arxiv: 2604.20295 · v1 · submitted 2026-04-22 · 💻 cs.RO

Recognition: unknown

ETac: A Lightweight and Efficient Tactile Simulation Framework for Learning Dexterous Manipulation

Zhe Xu , Feiyu Zhao , Xiyan Huang , Chenxi Xiao

Authors on Pith no claims yet

Pith reviewed 2026-05-10 00:39 UTC · model grok-4.3

classification 💻 cs.RO

keywords tactile simulationdexterous manipulationreinforcement learningsoft-body modelingelastomeric sensorsparallel simulationgrasping policy

0 comments

The pith

ETac models elastomeric tactile sensor deformations with a lightweight data-driven approach that matches finite element accuracy while running fast enough for large-scale reinforcement learning.

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

The paper introduces ETac as a simulation framework for tactile sensors in robotic hands. It aims to resolve the trade-off between simulation fidelity and speed that has limited learning of tactile-based manipulation skills. By using a data-driven model for deformation propagation instead of full physics simulations, ETac achieves deformation estimates similar to FEM methods but at much higher efficiency. This allows training reinforcement learning policies across thousands of parallel environments on a single GPU, leading to successful grasping policies for diverse objects. A sympathetic reader would care because it makes tactile feedback practical for scalable robot learning in contact-rich tasks.

Core claim

ETac employs a lightweight data-driven deformation propagation model to capture soft-body contact dynamics. When used as a simulation backend, it produces surface deformation estimates comparable to FEM and applies to modeling real tactile sensors. This enables training a blind grasping policy that uses large-area tactile feedback, achieving 84.45% average success rate across four object types at a throughput of 869 FPS over 4096 parallel environments on a single RTX 4090 GPU.

What carries the argument

The lightweight data-driven deformation propagation model, which approximates elastomeric soft-body interactions to enable efficient contact dynamics simulation.

If this is right

It supports reinforcement learning training at scale for tactile-dependent manipulation skills.
It demonstrates applicability for modeling real tactile sensors by matching FEM surface deformations.
The framework achieves high efficiency with 869 FPS across 4096 environments while maintaining simulation quality.
Resulting policies can manipulate diverse objects using blind grasping with tactile feedback.

Where Pith is reading between the lines

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

The approach could extend to other soft-body simulation needs in robotics beyond tactile sensors.
If the model generalizes well, it might reduce the need for expensive FEM computations in real-time robotic planning.
Testing on a wider variety of object shapes and materials would reveal the limits of the data-driven approximation.

Load-bearing premise

The data-driven deformation model accurately generalizes from training data to real-world tactile sensors and new objects without significant errors in contact forces or deformations.

What would settle it

Compare the simulated surface deformations from ETac against actual measurements from a physical tactile sensor on objects not used in training; if the error exceeds that of FEM or leads to policy failure in real transfer, the claim does not hold.

Figures

Figures reproduced from arXiv: 2604.20295 by Chenxi Xiao, Feiyu Zhao, Xiyan Huang, Zhe Xu.

**Figure 1.** Figure 1: ETac: a lightweight and efficient tactile simulation framework. (a) ETac estimates elastomer surface deformations with fidelity comparable to that of FEM. (b) It enables large-area tactile sensing for dexterous manipulators while supporting large-scale RL training. Currently, a common choice for state-of-the-art tactile simulators is based on soft-body physics engines. For instance, simulators built on t… view at source ↗

**Figure 2.** Figure 2: Overview of the ETac pipeline. The manipulator’s surface mesh is discretized into nodes, whose contact dynamics are learned by a hybrid deformation propagation model. We calibrate the propagation model using data from FEM simulation to ensure fidelity. The computed displacement field of the nodes serves as the tactile input to the RL policy during skill learning. method first discretizes the manipulator’s … view at source ↗

**Figure 3.** Figure 3: Computational process of the propagation model. Displacements of passive nodes (nodes not in direct contact) are estimated using a decay-based model combined with a residual correction network. features using shared MLPs and aggregates them via MaxPooling into a 64-dimensional global feature vector. This global feature is further encoded by another MLP, then broadcast and concatenated with per-node featur… view at source ↗

**Figure 4.** Figure 4: Comparison of elastomer surface displacement fields estimated by different methods. Arrows show node-wise displacement direction and magnitude, with color shifting to red as z-axis displacement increases. TABLE II: The average RMSE (mm) between estimated displacements and the FEM simulation results. Flat Elastomer Curved Elastomer TacSL [19] 0.194 ± 0.076 0.445 ± 0.079 Taxim [18] 0.163 ± 0.048 0.447 ± 0.35… view at source ↗

**Figure 5.** Figure 5: Predicting signals of real sensors. (a) Paired data collection in real and simulated settings. (b) Comparison of sensor output predictions using ETac and FEM data. “Unseen Trajectory” denotes novel loading patterns with seen indenters. of the full model with better physical prior knowledge for modeling attenuation. Their combination achieves the best performance in deformation estimation. Real-to-Sim Tacti… view at source ↗

**Figure 6.** Figure 6: Demonstrations of Blind Grasping Skill. The process of successfully grasping 4 different objects over time. TABLE III: Success rates (%) of blind grasping policies under different settings. Cube Egg Rock Sodacan Average Baseline (Object Pose) 63.79 ± 36.25 72.05 ± 30.22 47.66 ± 43.63 68.70 ± 12.12 62.97 ± 37.82 Fingertip Sensors 74.35 ± 5.24 86.46 ± 10.57 70.36 ± 15.30 60.43 ± 3.01 72.90 ± 21.06 Full-hand … view at source ↗

**Figure 7.** Figure 7: Comparison of RL training performance. achieved, ETac scales efficiently with more environments: reaching 669, 956, 878, and 869 total FPS at 64, 256, 1024, and 4096 environments, respectively. This outperforms Taxim (508, 620, 650 FPS at 64, 256, 1024 environments, but outof-memory when scale up to 4096 envs) and is comparable to TacSL (698, 975, 918, 886 FPS at 64, 256, 1024, 4096 envs). ETac’s memory u… view at source ↗

**Figure 8.** Figure 8: Network used for sensor output prediction. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗

read the original abstract

Tactile sensors are increasingly integrated into dexterous robotic manipulators to enhance contact perception. However, learning manipulation policies that rely on tactile sensing remains challenging, primarily due to the trade-off between fidelity and computational cost of soft-body simulations. To address this, we present ETac, a tactile simulation framework that models elastomeric soft-body interactions with both high fidelity and efficiency. ETac employs a lightweight data-driven deformation propagation model to capture soft-body contact dynamics, achieving high simulation quality and boosting efficiency that enables large-scale policy training. When serving as the simulation backend, ETac produces surface deformation estimates comparable to FEM and demonstrates applicability for modeling real tactile sensors. Then, we showcase its capability in training a blind grasping policy that leverages large-area tactile feedback to manipulate diverse objects. Running on a single RTX 4090 GPU, ETac supports reinforcement learning across 4,096 parallel environments, achieving a total throughput of 869 FPS. The resulting policy reaches an average success rate of 84.45% across four object types, underscoring ETac's potential to make tactile-based skill learning both efficient and scalable.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note private letter to a colleague

ETac gives a fast data-driven tactile sim that runs RL at scale, but the FEM comparability claim rests on thin validation details.

read the letter

The core of this paper is a lightweight data-driven model for propagating deformations in soft elastomeric tactile sensors. It positions itself as a practical alternative to full FEM solvers, and the authors show it can drive parallel RL training for a blind grasping task that reaches 84.45% success across four objects while delivering 869 FPS total throughput on one RTX 4090 with 4096 environments. That efficiency number is concrete and directly useful for anyone trying to scale tactile-based policies without vision. The framework also claims to model real sensors, which opens a path for sim-to-real work in contact-rich manipulation. What the paper does cleanly is demonstrate the end-to-end payoff: the speed gain lets them run large-scale training that would be impractical with heavier simulators. The soft spot is the accuracy side. The abstract states the deformations are comparable to FEM and applicable to real sensors, yet it gives no error metrics, no breakdown of training data sources, and no tests on unseen geometries or multi-contact cases. Without those numbers it is hard to know how much the reported success rate depends on simulation-specific artifacts versus transferable tactile information. The stress-test concern about generalization therefore lands. This work is aimed at robotics researchers who need faster soft-body simulation for RL loops in dexterous or vision-denied settings. It deserves a serious referee because the efficiency results are specific and the application is relevant, even though the deformation validation will need tightening in revision.

Referee Report

2 major / 1 minor

Summary. The paper presents ETac, a lightweight data-driven tactile simulation framework for modeling elastomeric soft-body contact dynamics. It claims that a propagation model yields surface deformations comparable to FEM while enabling high-throughput RL (869 FPS across 4096 parallel environments on a single RTX 4090), and demonstrates this via a blind grasping policy achieving 84.45% average success on four object types.

Significance. If the fidelity claims hold with quantified validation, ETac could address a key bottleneck in tactile RL by trading off simulation cost and accuracy, supporting large-scale policy training for contact-rich dexterous tasks. The reported parallel throughput and RL demonstration are concrete strengths that would be valuable if the underlying deformation model generalizes.

major comments (2)

[Abstract] Abstract: the assertion that ETac 'produces surface deformation estimates comparable to FEM' lacks any quantitative support such as mean deformation error, contact force RMSE, or baseline comparisons; no validation dataset size, object geometries, or test conditions are specified. This directly undermines the central fidelity claim required for the RL results to be interpretable.
[Experimental results] Experimental results (implied in the RL demonstration paragraph): the data-driven deformation propagation model is described as trained on interactions but provides no details on training data source, loss function, network architecture, or explicit tests for generalization to unseen curvatures, materials, or multi-contact scenarios. Without these, the 84.45% success rate cannot be distinguished from simulation-specific artifacts.

minor comments (1)

[Abstract] The efficiency numbers (869 FPS, 4096 environments) are presented without breakdown of per-environment cost or comparison to alternative simulators; adding a table with these metrics would strengthen the efficiency claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will revise the paper to strengthen the presentation of quantitative validation and model training details.

read point-by-point responses

Referee: [Abstract] Abstract: the assertion that ETac 'produces surface deformation estimates comparable to FEM' lacks any quantitative support such as mean deformation error, contact force RMSE, or baseline comparisons; no validation dataset size, object geometries, or test conditions are specified. This directly undermines the central fidelity claim required for the RL results to be interpretable.

Authors: We agree that the abstract would benefit from explicit quantitative support. We will revise the abstract to incorporate key metrics from our validation experiments, including mean surface deformation error, contact force RMSE, baseline comparisons to FEM, validation dataset size, object geometries tested, and test conditions. These additions will make the fidelity claim more precise and directly support the interpretability of the RL results. revision: yes
Referee: [Experimental results] Experimental results (implied in the RL demonstration paragraph): the data-driven deformation propagation model is described as trained on interactions but provides no details on training data source, loss function, network architecture, or explicit tests for generalization to unseen curvatures, materials, or multi-contact scenarios. Without these, the 84.45% success rate cannot be distinguished from simulation-specific artifacts.

Authors: We acknowledge the need for greater transparency on the data-driven model. We will add a dedicated subsection detailing the training data source, loss function, network architecture of the propagation model, and explicit generalization tests across unseen curvatures, materials, and multi-contact scenarios. This will allow readers to better evaluate the 84.45% success rate and confirm it is not due to simulation artifacts. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper describes ETac as a data-driven deformation propagation model trained to approximate soft-body contact dynamics, with empirical claims of FEM-comparable surface estimates and downstream RL success rates (84.45% across objects at 869 FPS). No load-bearing step reduces by construction to its own inputs: the model is not self-defined via the target quantities, no fitted parameters are relabeled as independent predictions, and no uniqueness theorems or ansatzes are imported via self-citation chains. The derivation remains self-contained against external benchmarks (FEM comparisons and real-sensor applicability), with the RL throughput and success metrics arising from parallel execution rather than tautological reuse of training data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claim rests on the unstated details of how the data-driven deformation model is trained and validated; no free parameters, axioms, or invented entities are specified in the abstract.

pith-pipeline@v0.9.0 · 5502 in / 1050 out tokens · 22903 ms · 2026-05-10T00:39:03.331838+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

38 extracted references · 16 canonical work pages · 2 internal anchors

[1]

Learning to grasp with- out seeing,

A. Murali, Y . Li, D. Gandhi, and A. Gupta, “Learning to grasp with- out seeing,” inInternational Symposium on Experimental Robotics, pp. 375–386, Springer, 2018

2018
[2]

Proprioceptive state estimation for amphibious tactile sensing,

N. Guo, X. Han, S. Zhong, Z. Zhou, J. Lin, J. S. Dai, F. Wan, and C. Song, “Proprioceptive state estimation for amphibious tactile sensing,”IEEE Transactions on Robotics, 2024

2024
[3]

Grasping in the dark: Compliant grasping using shadow dexterous hand and biotac tactile sensor,

K. Ganguly, B. Sadrfaridpour, K. B. Kidambi, C. Fermller, and Y . Aloi- monos, “Grasping in the dark: Compliant grasping using shadow dexterous hand and biotac tactile sensor,”arXiv e-prints, pages arXiv– 2011, 2020

2011
[4]

Embedding high-resolution touch across robotic hands enables adaptive human-like grasping,

Z. Zhao, W. Li, Y . Li, T. Liu, B. Li, M. Wang, K. Du, H. Liu, Y . Zhu, Q. Wang,et al., “Embedding high-resolution touch across robotic hands enables adaptive human-like grasping,”arXiv preprint arXiv:2412.14482, 2024

work page arXiv 2024
[5]

Robotsweater: Scalable, generalizable, and customizable machine-knitted tactile skins for robots,

Z. Si, T. C. Yu, K. Morozov, J. McCann, and W. Yuan, “Robotsweater: Scalable, generalizable, and customizable machine-knitted tactile skins for robots,” in2023 IEEE International Conference on Robotics and Automation (ICRA), pp. 10352–10358, IEEE, 2023

2023
[6]

Social gesture recognition in sphri: Leveraging fabric-based tactile sensing on humanoid robots,

D. Crowder, K. Vandyck, X. Sun, J. McCann, and W. Yuan, “Social gesture recognition in sphri: Leveraging fabric-based tactile sensing on humanoid robots,”arXiv preprint arXiv:2503.03234, 2025

work page arXiv 2025
[7]

A tactile sensing foot for single robot leg stabilization,

G. Zhang, Y . Du, Y . Zhang, and M. Y . Wang, “A tactile sensing foot for single robot leg stabilization,” in2021 IEEE International Conference on Robotics and Automation (ICRA), pp. 14076–14082, IEEE, 2021

2021
[8]

Learning in-hand translation using tactile skin with shear and normal force sensing,

J. Yin, H. Qi, J. Malik, J. Pikul, M. Yim, and T. Hellebrekers, “Learning in-hand translation using tactile skin with shear and normal force sensing,”arXiv preprint arXiv:2407.07885, 2024

work page arXiv 2024
[9]

Tactile sensing in intelligent robotic manipulation–a review,

J. Tegin and J. Wikander, “Tactile sensing in intelligent robotic manipulation–a review,”Industrial Robot: An International Journal, vol. 32, no. 1, pp. 64–70, 2005

2005
[10]

Recent progress in tactile sensing and sensors for robotic manipulation: can we turn tactile sensing into vision?,

A. Yamaguchi and C. G. Atkeson, “Recent progress in tactile sensing and sensors for robotic manipulation: can we turn tactile sensing into vision?,”Advanced Robotics, vol. 33, no. 14, pp. 661–673, 2019

2019
[11]

Lessons from learning to spin" pens".arXiv preprint arXiv:2407.18902, 2024

J. Wang, Y . Yuan, H. Che, H. Qi, Y . Ma, J. Malik, and X. Wang, “Lessons from learning to spin” pens”,”arXiv preprint arXiv:2407.18902, 2024

work page arXiv 2024
[12]

Dextouch: Learning to seek and manipulate objects with tactile dexterity,

K.-W. Lee, Y . Qin, X. Wang, and S.-C. Lim, “Dextouch: Learning to seek and manipulate objects with tactile dexterity,”arXiv preprint arXiv:2401.12496, 2024

work page arXiv 2024
[13]

Affine body dynamics: Fast, stable & intersection-free simulation of stiff materials,

L. Lan, D. M. Kaufman, M. Li, C. Jiang, and Y . Yang, “Affine body dynamics: Fast, stable & intersection-free simulation of stiff materials,” arXiv preprint arXiv:2201.10022, 2022

work page arXiv 2022
[14]

Difftactile: A physics-based differentiable tactile simulator for contact-rich robotic manipulation,

Z. Si, G. Zhang, Q. Ben, B. Romero, Z. Xian, C. Liu, and C. Gan, “Difftactile: A physics-based differentiable tactile simulator for contact-rich robotic manipulation,”arXiv preprint arXiv:2403.08716, 2024

work page arXiv 2024
[15]

Efficient tactile simulation with differentiability for robotic manipulation,

J. Xu, S. Kim, T. Chen, A. R. Garcia, P. Agrawal, W. Matusik, and S. Sueda, “Efficient tactile simulation with differentiability for robotic manipulation,” inConference on Robot Learning, pp. 1488– 1498, PMLR, 2023

2023
[16]

Learning to walk in minutes using massively parallel deep reinforcement learning,

N. Rudin, D. Hoeller, P. Reist, and M. Hutter, “Learning to walk in minutes using massively parallel deep reinforcement learning,” in Conference on Robot Learning, pp. 91–100, PMLR, 2022

2022
[17]

Tacto: A fast, flexible, and open-source simulator for high-resolution vision-based tactile sensors,

S. Wang, M. Lambeta, P.-W. Chou, and R. Calandra, “Tacto: A fast, flexible, and open-source simulator for high-resolution vision-based tactile sensors,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 3930–3937, 2022

2022
[18]

Taxim: An example-based simulation model for gelsight tactile sensors,

Z. Si and W. Yuan, “Taxim: An example-based simulation model for gelsight tactile sensors,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 2361–2368, 2022

2022
[19]

TacSL: A library for visuotactile sensor simu- lation and learning

I. Akinola, J. Xu, J. Carius, D. Fox, and Y . Narang, “Tacsl: A library for visuotactile sensor simulation and learning,”arXiv preprint arXiv:2408.06506, 2024

work page arXiv 2024
[20]

Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning

V . Makoviychuk, L. Wawrzyniak, Y . Guo, M. Lu, K. Storey, M. Mack- lin, D. Hoeller, N. Rudin, A. Allshire, A. Handa,et al., “Isaac gym: High performance gpu-based physics simulation for robot learning,” arXiv preprint arXiv:2108.10470, 2021

work page internal anchor Pith review arXiv 2021
[21]

Gelsight svelte hand: A three-finger, two-dof, tactile-rich, low-cost robot hand for dexterous manipulation,

J. Zhao and E. H. Adelson, “Gelsight svelte hand: A three-finger, two-dof, tactile-rich, low-cost robot hand for dexterous manipulation,” ArXiv, vol. abs/2309.10886, 2023

work page arXiv 2023
[22]

Eyesight hand: Design of a fully-actuated dexterous robot hand with inte- grated vision-based tactile sensors and compliant actuation,

B. Romero, H.-S. Fang, P. Agrawal, and E. H. Adelson, “Eyesight hand: Design of a fully-actuated dexterous robot hand with inte- grated vision-based tactile sensors and compliant actuation,”ArXiv, vol. abs/2408.06265, 2024

work page arXiv 2024
[23]

Rotating without seeing: Towards in-hand dexterity through touch,

Z.-H. Yin, B. Huang, Y . Qin, Q. Chen, and X. Wang, “Rotating without seeing: Towards in-hand dexterity through touch,”arXiv preprint arXiv:2303.10880, 2023

work page arXiv 2023
[24]

Robot synesthesia: In-hand manipulation with visuotactile sensing,

Y . Yuan, H. Che, Y . Qin, B. Huang, Z.-H. Yin, K.-W. Lee, Y . Wu, S.-C. Lim, and X. Wang, “Robot synesthesia: In-hand manipulation with visuotactile sensing,” in2024 IEEE International Conference on Robotics and Automation (ICRA), pp. 6558–6565, IEEE, 2024

2024
[25]

Sequential dexterity: Chaining dexterous policies for long-horizon manipulation,

Y . Chen, C. Wang, L. Fei-Fei, and C. K. Liu, “Sequential dexterity: Chaining dexterous policies for long-horizon manipulation,” 2023

2023
[26]

Taccel: Scaling up vision-based tactile robotics via high- performance gpu simulation,

Y . Li, W. Du, C. Yu, P. Li, Z. Zhao, T. Liu, C. Jiang, Y . Zhu, and S. Huang, “Taccel: Scaling up vision-based tactile robotics via high- performance gpu simulation,”arXiv preprint arXiv:2504.12908, 2025

work page arXiv 2025
[27]

A moving least squares material point method with displacement discontinuity and two-way rigid body coupling,

Y . Hu, Y . Fang, Z. Ge, Z. Qu, Y . Zhu, A. Pradhana, and C. Jiang, “A moving least squares material point method with displacement discontinuity and two-way rigid body coupling,”ACM Transactions on Graphics (TOG), vol. 37, no. 4, pp. 1–14, 2018

2018
[28]

Chainqueen: A real-time differentiable physical simulator for soft robotics,

Y . Hu, J. Liu, A. Spielberg, J. B. Tenenbaum, W. T. Freeman, J. Wu, D. Rus, and W. Matusik, “Chainqueen: A real-time differentiable physical simulator for soft robotics,” in2019 International conference on robotics and automation (ICRA), pp. 6265–6271, IEEE, 2019

2019
[29]

Kaolin: A pytorch library for accelerating 3d deep learn- ing research

C. Fuji Tsang, M. Shugrina, J. F. Lafleche, T. Takikawa, J. Wang, and C. e. a. Loop, “Kaolin: A pytorch library for accelerating 3d deep learn- ing research.” https://github.com/NVIDIAGameWorks/kaolin, 2022

2022
[30]

Interaction of elastic bodies via surface forces: 2. exponential decay,

O. I. Vinogradova and F. Feuillebois, “Interaction of elastic bodies via surface forces: 2. exponential decay,”Journal of colloid and interface science, vol. 268, no. 2, pp. 464–475, 2003

2003
[31]

Pointnet: Deep learning on point sets for 3d classification and segmentation,

C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” inProceedings of the IEEE conference on computer vision and pattern recognition, pp. 652–660, 2017

2017
[32]

Sim-to-real for robotic tactile sensing via physics-based simulation and learned latent projections,

Y . Narang, B. Sundaralingam, M. Macklin, A. Mousavian, and D. Fox, “Sim-to-real for robotic tactile sensing via physics-based simulation and learned latent projections,” in2021 IEEE International Conference on Robotics and Automation (ICRA), pp. 6444–6451, IEEE, 2021

2021
[33]

Transferring tactile data across sensors,

W. Z. E. Amri, M. Kuhlmann, and N. Navarro-Guerrero, “Transferring tactile data across sensors,”arXiv preprint arXiv:2410.14310, 2024

work page arXiv 2024
[34]

Shadowrobot

“Shadowrobot.” https://www.shadowrobot.com/ dexterous-hand-series/
[35]

Proximal Policy Optimization Algorithms

J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,”arXiv preprint arXiv:1707.06347, 2017

work page internal anchor Pith review Pith/arXiv arXiv 2017
[36]

Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy,

Y . Xu, W. Wan, J. Zhang, H. Liu, Z. Shan, H. Shen, R. Wang, H. Geng, Y . Weng, J. Chen,et al., “Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 4737–4746, 2023

2023
[37]

Twintac: A wide-range, highly sensitive tactile sen- sor with real-to-sim digital twin sensor model,

X. H. et al., “Twintac: A wide-range, highly sensitive tactile sen- sor with real-to-sim digital twin sensor model,”arXiv preprint arXiv:2509.10063, 2025

work page arXiv 2025
[38]

Attpnet: Attention- based deep neural network for 3d point set analysis,

Y . Yang, Y . Ma, J. Zhang, X. Gao, and M. Xu, “Attpnet: Attention- based deep neural network for 3d point set analysis,”Sensors, vol. 20, no. 19, p. 5455, 2020

2020