Pith. sign in

REVIEW 2 major objections 4 minor 2 cited by

LightTact: A Visual-Tactile Fingertip Sensor for Deformation-Independent Contact Sensing

T0 review · 2 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

Pith's one-line read LightTact makes contact visible by blocking all light except at the true contact point, so robots can sense even zero-force touch.

desk verdict A clever optical design with strong dark-background results; the 'visible iff contact' claim is conditional on diffuse scattering and needs quantitative evaluation. read the letter →

arxiv 2512.20591 v3 pith:3M55UN7P submitted 2025-12-23 cs.RO

classification cs.RO
keywords visual-tactilesensingcontactsegmentationdeformation-independenttotalinternalreflectionfingertipsensorrobotmanipulationvision-languagemodelsopticaldesign
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

This paper presents LightTact, a camera-based fingertip sensor that detects physical contact directly through optics rather than through deformation of a soft surface. The sensor's optical layout is engineered so that both external light and internal LED illumination are blocked at non-contact regions, and only light scattered at true contact points reaches the camera. In raw images, non-contact pixels stay near-black under ambient lighting up to 2010 lux, while contact pixels preserve the object's natural appearance. This yields simple, robust pixel-level contact segmentation that works across liquids, semi-liquids, ultra-soft materials, and rigid objects, with no minimum force threshold. If correct, this would remove a major blind spot in robot touch: interactions too gentle to indent a surface.

What carries the argument

The key mechanism is an 'ambient-blocking' optical configuration: a transparent medium with a soft touch surface and a rigid acrylic viewing surface set roughly perpendicular (θ_tv = 90°, above 2θ_c ≈ 86°). External light entering non-contact regions is rejected by refraction-then-TIR; internal LED rays reflecting off the touch surface are rejected when the LED lies within a computed admissible region; and only diffuse-scattered rays from true contact transmit through the viewing surface. This single design simultaneously suppresses background and captures natural appearance, requiring no deformation.

What would settle it

Press a near-perfect mirror or a black glass slide against the sensing surface and observe whether the contact region brightens relative to non-contact regions; if it does not, the bijective visibility claim fails. Also test a fully transparent acrylic block with a black backing to isolate whether diffuse scattering is truly required.

Watch

Extended reading notes

Core claim

The central claim is that a visual-tactile sensor can enforce a bijective contact-visibility relationship—a pixel is visible if and only if it is in true physical contact—by using a non-parallel touching-surface/viewing-surface wedge geometry and careful LED placement. All light paths that would enter the camera through non-contact regions are made to hit the viewing surface at angles exceeding the critical angle, causing total internal reflection into black absorbing surfaces; only diffuse scattering generated at contact points refracts into the camera. The paper provides a geometric analysis (wedge angle > 2× the critical angle) and validates the behavior experimentally with intensity stat

Load-bearing premise

The optical principle requires that a true contact surface diffusely scatters the internal LED light; if the object is a mirror, a strong absorber, or fully transparent, contact pixels may stay dark and the 'visible iff contact' equivalence breaks.

Editorial extensions

If this is right

  • Robots can sense contact with liquids, semi-liquids, and ultra-soft materials where deformation-based tactile sensors give no signal.
  • Pixel-level contact maps can be obtained by simple frame differencing and thresholding, without complex learning-based segmentation.
  • The sensor can operate in bright ambient lighting well above typical indoor levels, unlike prior TIR-based sensors.
  • The same RGB image provides both contact geometry and surface appearance, so vision-language models can reason over it directly.
  • Zero-force contact, such as a hanging thin film touching the surface, becomes detectable.

Reading between the lines

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

  • The diffuse-scattering assumption implies that mirror-finish or transparent objects may not produce visible contact pixels; testing those could bound the class of surfaces the sensor can 'see'.
  • The wedge geometry might be adapted to curved or flexible fingertips if a rigid viewing window can be integrated, suggesting alternative form factors beyond the flat fingertip.
  • Because the sensor outputs appearance-preserving images aligned with contact, it could serve as a front-end for learned visuotactile manipulation policies.
  • The demonstrated liquid and thin-film tasks point toward biomedical and hygiene applications, such as detecting liquid spreading or gentle tissue contact, where deformation-based sensing historically fails.
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

2 major / 4 minor

Summary. LightTact is a compact (12×18×34.5 mm) vision-based tactile fingertip sensor that aims to make physical contact directly visible through an optics-based, deformation-independent principle. The sensor uses a transparent wedge-shaped medium with a touching surface and a viewing surface. A side-view camera and a single internal LED are arranged so that external light and internal LED illumination at non-contact regions are rejected by total internal reflection at the viewing surface, while light diffusely scattered by a truly contacting object can refract toward the camera. The paper derives—or rather states—three geometric conditions on the wedge angle and LED placement, then describes a complete hardware implementation, a frame-differencing segmentation algorithm, calibration, and a series of experiments. Evaluations include light suppression under varying internal and external illumination, contact segmentation for liquids, semi-liquids, ultra-soft materials, and rigid objects with and without force, and robotic demonstrations of water spreading, facial-cream dipping, thin-film interaction, and VLM-guided resistor sorting. The central claim is a bijective contact–visibility relationship: a pixel is visible if and only if it is in true physical contact.

Significance. If the optical mechanism works as claimed, LightTact addresses a real shortcoming of deformation-based tactile sensors and of conventional frustrated-TIR designs, which typically require controlled lighting and fail under ambient illumination. The engineering contribution is strong: the sensor is small, low-cost, open-source, and produces high-contrast raw images with near-black non-contact pixels. The paper includes quantitative light-suppression data, a reproducible fabrication procedure, and multiple real-world manipulation demos that are difficult or impossible with existing tactile sensors. The idea of using a wedge angle larger than 2θc to reject non-contact light while transmitting contact-scattered light is plausible and is empirically supported across a range of everyday materials and lighting conditions. However, the headline 'bijective contact–visibility' claim is broader than the evidence, because the sensing mechanism relies on diffuse scattering at the contact surface; the paper does not test or analyze surfaces that are specular, transparent, or very low-albedo. This limits the generality of the central claim and needs to be addressed before publication.

major comments (2)
  1. [III-A(c), IV-B, III-D] The central claim is that a pixel is visible iff it is in true physical contact. However, the mechanism that makes contact visible is stated in Section III-A(c) as 'the contacting surface produces diffuse scattering of the LED illumination… a subset of these scattered rays… enters the camera.' This requires the contacting material to be sufficiently diffuse and reflective. The experiments in Section IV-B include liquids, semi-liquids, ultra-soft materials, and rigid objects, but do not include near-specular surfaces (e.g., polished metal, mirrors), transparent objects (e.g., clear glass or acrylic), or strongly absorbing/black surfaces. For such materials, the light scattered back into the medium could be far below the segmentation thresholds (t0=25, t1=20, t2=30, t3=40), producing false negatives and violating the bijective claim. Since the paper's main scientific contribution is exactl
  2. [III-A, Table I, Appendix VI-A] The three geometric design conditions in Table I (θtv > 2θc, θit < θtv − θc, and θtv < π/2 + θc) are central to the sensor's operation, but they are presented as assertions rather than derivations. The appendix only analyzes the alternative case θtv < 2θc; it does not prove that the chosen conditions are sufficient to guarantee the claimed rejection/transmission behaviors. In particular, the internal-illumination rejection condition θit < θtv − θc is crucial for preventing LED light from reaching the camera through non-contact regions, and no Snell's-law/TIR derivation is given for it. Without this derivation, the statement that the optical layout 'can, in theory, completely suppress both external light and internal illumination at non-contact regions' is not fully supported. Please provide a formal geometric-optics derivation (or a simulation) of these conditions, or at least a more rig
minor comments (4)
  1. [Abstract and Section IV-A] The abstract states that non-contact pixels remain near-black with mean gray value < 3. Table II shows this holds up to 2010 Lux external illumination, but at 3520 Lux the mean is 6.90. Please qualify the claim with the tested lighting range (e.g., 'up to 2010 Lux') or soften the statement.
  2. [Section III-D] The segmentation thresholds (N=10, t0=25, t1=20, t2=30, t3=40) are fixed values with no sensitivity analysis. It would strengthen the robustness claim to report how segmentation accuracy varies with these thresholds, or to explain how they were chosen.
  3. [Section III-A and Table I] The notation for the LED incidence angle is inconsistent: the text uses θit while Table I uses θ_it. Please unify.
  4. [Section IV-B] The comparison with baseline sensors (9DTact, GelSight, DelTact) in Figures 6 and 7 is qualitative. Including a quantitative metric (e.g., segmentation IoU against ground truth, or detection precision/recall) would make the robustness claim more convincing.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the optical derivation is self-contained; the diffuse-scattering assumption is an untested boundary condition, not a circular reduction.

full rationale

The claimed derivation chain starts from an optical layout and derives three conditions from Snell's law and TIR: θ_tv > 2θ_c for external-light rejection, LED placement θ_it < θ_tv - θ_c for internal-light rejection, and θ_tv < π/2 + θ_c for transmitting contact-scattered light. These are analytic conditions with stated refractive indices, and the sensor is then built to satisfy them (θ_tv = 90°, θ_c ≈ 43°). The experimental sections test the resulting behavior (near-black non-contact, bright contact) rather than fitting parameters and relabeling them as predictions. The segmentation thresholds (t0..t3) and LED level (430 Lux) are disclosed operational choices, not used to justify the optical principle. The only self-citation, 9DTact for the calibration tool, is a procedural borrowing and is not load-bearing for the central claim. The remaining concern—that the contact-implies-visibility direction relies on the contacting surface producing diffuse scattering, which may fail for specular, transparent, or low-albedo surfaces—is an unverified assumption and a boundary condition, not a circular reduction: the paper does not define contact in terms of scattering, nor does it derive the scattering assumption from the visibility claim. Therefore no circular step is exhibited.

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

The central claim rests on standard geometric optics (Snell's law, TIR) plus two domain assumptions: diffuse scattering at true contact and a clean air gap at non-contact regions. The free parameters are design and calibration choices (wedge angle, LED level, exposure, segmentation thresholds) rather than fitted physical constants. No new physical entities are postulated.

free parameters (4)
  • Wedge angle θtv = 90° (π/2)
    Hand-chosen to satisfy θtv > 2θc ≈ 86° and θtv < π/2 + θc ≈ 133°. Central to the optical suppression behavior, but not fitted to data.
  • Internal LED illumination level = 430 Lux (3.4 V drive)
    Selected empirically because higher levels overexpose contact regions while lower levels reduce appearance contrast; directly affects the reported non-contact mean gray < 3.
  • Segmentation thresholds t0..t3 and N = 25, 20, 30, 40; N=10
    Hand-chosen to separate contact from sensor noise; no sensitivity analysis or validation set is reported.
  • Camera exposure time = 20 ms
    Fixed to keep the dark background near-black; chosen manually rather than derived from first principles.
assumptions (4)
  • standard math Snell's law and total internal reflection with nm ≈ 1.45, na ≈ 1.0.
    Used to derive the rejection/transmission conditions in Section III-A (θtv > 2θc, θit < θtv − θc, θtv < π/2 + θc).
  • domain assumption At true contact the object's surface produces diffuse scattering of internal LED light; at non-contact regions the gel-air interface is clean and specular.
    Section III-A(c): contact visibility depends on diffuse scattering; the paper does not test mirror-like or fully transparent/specular contacting surfaces.
  • domain assumption Non-contact regions retain an air gap so TIR is not frustrated before true contact.
    Underlies the non-contact suppression; liquids or wetting films could prematurely frustrate TIR and create false positives. The paper does not test surface contamination or wetting over time.
  • domain assumption Internal LED rays reflect specularly at the non-contact touching surface and are confined by baffles and black coatings.
    Section III-B(b): the baffle geometry and black paint are needed to suppress residual leakage; Table II shows leakage at 3520 Lux, confirming imperfect absorption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LightTact: A Visual-Tactile Fingertip Sensor for Deformation-Independent Contact Sensing." pith.science (2026). https://pith.science/paper/3M55UN7P

@misc{pith2026251220591,
  author       = {Pith},
  title        = {Pith review of: LightTact: A Visual-Tactile Fingertip Sensor for Deformation-Independent Contact Sensing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3M55UN7P}},
  note         = {Machine review of arXiv:2512.20591}
}
read the original abstract

Contact often occurs without macroscopic surface deformation, such as during interaction with liquids, semi-liquids, or ultra-soft materials. However, most existing tactile sensors rely on deformation to infer contact, making such light-contact interactions difficult to perceive robustly. To address this, we present LightTact, a visual-tactile fingertip sensor that makes contact directly visible via a deformation-independent principle. LightTact features an ambient-blocking optical configuration that suppresses both external light and internal illumination at non-contact regions, while transmitting only the scattered light generated at true contacts. As a result, LightTact produces high-contrast raw images in which non-contact pixels remain near-black (mean gray value < 3) and contact pixels preserve the natural appearance of the contacting surface. Built on this, LightTact achieves accurate pixel-level contact segmentation that is robust to material properties, contact force, surface appearance, and environmental lighting. We further demonstrate that LightTact unlocks new robotic manipulation behaviors that require detection of extremely light contact, including water spreading, facial-cream dipping, and soft thin-film interaction. In addition, we show that LightTact's spatially aligned visual-tactile images can be directly interpreted by vision-language models.

Figures

Figures reproduced from arXiv: 2512.20591 by the authors.

Figure 1
Figure 1. LightTact provides direct, pixel-level contact sensing across liquids, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Optical layout and sensing principle of LightTact. (a) Core [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Gel fabrication for LightTact. (a) Casting of the transparent gel. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (10 more)
Figure 5
Figure 5. Figure 5: Camera calibration for LightTact. (a) LightTact captures imprints [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: LightTact achieves robust contact segmentation with objects that make extremely light contact without producing macroscopic surface deformation, [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: LightTact detects contact with water and segments the contact [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 7
Figure 7. Figure 7: LightTact reliably senses both light and firm contact from rigid [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Using LightTact, the robot can dip facial cream using a contact [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: Two LightTact sensors enable responsive interaction with an ultra [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 12
Figure 12. Figure 12: Optical analysis for the transparent medium under [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 16
Figure 16. Figure 16: LightTact sensing large-area contact on flat surfaces. Unlike [PITH_FULL_IMAGE:figures/full_fig_p012_16.png]
Figure 17
Figure 17. Figure 17: Three baseline sensors and LightTact. G. Details for Resistor Sorting In the resistor sorting task, our text prompt to VLMs is: The attached images are captured by two tactile sensors, which are mounted on the left and right side of a gripper. They are grabbing a resi…
Figure 15
Figure 15. Figure 15: Applying high-absorption black paint substantially darkens the [PITH_FULL_IMAGE:figures/full_fig_p012_15.png]

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. VibeAct: Vibration to Actions for Contact-Rich Reactive Robot Dexterity

    cs.RO 2026-06 unverdicted novelty 6.0 of 10

    VibeAct bridges real vibro-acoustic sensing and sim-based RL via a shared contact/slip representation, outperforming proprioception baselines on contact-rich dexterous tasks with successful real-world transfer.

  2. HiPi: Reproducible High-Fidelity Piezoresistive Sensors for Robotic Manipulation

    cs.RO 2026-06 unverdicted novelty 5.0 of 10

    HiPi integrates a compact readout PCB, STM32 MCU, optimized comms, and FPCB layers to deliver 220 Hz readout on 2048-taxel bimanual arrays while raising contact-geometry IoU from 0.428 to 0.797 versus a reproducible baseline.

Reference graph

Works this paper leans on

38 extracted references · 4 linked inside Pith · cited by 2 Pith papers

  1. [1]

    Tactile sensing—from humans to humanoids,

    R. S. Dahiya, G. Metta, M. Valle, and G. Sandini, “Tactile sensing—from humans to humanoids,”IEEE transactions on robotics, vol. 26, no. 1, pp. 1–20, 2009

  2. [2]

    Tactile robotics: An outlook,

    S. Luo, N. F. Lepora, W. Yuan, K. Althoefer, G. Cheng, and R. Dahiya, “Tactile robotics: An outlook,”IEEE Transactions on Robotics, 2025

  3. [3]

    Classification of vision-based tactile sensors: A review,

    H. Li, Y . Lin, C. Lu, M. Yang, E. Psomopoulou, and N. F. Lepora, “Classification of vision-based tactile sensors: A review,”IEEE Sensors Journal, 2025

  4. [4]

    When vision meets touch: A contemporary review for visuotactile sensors from the signal processing perspective,

    S. Li, Z. Wang, C. Wu, X. Li, S. Luo, B. Fang, F. Sun, X.-P. Zhang, and W. Ding, “When vision meets touch: A contemporary review for visuotactile sensors from the signal processing perspective,”IEEE Journal of Selected Topics in Signal Processing, vol. 18, no. 3, pp. 267–287, 2024

  5. [5]

    Vision-based tactile sensing: From performance parameters to device design,

    Y .-H. Xin, K.-M. Hu, R.-J. Xiang, Y .-L. Gao, J.-F. Zhou, G. Meng, and W.-M. Zhang, “Vision-based tactile sensing: From performance parameters to device design,” Applied Physics Reviews, vol. 12, no. 2, 2025

  6. [6]

    A survey of vision-based tactile sensors: Hardware, algorithm, application and future direction,

    K. He, “A survey of vision-based tactile sensors: Hardware, algorithm, application and future direction,” IEEE Transactions on Instrumentation and Measurement, 2025

  7. [7]

    Gelsight: High- resolution robot tactile sensors for estimating geometry and force,

    W. Yuan, S. Dong, and E. H. Adelson, “Gelsight: High- resolution robot tactile sensors for estimating geometry and force,”Sensors, vol. 17, no. 12, p. 2762, 2017

  8. [8]

    Combining finger vision and optical tactile sensing: Reducing and han- dling errors while cutting vegetables,

    A. Yamaguchi and C. G. Atkeson, “Combining finger vision and optical tactile sensing: Reducing and han- dling errors while cutting vegetables,” in2016 IEEE- RAS 16th International Conference on Humanoid Robots (Humanoids). IEEE, 2016, pp. 1045–1051

Show all 38 references
  1. [9]

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

    C. Lin, H. Zhang, J. Xu, L. Wu, and H. Xu, “9dtact: A compact vision-based tactile sensor for accurate 3d shape reconstruction and generalizable 6d force estimation,” IEEE Robotics and Automation Letters, vol. 9, no. 2, pp. 923–930, 2023

  2. [10]

    The tactip family: Soft optical tactile sensors with 3d- printed biomimetic morphologies,

    B. Ward-Cherrier, N. Pestell, L. Cramphorn, B. Win- stone, M. E. Giannaccini, J. Rossiter, and N. F. Lepora, “The tactip family: Soft optical tactile sensors with 3d- printed biomimetic morphologies,”Soft robotics, vol. 5, no. 2, pp. 216–227, 2018

  3. [11]

    Deltact: A vision-based tactile sensor using a dense color pattern,

    G. Zhang, Y . Du, H. Yu, and M. Y . Wang, “Deltact: A vision-based tactile sensor using a dense color pattern,” IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 10 778–10 785, 2022

  4. [12]

    Low-cost multi-touch sensing through frus- trated total internal reflection,

    J. Y . Han, “Low-cost multi-touch sensing through frus- trated total internal reflection,” inProceedings of the 18th annual ACM symposium on User interface software and technology, 2005, pp. 115–118

  5. [13]

    Robotic grasp control with high-resolution combined tactile and proximity sensing,

    K. Shimonomura, H. Nakashima, and K. Nozu, “Robotic grasp control with high-resolution combined tactile and proximity sensing,” in2016 IEEE International Confer- ence on Robotics and automation (ICRA). IEEE, 2016, pp. 138–143

  6. [14]

    Tirgel: A visuo-tactile sensor with total internal reflection mechanism for external observation and contact detection,

    S. Zhang, Y . Sun, J. Shan, Z. Chen, F. Sun, Y . Yang, and B. Fang, “Tirgel: A visuo-tactile sensor with total internal reflection mechanism for external observation and contact detection,”IEEE Robotics and Automation Letters, vol. 8, no. 10, pp. 6307–6314, 2023

  7. [15]

    Soft contact simulation and manipulation learning of deformable ob- jects with vision-based tactile sensor,

    Y . Sun, S. Zhang, Z. Chen, Z. Shen, F. Sun, C. Stefanini, D. Guo, S. Luo, J. Zhang, J. Shanet al., “Soft contact simulation and manipulation learning of deformable ob- jects with vision-based tactile sensor,”IEEE Transactions on Automation Science and Engineering, 2025

  8. [16]

    AmazingHand: Code and model to control the AH! (amazing hand),

    Pollen Robotics, “AmazingHand: Code and model to control the AH! (amazing hand),” https://github.com/ pollen-robotics/AmazingHand, 2025, gitHub repository, accessed 2025-12-14

  9. [17]

    Soft-bubble: A highly compliant dense geometry tactile sensor for robot manipulation,

    A. Alspach, K. Hashimoto, N. Kuppuswamy, and R. Tedrake, “Soft-bubble: A highly compliant dense geometry tactile sensor for robot manipulation,” in2019 2nd IEEE International Conference on Soft Robotics (RoboSoft). IEEE, 2019, pp. 597–604

  10. [18]

    In-hand object localization using a novel high-resolution visuotactile sensor,

    S. Cui, R. Wang, J. Hu, J. Wei, S. Wang, and Z. Lou, “In-hand object localization using a novel high-resolution visuotactile sensor,”IEEE Transactions on Industrial Electronics, vol. 69, no. 6, pp. 6015–6025, 2021

  11. [19]

    Tac3d: A novel vision- based tactile sensor for measuring forces distribution and estimating friction coefficient distribution,

    L. Zhang, Y . Wang, and Y . Jiang, “Tac3d: A novel vision- based tactile sensor for measuring forces distribution and estimating friction coefficient distribution,”arXiv preprint arXiv:2202.06211, 2022

  12. [20]

    High-resolution tac- tile sensor using the deformation of a reflection image,

    S. Saga, H. Kajimoto, and S. Tachi, “High-resolution tac- tile sensor using the deformation of a reflection image,” Sensor Review, vol. 27, no. 1, pp. 35–42, 2007

  13. [21]

    Precise shape re- construction by active pattern in total-internal-reflection- based tactile sensor,

    S. Saga, R. Taira, and K. Deguchi, “Precise shape re- construction by active pattern in total-internal-reflection- based tactile sensor,”IEEE Transactions on Haptics, vol. 7, no. 1, pp. 67–77, 2013

  14. [22]

    Method for optical comparison of skin friction-ridge patterns,

    W. Wyman, “Method for optical comparison of skin friction-ridge patterns,” Aug. 17 1965, uS Patent 3,200,701

  15. [23]

    The drawing prism: a versatile graphic input device,

    R. Greene, “The drawing prism: a versatile graphic input device,” inProceedings of the 12th annual conference on Computer graphics and interactive techniques, 1985, pp. 103–110

  16. [24]

    Finger-sts: Combined prox- imity and tactile sensing for robotic manipulation,

    F. R. Hogan, J.-F. Tremblay, B. H. Baghi, M. Jenkin, K. Siddiqi, and G. Dudek, “Finger-sts: Combined prox- imity and tactile sensing for robotic manipulation,”IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 10 865–10 872, 2022

  17. [25]

    Spectac: A visual- tactile dual-modality sensor using uv illumination,

    Q. Wang, Y . Du, and M. Y . Wang, “Spectac: A visual- tactile dual-modality sensor using uv illumination,” in 2022 international conference on robotics and automa- tion (ICRA). IEEE, 2022, pp. 10 844–10 850

  18. [26]

    Vitactip: Design and verification of a novel biomimetic physical vision-tactile fusion sensor,

    W. Fan, H. Li, W. Si, S. Luo, N. Lepora, and D. Zhang, “Vitactip: Design and verification of a novel biomimetic physical vision-tactile fusion sensor,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 1056–1062

  19. [27]

    Compdvision: Combining near-field 3d visual and tactile sensing using a compact compound-eye imaging system,

    L. Luo, B. Zhang, Z. Peng, Y . K. Cheung, G. Zhang, Z. Li, M. Y . Wang, and H. Yu, “Compdvision: Combining near-field 3d visual and tactile sensing using a compact compound-eye imaging system,” in2024 IEEE/RSJ In- ternational Conference on Intelligent Robots and Systems (IROS)...

  20. [28]

    Vi2tap: A cross-polarization based mechanism for perception transition in tactile-proximity sensing with applications to soft grippers,

    N. H. Nguyen, N. M. D. Le, Q. K. Luu, T. T. Nguyen, and V . A. Ho, “Vi2tap: A cross-polarization based mechanism for perception transition in tactile-proximity sensing with applications to soft grippers,”IEEE Robotics and Au- tomation Letters, 2025

  21. [29]

    Simultaneous tactile-visual perception for learning multimodal robot manipulation,

    Y . Li, Y . Chen, Z. Zhao, P. Li, T. Liu, S. Huang, and Y . Zhu, “Simultaneous tactile-visual perception for learning multimodal robot manipulation,”arXiv preprint arXiv:2512.09851, 2025

  22. [30]

    Seeing through your skin: Recognizing objects with a novel visuotactile sensor,

    F. R. Hogan, M. Jenkin, S. Rezaei-Shoshtari, Y . Girdhar, D. Meger, and G. Dudek, “Seeing through your skin: Recognizing objects with a novel visuotactile sensor,” inProceedings of the IEEE/CVF winter conference on applications of computer vision, 2021, pp. 1218–1227

  23. [31]

    Stereotac: A novel visuotactile sensor that combines tactile sensing with 3d vision,

    E. Roberge, G. Fornes, and J.-P. Roberge, “Stereotac: A novel visuotactile sensor that combines tactile sensing with 3d vision,”IEEE Robotics and Automation Letters, vol. 8, no. 10, pp. 6291–6298, 2023

  24. [32]

    Vistac toward a unified multimodal sensing finger for robotic manipulation,

    S. Athar, G. Patel, Z. Xu, Q. Qiu, and Y . She, “Vistac toward a unified multimodal sensing finger for robotic manipulation,”IEEE Sensors Journal, vol. 23, no. 20, pp. 25 440–25 450, 2023

  25. [33]

    (2022) Gelsight-mini

    GelSight. (2022) Gelsight-mini. [Online]. Available: https://www.gelsight.com/gelsightmini/

  26. [34]

    Physically grounded vision-language models for robotic manipulation,

    J. Gao, B. Sarkar, F. Xia, T. Xiao, J. Wu, B. Ichter, A. Majumdar, and D. Sadigh, “Physically grounded vision-language models for robotic manipulation,” in 2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 12 462–12 469

  27. [35]

    Gemini robotics: Bringing ai into the physical world,

    G. R. Team, S. Abeyruwan, J. Ainslie, J.-B. Alayrac, M. G. Arenas, T. Armstrong, A. Balakrishna, R. Baruch, M. Bauza, M. Blokzijlet al., “Gemini robotics: Bringing ai into the physical world,”arXiv preprint arXiv:2503.20020, 2025

  28. [36]

    Demonstrating the octopi-1.5 visual-tactile-language model,

    S. Yu, K. Lin, and H. Soh, “Demonstrating the octopi-1.5 visual-tactile-language model,”arXiv preprint arXiv:2507.09985, 2025

  29. [37]

    Tactile-vla: Unlocking vision-language-action model’s physical knowledge for tactile generalization,

    J. Huang, S. Wang, F. Lin, Y . Hu, C. Wen, and Y . Gao, “Tactile-vla: Unlocking vision-language-action model’s physical knowledge for tactile generalization,”arXiv preprint arXiv:2507.09160, 2025

  30. [38]

    Omnivtla: Vision-tactile-language-action model with semantic-aligned tactile sensing,

    Z. Cheng, Y . Zhang, W. Zhang, H. Li, K. Wang, L. Song, and H. Zhang, “Omnivtla: Vision-tactile-language-action model with semantic-aligned tactile sensing,”arXiv preprint arXiv:2508.08706, 2025. APPENDIX A. Camera Placement underθ tv <2θ c In this section, we discuss our moti...

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.