Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Compact tactile force fields beat high-res touch images for robotic disassembly

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 18:29 UTC pith:Z7YKFFKJ

load-bearing objection A well-structured empirical study on tactile representation for disassembly, but the main claim is confounded by missing resolution-matched controls and an underspecified force-field extraction. the 3 major comments →

arxiv 2603.08560 v2 pith:Z7YKFFKJ submitted 2026-03-09 cs.RO

CONTACT: CONtact-aware TACTile Learning for Robotic Disassembly

Yosuke Saka , Jyun-Chi Hu , Adeesh Desai , Zhiyuan Zhang , Bihao Zhang , Quan Khanh Luu , Md Rakibul Islam Prince , Minghui Zheng
show 1 more author
Yu She
This is my paper
classification cs.RO
keywords tactile sensingrobotic disassemblyforce-field representationdiffusion policymultimodal manipulationcontact-rich tasksdeformable objectsimitation learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper tries to establish that, for contact-rich robotic disassembly, the representation of tactile information matters more than its presence. It compares vision alone, vision plus high-resolution tactile images (TacRGB), and vision plus a compact tactile force field (TacFF) under a single imitation-learning policy. Across five simulation and five real-world tasks, TacFF consistently yields the highest success rates, with the biggest gains in tight-tolerance, asymmetric-resistance, and deformable scenarios. The paper also shows that naively combining TacRGB and TacFF degrades performance, indicating that unstructured and structured tactile signals dilute each other. A sympathetic reader would care because disassembly is a contact-dominated, force-sensitive process where visual cues are often ambiguous, and the paper offers a concrete representation choice that improves robustness.

Core claim

The central claim is that a compact 10x14 force-field representation of touch — encoding shear direction and magnitude plus normal force at each cell — is the effective tactile modality for contact-rich disassembly, and that adding it to vision outperforms both vision alone and vision plus high-resolution tactile imagery. The evidence includes simulation results where TacFF raises success in a tight socket from 36.9% to 44.3% (S2) and real-world results where it reaches 70% in the matching R2 task, and 55% on a deformable clip (R5) where vision-only collapses to 0% under dim lighting. The paper's task-by-task pattern shows tactile sensing matters most when geometric constraints tighten, resi

What carries the argument

TacFF (tactile force field) is the central object: a 10x14 grid in which each cell encodes shear-x, shear-y, and normal force components, computed from optical flow and depth reconstruction of the contact surface. It is the load-bearing modality in the paper's comparison, because the success-rate gains appear only when this structured force summary is included, not when high-resolution tactile images (TacRGB) are added. The policy is a diffusion-model-based visuomotor controller that encodes front and wrist RGB, optional tactile inputs, and end-effector pose with ResNet-18 backbones, concatenates two-step history features, and denoises action chunks (horizon 16, execution of the first 8 step

Load-bearing premise

The entire comparison rests on the assumption that TacFF is computed faithfully from optical flow and depth reconstruction — an algorithm the paper never specifies — such that the 10x14 grid in simulation and on the real sensor carries the same physical meaning; if the computation is fragile or inconsistent, the modality comparison is confounded by representation quality.

What would settle it

A reader could settle the modality claim by training the policy with (1) the raw optical-flow field before it is reduced to the 10x14 grid, and (2) a spatially shuffled or permuted TacFF grid, on the same tasks. If the raw or shuffled versions match TacFF's success rates, the compact structured encoding is not the cause. Separately, if a TacFF grid computed from simulated depth reconstruction fails to produce comparable success when the real sensor is replaced with a different shear/normal calibration, the sim-to-real consistency assumption is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • For contact-rich disassembly, policies should be built on compact structured force-field tactile input rather than high-resolution tactile imagery.
  • Tactile sensing acts as a stabilizing signal: in dim lighting, TacFF maintained success on a deformable clip task while vision-only dropped to zero.
  • Naive concatenation of heterogeneous tactile representations can reduce success rates, so multimodal fusion for force-sensitive tasks needs structured or attention-based integration.
  • The task-dependent pattern generalizes: gains concentrate in tight-tolerance, asymmetric-resistance, and deformable scenarios, not in loose geometry-dominant ones.

Where Pith is reading between the lines

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

  • A direct test of the representation claim would be to train the same policy with a spatially shuffled TacFF grid; if shuffled grids retain the gains, the spatial structure of the force field is not what carries the information.
  • The paper leaves the TacFF computation unspecified; if the optical-flow-to-force-field algorithm is published and shown to transfer across different GelSight-like sensors, the result would likely extend to other force-sensitive manipulation skills such as insertion, prying, and cable mating.
  • The fusion failure suggests a likely research direction: learning a gating or attention mechanism over tactile streams could recover the lost gains, a hypothesis the paper itself flags.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper investigates tactile sensing for robotic disassembly using a unified Diffusion Policy framework. It introduces five simulated rigid-body tasks (S1–S5) and five real-world tasks (R1–R5, including deformable components), and compares three sensing configurations: Vision Only, Vision + TacRGB (high-resolution tactile image), and Vision + TacFF (compact 10×14 force-field grid). Across simulation and real-world experiments, TacFF-based policies achieve the highest success rates, with the largest margins in contact-dependent and deformable tasks. The paper also reports that naive concatenation of TacRGB and TacFF degrades performance, arguing that structured force representations are more effective than high-dimensional tactile imagery.

Significance. If the central claim holds, the paper makes a valuable, actionable contribution to tactile representation design for contact-rich manipulation: it suggests that compact force-field encodings can outperform high-resolution tactile images, and it provides a disassembly benchmark spanning increasing contact complexity. The strengths are the unified policy framework (all modalities compared under identical training protocols), a relatively large simulation evaluation (1500 rollouts per configuration per task), and real-world tasks that include deformable contact interactions, which are under-represented in prior tactile manipulation benchmarks. The consistent pattern across simulation and real-world for the shared tasks (S1/R1, S2/R2, S3/R3) is a useful indicator. However, the load-bearing conclusion about representation semantics is currently underdetermined by the experiments: the comparison does not control for input dimensionality, and the TacFF representation is not specified at a level that permits reproduction or assessment of sim-to-real fidelity.

major comments (3)
  1. [§III.A and Tables I–II] The central claim is that TacFF's advantage over TacRGB is caused by structured force semantics, not by lower input dimensionality. But the comparison is confounded: TacRGB is a 240×160 image, while TacFF is a 10×14 grid, and both are encoded by the same ResNet-18. With only 50 demonstrations and 400 training epochs, a higher-dimensional input may underperform due to overfitting or representation sparsity rather than because force content is worse. I ask for a matched-resolution control: downsample TacRGB to 10×14 (or upsample/project TacFF to a comparable visual form) and retrain. Without such a control, the abstract and conclusions' claim that 'structured force fields, not high-resolution imagery' is the active ingredient is not supported.
  2. [§II.B, TacFF definition] The TacFF representation is described only qualitatively: 'computed from optical flow and depth reconstruction' with shear components and normal force over a grid. No algorithm, equations, calibration procedure, or mapping between GelSight R1.5 measurements and the 10×14 grid is given. It is also unclear whether the simulation (TacSL) and the real GelSight sensor use the same extraction pipeline and normalization; Fig. 3 suggests visual similarity but does not quantify agreement. Because Table II's central comparison depends on the quality and consistency of TacFF, the absence of this detail is a reproducibility and validity gap. Please provide the extraction method (or cite a public implementation), the calibration steps, and at least a qualitative/quantitative validation that the sim and real force fields correspond.
  3. [§III.C, Tables II, IV, V] The real-world results rest on 20 trials per configuration per task, with no confidence intervals or statistical tests. Several cell-to-cell swings are large (e.g., R2: TacRGB 30% vs TacFF 70%; R4: 5% vs 75%), but with n=20 the standard error is around 10 percentage points at p≈0.5, so some differences that drive the 'consistently highest' claim may not be statistically reliable. The dim-lighting and fusion results (Tables IV and V) use the same small n. Table I reports 1500 rollouts per configuration, but no standard errors or seed-level variability are shown, and some margins (e.g., S3: 53.5 vs 50.0; S5: 61.9 vs 59.5) are small. Please include confidence intervals, per-seed breakdowns, and/or significance tests for the key comparisons, and if possible increase the real-world rollout count or clearly state the precision limitation.
minor comments (5)
  1. [§I, first paragraph] The sentence 'Robotic disassembly is a critical capability for recycling, maintenance, and remanufacturing.' is duplicated verbatim.
  2. [Fig. 1] The bar chart's horizontal axis is labeled 'Modalities Design', which is unclear; the individual bars are not labeled with the task or modality names. Please make the figure self-contained.
  3. [§III.A, action space] The action is described as 'restricted to 3D translation and yaw rotation' but is then defined as a 5D command including a gripper command; clarify the dimension count and what the gripper command encodes.
  4. [§II.B / §III.A] The paper states that all observations are 'normalized to [−1, 1]', but it does not say how TacRGB and TacFF are normalized (per-channel min–max, per-sensor calibration, etc.). This matters for reproducibility of the modality comparison.
  5. [§III.D.2, Table V] The combined TacRGB+TacFF configuration is only tested on two real tasks with 20 rollouts each. The conclusion that 'naive concatenation dilutes information' would be stronger with more tasks and more trials, or with a companion simulation experiment where rollouts are plentiful.

Circularity Check

0 steps flagged

No circular derivation; empirical modality comparison is measured, not fitted.

full rationale

The paper's central claims (TacFF outperforming TacRGB and Vision Only) rest on measured success rates from trained policies evaluated on held-out rollouts (Tables I, II, IV, V), not on a derivation that reduces outputs to inputs. TacFF is introduced as a representation computed from optical flow and depth reconstruction (Section II.B), but no equations are given, so there is no formal derivation chain to be circular. The main threats to the conclusion are confounds: TacRGB is 240x160 while TacFF is 10x14 (Section III.A), so the comparison does not isolate compactness from force semantics, and the TacFF extraction algorithm is unspecified, leaving sim-to-real consistency unverified. These are correctness and reproducibility concerns, not circularity. The paper's self-citations (e.g., [24], [38], [31]-[33]) provide background, a simulator, and tactile-sensor references; none is used as a load-bearing uniqueness theorem or as proof that TacFF must win. The Diffusion Policy backbone [5] is external, and no fitted parameter is renamed as a prediction. No circular step is therefore identified.

Axiom & Free-Parameter Ledger

3 free parameters · 3 axioms · 0 invented entities

No new physical entities are introduced. The main ledger items are experimental design choices and unstated calibration assumptions around the TacFF grid.

free parameters (3)
  • TacFF grid resolution (10×14) and its shear/normal normalization = 10×14 grid; components normalized to [−1, 1]
    Chosen by hand without ablation; the central claim that TacFF is better could in principle depend on the grid granularity and normalization, although this is an experimental design choice rather than a fitted constant.
  • Diffusion Policy architecture hyperparameters (horizon H=16, DDIM K=10, 8 executed actions, 400 epochs) = H=16, K=10, 8 actions, 400 epochs
    Fixed across all compared configurations, so it does not bias the modality comparison, but the absolute success rates depend on these values.
  • Randomization ranges (±5 cm, 10°) = Ranges stated
    These define task difficulty and therefore the size of the modality gap. They were chosen by hand, not swept.
axioms (3)
  • domain assumption The simulation TacSL tactile rendering is a faithful proxy for the real GelSight sensor for the purpose of policy learning transfer.
    Invoked in Section III.A for all simulated tasks; if the sim-to-real tactile gap is large, the sim conclusions about TacFF vs TacRGB may not transfer.
  • domain assumption Success rate over 20 real-world rollouts per cell is a sufficient statistic to rank modalities.
    Used in Tables II, IV, V; with n=20, a 10% difference is a single trial, and the paper reports no CIs or repeated-trials statistics.
  • domain assumption The GelSight-based optical-flow force-field computation yields calibrated physical forces.
    The exact relation between the raw GelSight deformation and the shear/normal grid is not specified, yet the whole conclusion rests on this grid encoding being informative and consistent.

pith-pipeline@v1.3.0-alltime-deepseek · 10437 in / 5896 out tokens · 43249 ms · 2026-08-02T18:29:33.368051+00:00 · methodology

0 comments
read the original abstract

Robotic disassembly involves contact-rich interactions in which successful manipulation depends not only on geometric alignment but also on force-dependent state transitions. While vision-based policies perform well in structured settings, their reliability often degrades in tight-tolerance, contact-dominated, or deformable scenarios. In this work, we systematically investigate the role of tactile sensing in robotic disassembly through both simulation and real-world experiments. We construct five rigid-body disassembly tasks in simulation with increasing geometric constraints and extraction difficulty. We further design five real-world tasks, including three rigid and two deformable scenarios, to evaluate contact-dependent manipulation. Within a unified learning framework, we compare three sensing configurations: Vision Only, Vision + tactile RGB (TacRGB), and Vision + tactile force field (TacFF). Across both simulation and real-world experiments, TacFF-based policies consistently achieve the highest success rates, with particularly notable gains in contact-dependent and deformable settings. Notably, naive fusion of TacRGB and TacFF underperforms either modality alone, indicating that simple concatenation can dilute task-relevant force information. Our results show that tactile sensing plays a critical, task-dependent role in robotic disassembly, with structured force-field representations being particularly effective in contact-dominated scenarios.

Figures

Figures reproduced from arXiv: 2603.08560 by Adeesh Desai, Bihao Zhang, Jyun-Chi Hu, Md Rakibul Islam Prince, Minghui Zheng, Quan Khanh Luu, Yosuke Saka, Yu She, Zhiyuan Zhang.

Figure 1
Figure 1. Figure 1: Structured task design and multimodal evaluation for robotic [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Object geometries and interaction primitives of simulation and real-world disassembly tasks. Each task consists of an initial (Original) and goal [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of multimodal observations over time in simulation (left) and real-world (right) disassembly. Columns denote task stages, and rows [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visual ambiguity versus tactile disambiguation during grasp in Task [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Real-world experimental setup, including the Franka Emika Panda [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 2 Pith papers

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

  1. Imagining the Sense of Touch: Touch-Informed Manipulation via Imagined Tactile Representations

    cs.RO 2026-07 unverdicted novelty 5.0

    TacImag framework trains on paired visuotactile data to predict tactile observations from vision, improving performance on six simulated and four real-world manipulation tasks.

  2. Redefining End-of-Life: Intelligent Automation for Electronics Remanufacturing Systems

    eess.SY 2026-04 unverdicted novelty 2.0

    A literature review of intelligent automation approaches using robotics, AI, and control for disassembly, inspection, sorting, and reprocessing of end-of-life electronics.

Reference graph

Works this paper leans on

40 extracted references · 10 linked inside Pith · cited by 2 Pith papers

  1. [1]

    D. E. Whitney. Quasi-static assembly of compliantly supported rigid parts.Journal of Dynamic Systems, Measurement, and Control, 104(1):65–77, March 1982

    1982

  2. [2]

    Intuitive peg-in-hole assembly strategy with a compliant manipulator

    Hyeonjun Park, Ji-Hun Bae, Jae-Han Park, Moon-Hong Baeg, and Jae- heung Park. Intuitive peg-in-hole assembly strategy with a compliant manipulator. InIEEE ISR 2013, pages 1–5. IEEE, 2013

    2013

  3. [3]

    Challenges of robotic disassembly in practice.Procedia CIRP, 105:513–518, 2022

    Gwendolyn Foo, Sami Kara, and Maurice Pagnucco. Challenges of robotic disassembly in practice.Procedia CIRP, 105:513–518, 2022

    2022

  4. [4]

    Learning fine-grained bimanual manipulation with low-cost hardware

    Tony Z Zhao, Vikash Kumar, Sergey Levine, and Chelsea Finn. Learning fine-grained bimanual manipulation with low-cost hardware. InProceedings of Robotics: Science and Systems, 2023

    2023

  5. [5]

    Diffusion policy: Visuomotor policy learning via action diffusion

    Cheng Chi, Siyuan Feng, Yilun Du, Zhenjia Xu, Eric Cousineau, Benjamin Burchfiel, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion. InProceedings of Robotics: Science and Systems, 2023

    2023

  6. [6]

    3D Diffusion Policy: Generalizable visuomotor policy learning via simple 3d representations

    Yanjie Ze, Gu Zhang, Kangning Zhang, Chenyuan Hu, Muhan Wang, and Huazhe Xu. 3D Diffusion Policy: Generalizable visuomotor policy learning via simple 3d representations. InProceedings of Robotics: Science and Systems, 2024

    2024

  7. [7]

    Canonical policy: Learning canonical 3d representation for equivariant policy.arXiv preprint arXiv:2505.18474, 2025

    Zhiyuan Zhang, Zhengtong Xu, Jai Nanda Lakamsani, and Yu She. Canonical policy: Learning canonical 3d representation for equivariant policy.arXiv preprint arXiv:2505.18474, 2025

    arXiv 2025

  8. [8]

    Equiform: Noise-robust se(3)-equivariant policy learning from 3d point clouds, 2026

    Zhiyuan Zhang and Yu She. Equiform: Noise-robust se(3)-equivariant policy learning from 3d point clouds, 2026

    2026

  9. [9]

    Vibecheck: Using active acoustic tactile sensing for contact- rich manipulation

    Kaidi Zhang, Do-Gon Kim, Eric T Chang, Hua-Hsuan Liang, Zhan- peng He, Kathryn Lampo, Philippe Wu, Ioannis Kymissis, and Matei Ciocarlie. Vibecheck: Using active acoustic tactile sensing for contact- rich manipulation. In2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 12278–12285. IEEE, 2025

    2025

  10. [10]

    Openvla: An open-source vision-language-action model.arXiv preprint arXiv:2406.09246, 2024

    Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti, Ted Xiao, Ashwin Balakrishna, Suraj Nair, Rafael Rafailov, Ethan Foster, Grace Lam, Pannag Sanketi, et al. Openvla: An open-source vision-language-action model.arXiv preprint arXiv:2406.09246, 2024

    Pith/arXiv arXiv 2024

  11. [11]

    Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, et al.π 0.5: a vision-language-action model with open-world generalization.arXiv preprint arXiv:2504.16054, 2025

    Pith/arXiv arXiv 2025

  12. [12]

    Augmenting robotic disassembly skill: combining compliance control strategy with reinforcement learning for twist-pulling disassembly

    Yue Zang, Xiazhen Xu, Yongquan Zhang, Amir M Hajiyavand, Jiaqi Ye, and Yongjing Wang. Augmenting robotic disassembly skill: combining compliance control strategy with reinforcement learning for twist-pulling disassembly. In2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 20297–20302. IEEE, 2025

    2025

  13. [13]

    Heng Zhang, Gokhan Solak, Gustavo J. G. Lahr, and Arash Ajoudani. Srl-vic: A variable stiffness-based safe reinforcement learning for contact-rich robotic tasks.IEEE Robotics and Automation Letters, 9(6):5631–5638, 2024

    2024

  14. [14]

    Safe learning for contact-rich robot tasks: A survey from classical learning-based methods to safe foundation models

    Heng Zhang, Rui Dai, Gokhan Solak, Pokuang Zhou, Yu She, and Arash Ajoudani. Safe learning for contact-rich robot tasks: A survey from classical learning-based methods to safe foundation models. arXiv preprint arXiv:2512.11908, 2025

    arXiv 2025

  15. [15]

    A survey on imi- tation learning for contact-rich tasks in robotics.arXiv preprint arXiv:2506.13498, 2025

    Toshiaki Tsuji, Yasuhiro Kato, Gokhan Solak, Heng Zhang, Tadej Petriˇc, Francesco Nori, and Arash Ajoudani. A survey on imi- tation learning for contact-rich tasks in robotics.arXiv preprint arXiv:2506.13498, 2025

    Pith/arXiv arXiv 2025

  16. [16]

    Jeon Ho Kang, Sagar Joshi, Ruopeng Huang, and Satyandra K. Gupta. Robotic compliant object prying using diffusion policy guided by vision and force observations, 2025

    2025

  17. [17]

    Task-context-aware diffusion policy with language guidance for multi- task disassembly

    Jeon Ho Kang, Sagar Joshi, Neel Dhanaraj, and Satyandra K Gupta. Task-context-aware diffusion policy with language guidance for multi- task disassembly. In2025 IEEE 21st International Conference on Automation Science and Engineering (CASE), pages 609–616. IEEE, 2025

    2025

  18. [18]

    Segment everything everywhere all at once.Advances in neural information processing systems, 36:19769–19782, 2023

    Xueyan Zou, Jianwei Yang, Hao Zhang, Feng Li, Linjie Li, Jianfeng Wang, Lijuan Wang, Jianfeng Gao, and Yong Jae Lee. Segment everything everywhere all at once.Advances in neural information processing systems, 36:19769–19782, 2023

    2023

  19. [19]

    Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193, 2023

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193, 2023

    Pith/arXiv arXiv 2023

  20. [20]

    Dinov3.arXiv preprint arXiv:2508.10104, 2025

    Oriane Sim ´eoni, Huy V V o, Maximilian Seitzer, Federico Baldas- sarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Micha ¨el Ramamonjisoa, et al. Dinov3.arXiv preprint arXiv:2508.10104, 2025

    Pith/arXiv arXiv 2025

  21. [21]

    Point transformer v3: Simpler faster stronger

    Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xihui Liu, Yu Qiao, Wanli Ouyang, Tong He, and Hengshuang Zhao. Point transformer v3: Simpler faster stronger. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4840–4851, 2024

    2024

  22. [22]

    Droid: A large-scale in-the-wild robot manipulation dataset.arXiv preprint arXiv:2403.12945, 2024

    Alexander Khazatsky, Karl Pertsch, Suraj Nair, Ashwin Balakrishna, Sudeep Dasari, Siddharth Karamcheti, Soroush Nasiriany, Mohan Ku- mar Srirama, Lawrence Yunliang Chen, Kirsty Ellis, et al. Droid: A large-scale in-the-wild robot manipulation dataset.arXiv preprint arXiv:2403.12945, 2024

    Pith/arXiv arXiv 2024

  23. [23]

    Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation.arXiv preprint arXiv:2506.18088, 2025

    Tianxing Chen, Zanxin Chen, Baijun Chen, Zijian Cai, Yibin Liu, Qiwei Liang, Zixuan Li, Xianliang Lin, Yiheng Ge, Zhenyu Gu, et al. Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation.arXiv preprint arXiv:2506.18088, 2025

    Pith/arXiv arXiv 2025

  24. [24]

    Manifeel: Benchmarking and under- standing visuotactile manipulation policy learning.arXiv preprint arXiv:2505.18472, 2025

    Quan Khanh Luu, Pokuang Zhou, Zhengtong Xu, Zhiyuan Zhang, Qiang Qiu, and Yu She. Manifeel: Benchmarking and under- standing visuotactile manipulation policy learning.arXiv preprint arXiv:2505.18472, 2025

    arXiv 2025

  25. [25]

    Tactile mnist: Benchmarking active tactile perception.arXiv preprint arXiv:2506.06361, 2025

    Tim Schneider, Guillaume Duret, Cristiana de Farias, Roberto Calan- dra, Liming Chen, and Jan Peters. Tactile mnist: Benchmarking active tactile perception.arXiv preprint arXiv:2506.06361, 2025

    Pith/arXiv arXiv 2025

  26. [26]

    Vibrissae-inspired vision-based magnetic-actuated whisker.Nature Communications, 2025

    Zhixian Hu, Yi Cheng, Juan Wachs, and Yu She. Vibrissae-inspired vision-based magnetic-actuated whisker.Nature Communications, 2025

    2025

  27. [27]

    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. Sensors, 17(12):2762, 2017

    2017

  28. [28]

    Geltip: A finger-shaped optical tactile sensor for robotic manipulation

    Daniel Fernandes Gomes, Zhonglin Lin, and Shan Luo. Geltip: A finger-shaped optical tactile sensor for robotic manipulation. In2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 9903–9909. IEEE, 2020

    2020

  29. [29]

    Gelroller: A rolling vision-based tactile sensor for large surface reconstruction using self-supervised photometric stereo method

    Zhiyuan Zhang, Huan Ma, Yulin Zhou, Jingjing Ji, and Hua Yang. Gelroller: A rolling vision-based tactile sensor for large surface reconstruction using self-supervised photometric stereo method. In 2024 IEEE International Conference on Robotics and Automation (ICRA), pages 7961–7967. IEEE, 2024

    2024

  30. [30]

    Gelflow: Self- supervised learning of optical flow for vision-based tactile sensor displacement measurement

    Zhiyuan Zhang, Hua Yang, and Zhouping Yin. Gelflow: Self- supervised learning of optical flow for vision-based tactile sensor displacement measurement. InInternational Conference on Intelligent Robotics and Applications, pages 26–37. Springer, 2023

    2023

  31. [31]

    Tactile-reactive gripper with an active palm for dexterous manipulation.npj Robotics, 4(1):13, 2026

    Yuhao Zhou, Won Suk Lee, Yan Gu, and Yu She. Tactile-reactive gripper with an active palm for dexterous manipulation.npj Robotics, 4(1):13, 2026

    2026

  32. [32]

    Vistac toward a unified multimodal sensing finger for robotic manip- ulation.IEEE Sensors Journal, 23(20):25440–25450, 2023

    Sheeraz Athar, Gaurav Patel, Zhengtong Xu, Qiang Qiu, and Yu She. Vistac toward a unified multimodal sensing finger for robotic manip- ulation.IEEE Sensors Journal, 23(20):25440–25450, 2023

    2023

  33. [33]

    Vibtac: A high-resolution high-bandwidth tactile sensing finger for multi-modal perception in robotic manipulation.IEEE Transactions on Haptics, 18(3):507–518, 2025

    Sheeraz Athar, Xinwei Zhang, Jun Ueda, Ye Zhao, and Yu She. Vibtac: A high-resolution high-bandwidth tactile sensing finger for multi-modal perception in robotic manipulation.IEEE Transactions on Haptics, 18(3):507–518, 2025

    2025

  34. [34]

    Visuo-tactile world models.arXiv preprint arXiv:2602.06001, 2026

    Carolina Higuera, Sergio Arnaud, Byron Boots, Mustafa Mukadam, Francois Robert Hogan, and Franziska Meier. Visuo-tactile world models.arXiv preprint arXiv:2602.06001, 2026

    arXiv 2026

  35. [35]

    Residual rotation correction using tactile equivariance

    Yizhe Zhu, Zhang Ye, Boce Hu, Haibo Zhao, Yu Qi, Dian Wang, and Robert Platt. Residual rotation correction using tactile equivariance. arXiv preprint arXiv:2511.07381, 2025

    arXiv 2025

  36. [36]

    Vitacformer: Learning cross-modal representation for visuo- tactile dexterous manipulation.arXiv preprint arXiv:2506.15953, 2025

    Liang Heng, Haoran Geng, Kaifeng Zhang, Pieter Abbeel, and Jitendra Malik. Vitacformer: Learning cross-modal representation for visuo- tactile dexterous manipulation.arXiv preprint arXiv:2506.15953, 2025

    Pith/arXiv arXiv 2025

  37. [37]

    Isaac gym: High perfor- mance gpu-based physics simulation for robot learning, 2021

    Viktor Makoviychuk, Lukasz Wawrzyniak, Yunrong Guo, Michelle Lu, Kier Storey, Miles Macklin, David Hoeller, Nikita Rudin, Arthur Allshire, Ankur Handa, and Gavriel State. Isaac gym: High perfor- mance gpu-based physics simulation for robot learning, 2021

    2021

  38. [38]

    Tacsl: A library for visuotactile sensor simulation and learning.IEEE Transactions on Robotics, 2025

    Iretiayo Akinola, Jie Xu, Jan Carius, Dieter Fox, and Yashraj Narang. Tacsl: A library for visuotactile sensor simulation and learning.IEEE Transactions on Robotics, 2025

    2025

  39. [39]

    Gelsight wedge: Measuring high-resolution 3d contact geometry with a compact robot finger

    Shaoxiong Wang, Yu She, Branden Romero, and Edward Adelson. Gelsight wedge: Measuring high-resolution 3d contact geometry with a compact robot finger. In2021 IEEE international conference on robotics and automation (ICRA), pages 6468–6475. IEEE, 2021

    2021

  40. [40]

    Denoising diffusion implicit models.arXiv preprint arXiv:2010.02502, 2020

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models.arXiv preprint arXiv:2010.02502, 2020

    Pith/arXiv arXiv 2010