arxiv: 2604.05531 · v1 · submitted 2026-04-07 · 💻 cs.RO

Simulation-Driven Evolutionary Motion Parameterization for Contact-Rich Granular Scooping with a Soft Conical Robotic Hand

Yongliang Wang , Cristian C. Beltran-Hernandez , Tomoya Takahashi , Masashi Hamaya This is my paper

Pith reviewed 2026-05-10 19:59 UTC · model grok-4.3

classification 💻 cs.RO

keywords soft roboticsscoopingevolutionary optimizationphysics simulationgranular materialstrajectory optimizationdeformable hands

0 comments p. Extension

The pith

A physics-based simulation of a soft conical hand combined with evolutionary optimization enables reliable scooping of granular materials.

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

The paper develops a simulation model to capture how a soft conical robotic hand passively changes shape from flat to conical during scooping contact. It then uses an evolutionary strategy to automatically find good motion trajectories for the robot arm without needing manual adjustments. This addresses the control difficulties caused by the hand's flexibility. The approach is tested in both simulation and on physical robots, showing it can handle varied scooping conditions that prior methods struggled with.

Core claim

We present a physics-based simulation approach that accurately models the soft tool's morphing behavior from flat sheets to adaptive conical structures. Combined with an evolutionary strategy framework, this enables automatic optimization of scooping trajectories. Validation in simulation and real-robot experiments demonstrates strong generalization across challenging granular scooping tasks.

What carries the argument

The physics-based simulation model of the deformable soft conical robotic hand's passive reconfiguration, paired with an evolutionary strategy for trajectory optimization.

If this is right

The optimized trajectories transfer effectively to real robots.
It handles a range of challenging scooping tasks beyond previous methods.
No manual tuning of parameters is required for different conditions.
The method improves scooping efficiency by adapting to container surfaces.

Where Pith is reading between the lines

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

This suggests soft robotic tools can be controlled more practically in variable environments by relying on simulation rather than sensing.
Similar simulation-driven optimization might apply to other soft robot manipulation tasks involving contact with granular or deformable materials.

Load-bearing premise

The physics-based simulation accurately captures the passive reconfiguration dynamics and morphing behavior of the soft conical hand from flat sheets to adaptive structures during contact-rich scooping.

What would settle it

An experiment showing that the real soft hand's deformation during scooping deviates significantly from the simulation predictions, leading to unsuccessful trajectory optimization in practice.

Figures

Figures reproduced from arXiv: 2604.05531 by Cristian C. Beltran-Hernandez, Masashi Hamaya, Tomoya Takahashi, Yongliang Wang.

**Figure 1.** Figure 1: Contact-rich granular scooping system: The framework couples real-world execution (left) with its digital twin in simulation (right). From RGB-D perception and feature extraction, container geometry is abstracted into seed trajectories. In simulation, the covariance matrix adaptation evolution strategy (CMA-ES) optimizes both trajectory and hand roll angle, producing the best solution. This optimized strat… view at source ↗

**Figure 2.** Figure 2: The real soft-hand [13] rolling from 20◦ to 120◦ in 20◦ increments. differentiable methods have been developed for efficient gelbased surface tactile simulation [30]. Origami simulators based on rigid-panel models [31], [32] capture folding well and inspire origami/kirigami robotic tools, but mainly serve for visualization. Our work instead integrates deformable tool simulation into a contact-rich scoopin… view at source ↗

**Figure 3.** Figure 3: The top row shows, from left to right, the Model View, Wireframe [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 4.** Figure 4: Overall framework of our system: from RGB–D perception and ring–compass abstraction, through seed trajectory generation and parameterization, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 5.** Figure 5: From left to right: anchor skeleton, parabolic fit in uv-plane, and [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗

**Figure 7.** Figure 7: Generation and refinement of scooping trajectories. From left to [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗

**Figure 8.** Figure 8: Containers used in experiments: 12 objects in three sizes (large, medium, small), including regular and irregular shapes. (Width x Height) TABLE I SCOOPING RESULTS UNDER DIFFERENT ACTION PARAMETERIZATIONS ACROSS 3 CONTAINER SIZES IN THE 10 BALLS TASK. Action Parameterization Parameters Scooped Big Medium Small Initialization (k, a, b, ∆angle) 0 0 0 No k1 (k2, a1, b1, ∆angle) 2 1 1 No k2 (k1, a1, b1, ∆angle… view at source ↗

**Figure 9.** Figure 9: Action sequence of the SH executing the optimized trajectory transferred from simulation to reality. The figure illustrates key phases of the [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗

read the original abstract

Tool-based scooping is vital in robot-assisted tasks, enabling interaction with objects of varying sizes, shapes, and material states. Recent studies have shown that flexible, reconfigurable soft robotic end-effectors can adapt their shape to maintain consistent contact with container surfaces during scooping, improving efficiency compared to rigid tools. These soft tools can adjust to varying container sizes and materials without requiring complex sensing or control. However, the inherent compliance and complex deformation behavior of soft robotics introduce significant control complexity that limits practical applications. To address this challenge, this paper presents the development of a physics-based simulation model of a deformable soft conical robotic hand that captures its passive reconfiguration dynamics and enables systematic trajectory optimization for scooping tasks. We propose a novel physics-based simulation approach that accurately models the soft tool's morphing behavior from flat sheets to adaptive conical structures, combined with an evolutionary strategy framework that automatically optimizes scooping trajectories without manual parameter tuning. We validate the optimized trajectories through both simulation and real-robot experiments. The results demonstrate strong generalization and successfully address a range of challenging tasks previously beyond the reach of existing approaches. Videos of our experiments are available online: https://sites.google.com/view/scoopsh

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

The paper gives a workable sim-to-real pipeline that uses a physics model of passive soft-hand morphing plus evolutionary optimization to tune scooping trajectories, but the validation rests on thin quantitative evidence for how well the simulator matches reality.

read the letter

The main takeaway is that they built a physics-based simulator for a soft conical hand that starts as flat sheets and passively reconfigures under contact, then used an evolutionary strategy to optimize the scooping motion parameters inside that simulator before transferring the trajectories to hardware. This specific combination for granular scooping looks like a fresh application rather than a direct repeat of prior sim-plus-evolution work in robotics. They also ran real-robot tests and point to videos, which is the right move for a claim about practical utility in contact-rich tasks. That part is solid and gives the work something concrete to stand on. The soft spot is the sim-to-real gap. The whole argument for generalization and handling tasks beyond existing methods depends on the simulator correctly reproducing the hand's deformation, contact patches, and force behavior. The abstract and stress-test note both flag the absence of numbers on cone angle, curvature, or deformation error between sim and hardware, so it is difficult to judge whether the optimized trajectories are robust or simply benefited from the real hand's compliance. Minor issues like missing details on friction modeling or wrinkling would be easy to fix, but they matter here because the optimization never sees the real world. This is aimed at researchers doing soft manipulation in granular or unstructured settings who want an automated way to tune motions instead of manual iteration. A reader working on similar end-effectors would find the experimental setup and the evolutionary framework worth looking at, even if the metrics need expansion. It deserves peer review because the method is clearly described, the hardware results exist, and the core idea is falsifiable once the sim-reality numbers are added.

Referee Report

2 major / 1 minor

Summary. The paper develops a physics-based simulation model of a soft conical robotic hand to capture its passive reconfiguration from flat sheets to adaptive conical structures during contact-rich scooping. It combines this with an evolutionary strategy to optimize scooping trajectories without manual tuning, then validates the resulting motions in both simulation and real-robot experiments, claiming strong generalization across challenging granular scooping tasks that exceed prior methods.

Significance. If the simulation fidelity for passive morphing and contact dynamics holds and the optimized trajectories transfer robustly, the work offers a systematic, simulation-driven route to parameterizing compliant soft-robot behaviors in manipulation. This could reduce reliance on complex sensing or hand-tuning for reconfigurable end-effectors and extend to other contact-rich tasks. The evolutionary optimization inside an independent physics simulator is a methodological strength that supports reproducibility.

major comments (2)

[Abstract] Abstract: the central claim that optimized trajectories demonstrate 'strong generalization' and address 'a range of challenging tasks previously beyond the reach of existing approaches' is not accompanied by any quantitative metrics, success rates, error analysis, or sim-to-real deformation comparisons (cone angle, curvature, contact patch, or force matching). This directly weakens the validation of the sim-to-real transfer that the evolutionary search depends upon.
[Validation / Results] The load-bearing assumption that the physics-based simulator accurately reproduces the hand's passive flattening-to-conical morphing and resulting contact forces under granular interaction is stated but not quantified. Without reported metrics on how well simulated deformation matches hardware (e.g., in the validation or results sections), it is unclear whether the optimized parameters succeed due to model fidelity or simply because the real hand is compliant.

minor comments (1)

The online video link is referenced but the manuscript does not describe which specific behaviors or failure modes are shown, limiting the reader's ability to interpret the experimental evidence.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and for recognizing the methodological contributions of the simulation-driven evolutionary approach. We will revise the manuscript to strengthen the quantitative support for our claims on generalization and sim-to-real fidelity.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim that optimized trajectories demonstrate 'strong generalization' and address 'a range of challenging tasks previously beyond the reach of existing approaches' is not accompanied by any quantitative metrics, success rates, error analysis, or sim-to-real deformation comparisons (cone angle, curvature, contact patch, or force matching). This directly weakens the validation of the sim-to-real transfer that the evolutionary search depends upon.

Authors: We agree that the abstract would be strengthened by explicit quantitative backing for the generalization claims. In the revised manuscript we will update the abstract to reference specific metrics drawn from the experimental results, including task success rates, quantitative sim-to-real deformation comparisons (cone angle, curvature, contact patch), and force-matching data where measured. We will also add a short error analysis summary to directly tie the reported performance to the sim-to-real transfer that underpins the evolutionary optimization. revision: yes
Referee: [Validation / Results] The load-bearing assumption that the physics-based simulator accurately reproduces the hand's passive flattening-to-conical morphing and resulting contact forces under granular interaction is stated but not quantified. Without reported metrics on how well simulated deformation matches hardware (e.g., in the validation or results sections), it is unclear whether the optimized parameters succeed due to model fidelity or simply because the real hand is compliant.

Authors: We acknowledge that explicit quantitative validation of the simulator's fidelity is necessary. We will add a dedicated validation subsection (or expand the existing results section) that reports direct metrics comparing simulated and physical deformations, such as cone angle, curvature, contact patch geometry, and available force measurements. This addition will clarify the degree of model accuracy and justify the use of simulation for trajectory optimization. The current manuscript relies primarily on successful hardware transfer and visual agreement; we agree that numerical fidelity metrics are required for rigor. revision: yes

Circularity Check

0 steps flagged

No circularity: optimization and validation remain independent of target data

full rationale

The paper constructs a physics-based simulator of the soft hand's passive morphing, runs an evolutionary optimizer entirely inside that simulator to produce trajectories, and then transfers those trajectories to hardware for validation. No equation, parameter, or claim reduces to a fitted quantity defined by the real-robot outcomes; the sim-to-real transfer is presented as an external test rather than a self-referential loop. No self-citations, uniqueness theorems, or ansatzes are invoked in a load-bearing way within the provided derivation chain. The generalization claim rests on the independent real-robot experiments, not on any redefinition or renaming of the simulation outputs themselves.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; the simulation model and evolutionary framework are described at a high level without detailing fitted constants or unproven assumptions.

pith-pipeline@v0.9.0 · 5525 in / 1169 out tokens · 33535 ms · 2026-05-10T19:59:55.478355+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The SH sheet is modeled as a mass–spring–damping system... elastic energy E = ½ Σ ks(‖xi−xj‖−l0ij)² + ½ Σ kb(θpq−θ0pq)²
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

CMA-ES... optimizes both trajectory and hand roll angle

What do these tags mean?

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

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages

[1]

Scone: A food scooping robot learning framework with active perception,

Y .-L. Tai, Y . C. Chiu, Y .-W. Chao, and Y .-T. Chen, “Scone: A food scooping robot learning framework with active perception,” in Conference on Robot Learning, pp. 849–865, PMLR, 2023

work page 2023
[2]

L2d2: Robot learning from 2d drawings,

S. A. Mehta, H. Nemlekar, H. Sumant, and D. P. Losey, “L2d2: Robot learning from 2d drawings,”arXiv preprint arXiv:2505.12072, 2025

work page arXiv 2025
[3]

Learning compositional models of robot skills for task and motion planning,

Z. Wang, C. R. Garrett, L. P. Kaelbling, and T. Lozano-P ´erez, “Learning compositional models of robot skills for task and motion planning,”The International Journal of Robotics Research, vol. 40, no. 6-7, pp. 866–894, 2021

work page 2021
[4]

Learning robotic manipulation of granular media,

C. Schenck, J. Tompson, S. Levine, and D. Fox, “Learning robotic manipulation of granular media,” inConference on Robot Learning, pp. 239–248, PMLR, 2017

work page 2017
[5]

Primp: Probabilistically-informed motion primitives for efficient affordance learning from demonstration,

S. Ruan, W. Liu, X. Wang, X. Meng, and G. S. Chirikjian, “Primp: Probabilistically-informed motion primitives for efficient affordance learning from demonstration,”IEEE Transactions on Robotics, vol. 40, pp. 2868–2887, 2024

work page 2024
[6]

Soda–soft origami dynamic utensil for assisted feeding,

Y . R. Song and Y . Luo, “Soda–soft origami dynamic utensil for assisted feeding,” in2025 IEEE 8th International Conference on Soft Robotics (RoboSoft), pp. 1–8, IEEE, 2025

work page 2025
[7]

Kiri-spoon: A kirigami utensil for robot-assisted feeding,

M. Keely, B. Franco, C. Grothoff, R. K. Jenamani, T. Bhattacharjee, D. P. Losey, and H. Nemlekar, “Kiri-spoon: A kirigami utensil for robot-assisted feeding,”arXiv preprint arXiv:2501.01323, 2025

work page arXiv 2025
[8]

Learning thin deformable object manipulation with a multi-sensory integrated soft hand,

C. Zhao, C. Jiang, L. Luo, S. Yuan, Q. Chen, and H. Yu, “Learning thin deformable object manipulation with a multi-sensory integrated soft hand,”IEEE Transactions on Robotics, 2025

work page 2025
[9]

The double-scoop gripper: A tendon-driven soft-rigid end-effector for food handling exploiting constraints in narrow spaces,

L. Franco, E. Turco, V . Bo, M. Pozzi, M. Malvezzi, D. Prattichizzo, and G. Salvietti, “The double-scoop gripper: A tendon-driven soft-rigid end-effector for food handling exploiting constraints in narrow spaces,” in2024 IEEE International Conference on Robotics and Automation (ICRA), pp. 4170–4176, IEEE, 2024

work page 2024
[10]

Diff-control: A stateful diffusion-based policy for imitation learning,

X. Liu, Y . Zhou, F. Weigend, S. Sonawani, S. Ikemoto, and H. B. Amor, “Diff-control: A stateful diffusion-based policy for imitation learning,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 7453–7460, IEEE, 2024. TABLE II SCOOPING RESULTS OVER10TRIALS FOR THREE TASKS IN A REAL ROBOT. Task Trial 1 2 3 4 5 6 7 8 9 1...

work page 2024
[11]

Screw geometry meets bandits: Incremental acquisition of demonstrations to generate manipulation plans,

D. Das, A. Patankar, N. Chakraborty, C. Ramakrishnan, and I. Ra- makrishnan, “Screw geometry meets bandits: Incremental acquisition of demonstrations to generate manipulation plans,”arXiv preprint arXiv:2410.18275, 2024

work page arXiv 2024
[12]

S 2-diffusion: Generalizing from instance-level to category-level skills in robot manipulation,

Q. Yang, M. C. Welle, D. Kragic, and O. Andersson, “S 2-diffusion: Generalizing from instance-level to category-level skills in robot manipulation,”arXiv preprint arXiv:2502.09389, 2025

work page arXiv 2025
[13]

Scu-hand: Soft conical universal robotic hand for scooping granular media from containers of various sizes,

T. Takahashi, C. C. Beltran-Hernandez, Y . Kuroda, K. Tanaka, M. Hamaya, and Y . Ushiku, “Scu-hand: Soft conical universal robotic hand for scooping granular media from containers of various sizes,” arXiv preprint arXiv:2505.04162, 2025

work page arXiv 2025
[14]

Learning scooping deformable plastic objects using tactile sensors,

Y . Kageyama, M. Hamaya, K. Tanaka, A. Hashimoto, and H. Saito, “Learning scooping deformable plastic objects using tactile sensors,” in2024 IEEE 20th International Conference on Automation Science and Engineering (CASE), pp. 4020–4025, IEEE, 2024

work page 2024
[15]

Lava: Long-horizon visual action based food acquisition,

A. Bhaskar, R. Liu, V . D. Sharma, G. Shi, and P. Tokekar, “Lava: Long-horizon visual action based food acquisition,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 8929–8935, IEEE, 2024

work page 2024
[16]

Robotsmith: Generative robotic tool design for acquisition of complex manipulation skills, 2025

C. Lin, H. Yuan, Y . Wang, X. Qiu, T.-H. Wang, M. Guo, B. Wang, Y . Narang, D. Fox, and C. Gan, “Robotsmith: Generative robotic tool design for acquisition of complex manipulation skills,”arXiv preprint arXiv:2506.14763, 2025

work page arXiv 2025
[17]

Investigat- ing strategies enabling novice users to teach plannable hierarchical tasks to robots,

N. Moorman, A. Singh, M. Natarajan, E. Hedlund-Botti, M. Schrum, C. Yang, L. Seelam, M. C. Gombolay, and N. Gopalan, “Investigat- ing strategies enabling novice users to teach plannable hierarchical tasks to robots,”The International Journal of Robotics Research, p. 02783649241301075, 2024

work page 2024
[18]

Goats: Goal sampling adaptation for scooping with curriculum reinforcement learning,

Y . Niu, S. Jin, Z. Zhang, J. Zhu, D. Zhao, and L. Zhang, “Goats: Goal sampling adaptation for scooping with curriculum reinforcement learning,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1023–1030, IEEE, 2023

work page 2023
[19]

Autonomous excavation of challenging terrain using oscillatory primitives and adaptive impedance control,

N. Franceschini, P. Thangeda, M. Ornik, and K. Hauser, “Autonomous excavation of challenging terrain using oscillatory primitives and adaptive impedance control,”arXiv preprint arXiv:2409.18273, 2024

work page arXiv 2024
[20]

A soft-rigid hybrid gripper with lateral compliance and dexterous in-hand manipulation,

W. Zhu, C. Lu, Q. Zheng, Z. Fang, H. Che, K. Tang, M. Zhu, S. Liu, and Z. Wang, “A soft-rigid hybrid gripper with lateral compliance and dexterous in-hand manipulation,”IEEE/ASME Transactions on Mechatronics, vol. 28, no. 1, pp. 104–115, 2022

work page 2022
[21]

Riso: Combining rigid grippers with soft switchable adhesives,

S. A. Mehta, Y . Kim, J. Hoegerman, M. D. Bartlett, and D. P. Losey, “Riso: Combining rigid grippers with soft switchable adhesives,” in 2023 IEEE International Conference on Soft Robotics (RoboSoft), pp. 1–8, IEEE, 2023

work page 2023
[22]

Dexgrip: Multi- modal soft gripper with dexterous grasping and in-hand manipulation capacity,

X. Wang, L. Horrigan, J. Pinskier, G. Shi, V . Viswanathan, L. Liow, T. Bandyopadhyay, J. J. Chung, and D. Howard, “Dexgrip: Multi- modal soft gripper with dexterous grasping and in-hand manipulation capacity,” in2025 IEEE 8th International Conference on Soft Robotics (RoboSoft), pp. 1–6, IEEE, 2025

work page 2025
[23]

Fish mouth inspired origami gripper for robust multi-type underwater grasping,

H. Guo, J. Huang, I. Zhang, B. Liang, X. Ma, Y . Liu, and J. Zhou, “Fish mouth inspired origami gripper for robust multi-type underwater grasping,”arXiv preprint arXiv:2503.11049, 2025

work page arXiv 2025
[24]

Dih-tele: Dexterous in-hand teleoperation framework for learning multiobjects manipulation with tactile sensing,

J. Huang, K. Chen, J. Zhou, X. Lin, P. Abbeel, Q. Dou, and Y . Liu, “Dih-tele: Dexterous in-hand teleoperation framework for learning multiobjects manipulation with tactile sensing,”IEEE/ASME Trans- actions on Mechatronics, 2025

work page 2025
[25]

Design and control of soft robots using differentiable simulation,

M. B ¨acher, E. Knoop, and C. Schumacher, “Design and control of soft robots using differentiable simulation,”Current Robotics Reports, vol. 2, no. 2, pp. 211–221, 2021

work page 2021
[26]

Softgym: Benchmarking deep reinforcement learning for deformable object manipulation,

X. Lin, Y . Wang, J. Olkin, and D. Held, “Softgym: Benchmarking deep reinforcement learning for deformable object manipulation,” in Conference on Robot Learning, pp. 432–448, PMLR, 2021

work page 2021
[27]

Elastica: A compliant mechanics environment for soft robotic control,

N. Naughton, J. Sun, A. Tekinalp, T. Parthasarathy, G. Chowdhary, and M. Gazzola, “Elastica: A compliant mechanics environment for soft robotic control,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 3389–3396, 2021

work page 2021
[28]

Robotic object insertion with a soft wrist through sim-to-real privi- leged training,

Y . Fuchioka, C. C. Beltran-Hernandez, H. Nguyen, and M. Hamaya, “Robotic object insertion with a soft wrist through sim-to-real privi- leged training,” in2024 IEEE/RSJ International Conference on Intel- ligent Robots and Systems (IROS), pp. 9159–9166, IEEE, 2024

work page 2024
[29]

3d force and contact estimation for a soft-bubble visuotactile sensor using fem,

J.-C. Peng, S. Yao, and K. Hauser, “3d force and contact estimation for a soft-bubble visuotactile sensor using fem,” in2024 IEEE Inter- national Conference on Robotics and Automation (ICRA), pp. 5666– 5672, IEEE, 2024

work page 2024
[30]

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

work page 2023
[31]

A review on origami simula- tions: from kinematics, to mechanics, toward multiphysics,

Y . Zhu, M. Schenk, and E. T. Filipov, “A review on origami simula- tions: from kinematics, to mechanics, toward multiphysics,”Applied Mechanics Reviews, vol. 74, no. 3, p. 030801, 2022

work page 2022
[32]

Simulation of rigid origami,

T. Tachi, “Simulation of rigid origami,”Origami, vol. 4, no. 08, pp. 175–187, 2009

work page 2009
[33]

Fast segment anything,

X. Zhao, W. Ding, Y . An, Y . Du, T. Yu, M. Li, M. Tang, and J. Wang, “Fast segment anything,”arXiv preprint arXiv:2306.12156, 2023

work page arXiv 2023
[34]

Mujoco: A physics engine for model-based control,

E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,” in2012 IEEE/RSJ international conference on intelligent robots and systems, pp. 5026–5033, IEEE, 2012

work page 2012
[35]

Mujoco playground,

K. Zakka, B. Tabanpour, Q. Liao, M. Haiderbhai, S. Holt, J. Y . Luo, A. Allshire, E. Frey, K. Sreenath, L. A. Kahrs,et al., “Mujoco playground,”arXiv preprint arXiv:2502.08844, 2025

work page arXiv 2025
[36]

Adapting arbitrary normal mutation distributions in evolution strategies: The covariance matrix adaptation,

N. Hansen and A. Ostermeier, “Adapting arbitrary normal mutation distributions in evolution strategies: The covariance matrix adaptation,” inProceedings of IEEE international conference on evolutionary computation, pp. 312–317, IEEE, 1996

work page 1996