

full rationale

FoAR's central claim is an empirical comparison on externally scored tasks (whiteboard erasure, cucumber peel proportion, pepper segment uniformity), so the headline result cannot reduce to its own inputs by construction. The only step that superficially resembles co-adaptation is the future contact predictor: its ground-truth labels are extracted from force/torque thresholds (Sec. III-B, Appendix A), and the reactive controller triggers on force/torque thresholds δf=8 N, δt=5 N·m (Sec. IV-A). Even if the same thresholds are used, the predictor is a learned function of RGB and force/torque, not a fitted parameter renamed as a prediction; the reactive correction is a deploy-time heuristic whose effect is measured against external task scores. No equation in the paper sets the output equal to an input, and no load-bearing claim is justified by a self-citation chain. The Chopping evaluation omits most baselines (Sec. IV-C), and the static-threshold limitation is acknowledged in Sec. V, but both are evidence-coverage concerns, not circularity. Score 0.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The central design rests on several manually chosen constants: two wrench thresholds, a contact-probability threshold, a nudge step size, a loss weight, and the history/action horizons. The same thresholds generate the predictor's supervision and trigger the reactive control, so the ledger is dominated by task-specific calibration rather than learned or externally anchored quantities.

free parameters (8)
  • future contact threshold delta_phi = 0.9
    Threshold on predicted contact probability to switch between contact and non-contact buffers and to trigger reactive control (Section III.C).
  • force threshold delta_f = 8 N
    Used both to generate ground-truth contact labels and in reactive control to decide if contact force is insufficient (Section III.C, Appendix A).
  • torque threshold delta_t = 5 N*m
    Same dual role as delta_f.
  • reactive step epsilon = 0.006 m
    Fixed position correction magnitude applied along predicted displacement direction when insufficient wrench is detected (Algorithm 1).
  • loss weight alpha = 0.1
    Balances action loss and contact predictor loss (Section III.B).
  • history horizon To = 200 steps (~2s at 100Hz)
    Length of force/torque history encoded by the policy.
  • action horizon Ta and direction horizon Tf = 20 / 5 steps
    Ta is the predicted action chunk; Tf is the number of future action steps used to estimate direction in reactive control.
  • contact label window = [t-2s, t+2s]
    Whether a timestep counts as 'future contact' depends on force/torque exceeding thresholds anywhere in this window (Appendix A).
assumptions (4)
  • domain assumption Force/torque readings above task-specific thresholds indicate actual contact, and these thresholds are reliable across the evaluation trials.
    The entire supervision signal and reactive trigger depend on this equivalence (Appendix A).
  • ad hoc to paper A fixed 6 mm nudge in the predicted action direction is sufficient to establish or improve contact without causing instability or overshoot.
    Algorithm 1 Lines 12-14; no stability analysis or systematic sweep of epsilon is provided.
  • domain assumption The future contact predictor, trained on labels derived from the same thresholds used at deployment, generalizes to novel object placements and dynamic disturbances.
    Robustness experiments assume the predictor remains valid under whiteboard moves and rewritten figures (Section IV.E).
  • domain assumption RISE's pretrained or specified components and training setup are adequate backbones; FoAR inherits RISE's generalization.
    The paper builds directly on RISE [44] and inherits its hyperparameters; if RISE's assumptions fail, FoAR inherits the failure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FoAR: Force-Aware Reactive Policy for Contact-Rich Robotic Manipulation." pith.science (2026). https://pith.science/paper/ZHQSNJOH

@misc{pith2026241115753,
  author       = {Pith},
  title        = {Pith review of: FoAR: Force-Aware Reactive Policy for Contact-Rich Robotic Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZHQSNJOH}},
  note         = {Machine review of arXiv:2411.15753}
}
read the original abstract

Contact-rich tasks present significant challenges for robotic manipulation policies due to the complex dynamics of contact and the need for precise control. Vision-based policies often struggle with the skill required for such tasks, as they typically lack critical contact feedback modalities like force/torque information. To address this issue, we propose FoAR, a force-aware reactive policy that combines high-frequency force/torque sensing with visual inputs to enhance the performance in contact-rich manipulation. Built upon the RISE policy, FoAR incorporates a multimodal feature fusion mechanism guided by a future contact predictor, enabling dynamic adjustment of force/torque data usage between non-contact and contact phases. Its reactive control strategy also allows FoAR to accomplish contact-rich tasks accurately through simple position control. Experimental results demonstrate that FoAR significantly outperforms all baselines across various challenging contact-rich tasks while maintaining robust performance under unexpected dynamic disturbances. Project website: https://tonyfang.net/FoAR/

Figures

Figures reproduced from arXiv: 2411.15753 by the authors.

Figure 1
Figure 1. Overview of the FoAR Policy for Contact-Rich Robotic [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. FoAR Architecture. FoAR consists of a point cloud encoder [44], a force/torque encoder, a future contact predictor, and a diffusion action head [10]. The scene features and force features are fused under the guidance of the future contact predictor. Algorithm 1 FoAR Inference with Reactive Control 1: buffer.clear(); 2: contact buffer.clear(); ▷ clear the temporal ensemble buffer. 3: for timestep t ← 0 to Nmax − 1 do… view at source ↗
Figure 3
Figure 3. Tasks. We carefully design 3 challenging contact-rich tasks that focus on different aspects of the contact-rich manipulations. These tasks involve both non-contact phases and contact phases to evaluate the policy performance thoroughly. control [31, 38]. These approaches often require additional parameters, such as stiffness and contact force direction. In contrast, we demonstrate that our proposed future con￾tact p… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative Results of the Peeling Task. The green background indicates the robot fully peels the cucumber, while yellow and red backgrounds repre￾sent partial peeling and peeling failure, respectively. the RISE transformer, akin to [9, 30, 32, 37]; (2) RISE (force-con…
Figure 5
Figure 5. Figure 5: Evaluation Metrics of the Chopping Task. We encourage the robot to divide the pepper into several uniform small segments, without segments sticking together due to partial cuts. Method # Segments ↑ Norm. Length ASR (%) ↑ Avg. ↓ Std. ↓ Grasp Place RISE [44] 1.8 ± 0.6 0.…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. TA-VLA: Elucidating the Design Space of Torque-aware Vision-Language-Action Models

    cs.RO 2025-09 conditional novelty 6.0 of 10

    Feeding torque history as a single decoder token and adding torque prediction as an auxiliary objective improves pretrained VLA success rates on contact-rich manipulation, with large gains on button pushing and charge...

  2. Adaptive Visuo-Tactile Fusion with Predictive Force Attention for Dexterous Manipulation

    cs.RO 2025-05 conditional novelty 6.0 of 10

    A force-guided attention module and future-force prediction auxiliary task improve visuo-tactile fusion for dexterous manipulation, reaching 93% average success in real robot trials.

Reference graph

Works this paper leans on

55 extracted references · 39 canonical work pages · cited by 2 Pith papers

  1. [1]

    Learning Diffusion Policies from Demonstra- tions For Compliant Contact-rich Manipulation

    Malek Aburub et al. “Learning Diffusion Policies from Demonstra- tions For Compliant Contact-rich Manipulation”. In: arXiv preprint arXiv:2410.19235 (2024)

    arXiv 2024

  2. [2]

    Learning Force Control for Contact-Rich Manipulation Tasks With Rigid Position-Controlled Robots

    Cristian Camilo Beltran-Hernandez et al. “Learning Force Control for Contact-Rich Manipulation Tasks With Rigid Position-Controlled Robots”. In: IEEE Robotics and Automation Letters 5.4 (2020), pp. 5709–5716

    work page 2020

  3. [3]

    Anyskin: Plug-and-play skin sensing for robotic touch

    Raunaq Bhirangi et al. “Anyskin: Plug-and-play skin sensing for robotic touch”. In: arXiv preprint arXiv:2409.08276 (2024)

    arXiv 2024

  4. [4]

    ReSkin: Versatile, Replaceable, Lasting Tactile Skins

    Raunaq M. Bhirangi et al. “ReSkin: Versatile, Replaceable, Lasting Tactile Skins”. In: Conference on Robot Learning . 2021, pp. 587– 597

    work page 2021

  5. [5]

    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

  6. [6]

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

    Thanpimon Buamanee et al. “Bi-ACT: Bilateral Control-Based Im- itation Learning via Action Chunking with Transformer”. In: arXiv preprint arXiv:2401.17698 (2024)

    arXiv 2024

  7. [7]

    Six-Axis Force/Torque Sensors for Robotics Applications: A Re- view

    Max Yiye Cao, Stephen Laws, and Ferdinando Rodriguez y Baena. “Six-Axis Force/Torque Sensors for Robotics Applications: A Re- view”. In: IEEE Sensors Journal 21.24 (2021), pp. 27238–27251

    work page 2021

  8. [8]

    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)

    arXiv 2024

Show all 55 references
  1. [9]

    Visuo-Tactile Transformers for Manipulation

    Yizhou Chen et al. “Visuo-Tactile Transformers for Manipulation”. In: 2022

    2022

  2. [10]

    Diffusion Policy: Visuomotor Policy Learning via Action Diffusion

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

    2023

  3. [11]

    4D Spatio- Temporal ConvNets: Minkowski Convolutional Neural Networks

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. “4D Spatio- Temporal ConvNets: Minkowski Convolutional Neural Networks”. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 2019, pp. 3075–3084

    2019

  4. [12]

    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 Interna- tional Conference on Robotics and Automation . 2024, pp. 6892– 6903

    2024

  5. [13]

    Tactile-based Insertion for Dense Box-Packing

    Siyuan Dong and Alberto Rodriguez. “Tactile-based Insertion for Dense Box-Packing”. In: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . 2019, pp. 7953–7960

    2019

  6. [14]

    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

    2024

  7. [15]

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

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

    2024

  8. [16]

    Deep Residual Learning for Image Recognition

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

    2016

  9. [17]

    FurnitureBench: Reproducible Real-World Bench- mark for Long-Horizon Complex Manipulation

    Minho Heo et al. “FurnitureBench: Reproducible Real-World Bench- mark for Long-Horizon Complex Manipulation”. In: Robotics: Sci- ence and Systems . 2023

    2023

  10. [18]

    Denoising Diffusion Probabilistic Models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. “Denoising Diffusion Probabilistic Models”. In: Advances in Neural Information Process- ing Systems 33 (2020), pp. 6840–6851

    2020

  11. [19]

    Impedance Control: An Approach to Manipulation

    Neville Hogan. “Impedance Control: An Approach to Manipulation”. In: Journal of Dynamic Systems, Measurement, and Control 107 (1985), pp. 1–24

    1985

  12. [20]

    Robust Execution of Contact- Rich Motion Plans by Hybrid Force-Velocity Control

    Yifan Hou and Matthew T Mason. “Robust Execution of Contact- Rich Motion Plans by Hybrid Force-Velocity Control”. In: IEEE In- ternational Conference on Robotics and Automation. 2019, pp. 1933– 1939

    2019

  13. [21]

    Adaptive Compliance Policy: Learning Approxi- mate Compliance for Diffusion Guided Control

    Yifan Hou et al. “Adaptive Compliance Policy: Learning Approxi- mate Compliance for Diffusion Guided Control”. In: arXiv preprint arXiv:2410.09309 (2024)

    2024 arXiv

  14. [22]

    3D-ViTac: Learning Fine-Grained Manipulation with Visuo-Tactile Sensing

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

    2024 arXiv

  15. [23]

    Learning Force Control Policies for Compliant Manipulation

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

    2011

  16. [24]

    Learning Variable Compliance Control from a Few Demonstrations for Bimanual Robot with Haptic Feedback Teleop- eration System

    Tatsuya Kamijo, Cristian C Beltran-Hernandez, and Masashi Hamaya. “Learning Variable Compliance Control from a Few Demonstrations for Bimanual Robot with Haptic Feedback Teleop- eration System”. In: arXiv preprint arXiv:2406.14990 (2024)

    2024 arXiv

  17. [25]

    OpenVLA: An Open-Source Vision-Language- Action Model

    Moo Jin Kim et al. “OpenVLA: An Open-Source Vision-Language- Action Model”. In: arXiv preprint arXiv:2406.09246 (2024)

    2024 arXiv

  18. [26]

    ALPHA-α and Bi-ACT Are All You Need: Importance of Position and Force Information/Control for Imitation Learning of Unimanual and Bimanual Robotic Manipulation with Low-Cost System

    Masato Kobayashi, Thanpimon Buamanee, and Takumi Kobayashi. “ALPHA-α and Bi-ACT Are All You Need: Importance of Position and Force Information/Control for Imitation Learning of Unimanual and Bimanual Robotic Manipulation with Low-Cost System”. In: arXiv preprint arXiv:2411.099...

    2024 arXiv

  19. [27]

    DIGIT: A Novel Design for a Low-Cost Compact High-Resolution Tactile Sensor With Application to In- Hand Manipulation

    Mike Lambeta et al. “DIGIT: A Novel Design for a Low-Cost Compact High-Resolution Tactile Sensor With Application to In- Hand Manipulation”. In: IEEE Robotics and Automation Letters 5.3 (2020), pp. 3838–3845

    2020

  20. [28]

    Making Sense of Vision and Touch: Self- Supervised Learning of Multimodal Representations for Contact- Rich Tasks

    Michelle A. Lee et al. “Making Sense of Vision and Touch: Self- Supervised Learning of Multimodal Representations for Contact- Rich Tasks”. In: IEEE International Conference on Robotics and Automation. IEEE, 2019, pp. 8943–8950

    2019

  21. [29]

    Learning contact-rich manipulation skills with guided policy search

    Sergey Levine, Nolan Wagener, and Pieter Abbeel. “Learning contact-rich manipulation skills with guided policy search”. In: IEEE International Conference on Robotics and Automation . 2015, pp. 156–163

    2015

  22. [30]

    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

    2022

  23. [31]

    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: arXiv preprint arXiv:2410.07554 (2024)

    2024 arXiv

  24. [32]

    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

    2024

  25. [33]

    Con- trol of Generalized Contact Motion and Force in Physical Human- Robot Interaction

    Emanuele Magrini, Fabrizio Flacco, and Alessandro De Luca. “Con- trol of Generalized Contact Motion and Force in Physical Human- Robot Interaction”. In: IEEE International Conference on Robotics and Automation . 2015, pp. 2298–2304

    2015

  26. [34]

    Compliance and Force Control for Computer Controlled Manipulators

    Matthew T Mason. “Compliance and Force Control for Computer Controlled Manipulators”. In: IEEE Transactions on Systems, Man, and Cybernetics 11.6 (1981), pp. 418–432

    1981

  27. [35]

    Hearing Touch: Audio-Visual Pretraining for Contact-Rich Manipulation

    Jared Mejia et al. “Hearing Touch: Audio-Visual Pretraining for Contact-Rich Manipulation”. In: arXiv preprint arXiv:2405.08576 (2024)

    2024 arXiv

  28. [36]

    FORGE: Force-Guided Exploration for Robust Contact-Rich Manipulation under Uncertainty

    Michael Noseworthy et al. “FORGE: Force-Guided Exploration for Robust Contact-Rich Manipulation under Uncertainty”. In: arXiv preprint arXiv:2408.04587 (2024)

    2024 arXiv

  29. [37]

    Octo: An Open-Source Generalist Robot Policy

    Octo Model Team et al. “Octo: An Open-Source Generalist Robot Policy”. In: Robotics: Science and Systems . 2024

    2024

  30. [38]

    Hybrid Position/Force Control of Manip- ulators

    MH Raibert and JJ Craig. “Hybrid Position/Force Control of Manip- ulators”. In: Journal of Dynamic Systems, Measurement, and Control 103.2 (1981), pp. 126–133

    1981

  31. [39]

    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)

    2024 arXiv

  32. [40]

    Denoising Diffu- sion Implicit Models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. “Denoising Diffu- sion Implicit Models”. In: The International Conference on Learning Representations. 2021

    2021

  33. [41]

    A Survey of Robot Manipulation in Contact

    Markku Suomalainen, Yiannis Karayiannidis, and Ville Kyrki. “A Survey of Robot Manipulation in Contact”. In: Robotics and Au- tonomous Systems 156 (2022), p. 104224

    2022

  34. [42]

    A New Silicone Structure for uSkin - A Soft, Distributed, Digital 3-Axis Skin Sensor and Its Integration on the Humanoid Robot iCub

    Tito Pradhono Tomo et al. “A New Silicone Structure for uSkin - A Soft, Distributed, Digital 3-Axis Skin Sensor and Its Integration on the Humanoid Robot iCub”. In: IEEE Robotics and Automation Letters 3.3 (2018), pp. 2584–2591

    2018

  35. [43]

    Attention is All you Need

    Ashish Vaswani et al. “Attention is All you Need”. In: Advances in Neural Information Processing Systems . 2017, pp. 5998–6008

    2017

  36. [44]

    RISE: 3D Perception Makes Real-World Robot Imitation Simple and Effective

    Chenxi Wang et al. “RISE: 3D Perception Makes Real-World Robot Imitation Simple and Effective”. In: arXiv preprint arXiv:2404.12281 (2024)

    2024 arXiv

  37. [45]

    Historical Perspective and State of the Art in Robot Force Control

    Daniel E. Whitney. “Historical Perspective and State of the Art in Robot Force Control”. In: IEEE International Conference on Robotics and Automation . 1985, pp. 262–268

    1985

  38. [46]

    TacDiffusion: Force-domain Diffusion Policy for Precise Tactile Manipulation

    Yansong Wu et al. “TacDiffusion: Force-domain Diffusion Policy for Precise Tactile Manipulation”. In: arXiv preprint arXiv:2409.11047 (2024)

    2024 arXiv

  39. [47]

    CAGE: Causal Attention Enables Data- Efficient Generalizable Robotic Manipulation

    Shangning Xia et al. “CAGE: Causal Attention Enables Data- Efficient Generalizable Robotic Manipulation”. In: arXiv preprint arXiv:2410.14974 (2024)

    2024 arXiv

  40. [48]

    MimicTouch: Leveraging Multi-Modal Human Tac- tile Demonstrations for Contact-Rich Manipulation

    Kelin Yu et al. “MimicTouch: Leveraging Multi-Modal Human Tac- tile Demonstrations for Contact-Rich Manipulation”. In: Conference on Robot Learning . 2024

    2024

  41. [49]

    GelSight: High-Resolution Robot Tactile Sensors for Estimating Geometry and Force

    Wenzhen Yuan, Siyuan Dong, and Edward H. Adelson. “GelSight: High-Resolution Robot Tactile Sensors for Estimating Geometry and Force”. In: Sensors 17.12 (2017)

    2017

  42. [50]

    Transporter Networks: Rearranging the Visual World for Robotic Manipulation

    Andy Zeng et al. “Transporter Networks: Rearranging the Visual World for Robotic Manipulation”. In: Conference on Robot Learning. 2020, pp. 726–747

    2020

  43. [51]

    Transferable Tactile Transformers for Represen- tation Learning Across Diverse Sensors and Tasks

    Jialiang Zhao et al. “Transferable Tactile Transformers for Represen- tation Learning Across Diverse Sensors and Tasks”. In: Conference on Robot Learning . 2024

    2024

  44. [52]

    Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware

    Tony Z. Zhao et al. “Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware”. In: Robotics: Science and Systems . 2023

    2023

  45. [53]

    Admittance Visuomotor Policy Learning for General-Purpose Contact-Rich Manipulations

    Bo Zhou et al. “Admittance Visuomotor Policy Learning for General-Purpose Contact-Rich Manipulations”. In: arXiv preprint arXiv:2409.14440 (2024)

    2024 arXiv

  46. [54]

    RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

    Brianna Zitkovich et al. “RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control”. In: Conference on Robot Learning . 2023, pp. 2165–2183. APPENDIX A. Implementation Details Data Processing. Following RISE [44], we create the point cloud from a single-vi...

    2023

  47. [2048]

    Force/Torque Encoder

    The readout token has a dimension of 512. Force/Torque Encoder. The high-frequency force/torque observation of the last To = 200 steps (approximately 2 seconds given the frequency of 100Hz) is encoded via a 3-layer MLP of dimension (64, 128, 512). We use the same transformer a...

    2000

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.