pith. sign in

arxiv: 2605.31120 · v1 · pith:B2O6MFY7new · submitted 2026-05-29 · 💻 cs.GR · cs.AI· cs.LG

SWIM: Single-Instance Whole-Body Imitation for swiMming

Binglun Wang , Edmond S. L. Ho , He Wang This is my paper

Pith reviewed 2026-06-28 20:12 UTC · model grok-4.3

classification 💻 cs.GR cs.AIcs.LG
keywords swimming animationimitation learningphysically based animationcharacter controlgeneralizationfluid interactionsingle instance learning
0
0 comments X

The pith

SWIM learns a swimming controller from one motion clip that generalizes to new bodies, styles, and fluid conditions.

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

The paper presents SWIM as a method to create physically valid swimming animations by imitating a single reference motion. It seeks to establish that such imitation can produce stable full-body control despite the unpredictable forces from water, while also adapting to changes in body shape, swimming style, or environment. A reader would care because most prior work on character animation relies on many examples or simplified environments, leaving fluid tasks like swimming under-served. If the approach holds, it lowers the data requirement for training controllers in complex physical settings.

Core claim

SWIM is a new imitation method for swimming motions, which can learn from a single swimming motion and generalize to unseen environments, body conditions, and swimming styles. The method addresses the challenges of volatile fluid forces, lack of reference data, and slow simulation by producing data-efficient, stable, robust, and generalizable control that outperforms alternatives on multiple tasks.

What carries the argument

Single-instance whole-body imitation that trains a policy directly from one motion clip to handle continuous fluid interactions.

If this is right

  • Training for fluid-based motions becomes feasible with minimal reference data.
  • The same controller works across varied body proportions without retraining.
  • Control remains effective under disturbances that differ from the original motion.
  • Simulation cost during training stays manageable even for full-body fluid tasks.

Where Pith is reading between the lines

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

  • The single-clip approach could apply to other continuous-interaction domains such as flying or paddling.
  • It suggests that explicit fluid modeling during policy search may be less necessary than previously assumed for generalization.
  • Robotics applications might use the same imitation step to adapt virtual swimmers to real underwater hardware.

Load-bearing premise

A policy learned by imitating one motion clip remains stable when fluid forces change unpredictably without needing extra data or slower training.

What would settle it

Run the trained controller on a body shape or fluid viscosity far outside the single training clip and observe whether forward propulsion collapses or the character becomes unstable.

Figures

Figures reproduced from arXiv: 2605.31120 by Binglun Wang, Edmond S. L. Ho, He Wang.

Figure 1
Figure 1. Figure 1: Trained on a simple goal-reaching task along a straight trajectory in a small pool and a single swimming motion, SWIM can zero-shot to [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Left: forward and deviation reward. Middle: head reward. Right: training task. Red dots are goals. Blue dots are intermediate goals. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Total reward against total number of samples used in training [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Goal-reaching boundary sweep in the 5 m pool. From left to right, the subfigures report Pos, Prog, Dev, Roll, and Vel. The bottom axis [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative examples of cross-environment and cross-style generalization. Row 1: freestyle in oil. Row 2: freestyle in water with an [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Different flow speeds for goal-reaching in the 3 m pool. The policy is trained in quiet water and evaluated under different inflow speeds [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative comparison between methods in a goal-reaching task. The dashed line is the desired trajectory and the orange line is the [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Body geometry change with a 10cm long fin attached to the feet. Adding fins do enable the character to swim fast. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
read the original abstract

We propose a new method for synthesizing physically-based swimming motions. Physically-based character animation aims to generate physically valid, controllable, and natural-looking motions which can respond to unexpected disturbances, where one dictating factor of difficulty is the complexity of the task, especially the level of sophistication of the required interactions with the environment. Existing research has succeeded in various tasks in static and dynamic environments. We push the difficulty further to swimming, which requires full-body coordination and continuous interactions with fluids, a new level of complexity when it comes to interacting with the environment. This complexity imposes challenges in learning control under volatile environmental forces, generalizing control to different environments and swimming styles, lack of data references, and prohibitively slow physical simulation which is inevitable during control learning. To this end, we propose SWIM, a new imitation method for swimming motions, which can learn from a single swimming motion and generalize to unseen environments, body conditions, and swimming styles. Extensive evaluation and comparison demonstrate that SWIM is data-efficient, stable, robust, and generalizable, outperforming alternative methods across multiple classes of tasks and metrics.

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.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces SWIM, a single-instance imitation learning method for whole-body physically-based swimming animation. It claims to learn a policy from one reference motion clip that generalizes to unseen environments, body conditions, and swimming styles while remaining stable under volatile fluid forces, addressing data scarcity and slow simulation; extensive evaluations are said to show superiority over alternatives in data efficiency, stability, robustness, and generalization.

Significance. If the experimental claims are substantiated with quantitative evidence, the work would represent a meaningful advance in physically-based character animation by demonstrating data-efficient control for high-complexity fluid interactions, a domain where prior methods have been limited to simpler environments. The single-instance aspect, if achieved without hidden data augmentation or excessive simulation, would be a notable strength for tasks with prohibitive data or compute requirements.

major comments (2)
  1. [Abstract] Abstract: the central claim that a policy learned via imitation from one motion clip generalizes stably to unseen environments, bodies, and styles is asserted without any equations, architecture details, reward formulation, or experimental results (no tables, figures, or metrics). This mechanism is load-bearing for the generalization and stability assertions yet remains uninspectable from the provided text.
  2. [Abstract] Abstract: the statement that SWIM 'outperforms alternative methods across multiple classes of tasks and metrics' is made without reference to any specific baselines, quantitative scores, error bars, or statistical tests. This directly underpins the superiority claim and cannot be evaluated.
minor comments (1)
  1. [Abstract] The abstract lists four open challenges (volatile forces, generalization, lack of references, slow simulation) but does not indicate which components of SWIM (architecture, reward, randomization, or other) are intended to resolve each; a brief mapping would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the comments. The abstract is a concise summary per standard practice, with full technical details and results in the body of the paper. We address each point below and will make targeted revisions to the abstract for improved clarity.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that a policy learned via imitation from one motion clip generalizes stably to unseen environments, bodies, and styles is asserted without any equations, architecture details, reward formulation, or experimental results (no tables, figures, or metrics). This mechanism is load-bearing for the generalization and stability assertions yet remains uninspectable from the provided text.

    Authors: Abstracts are intentionally high-level and do not contain equations or results to remain concise. The policy architecture, single-instance imitation objective, fluid-interaction reward formulation, and generalization mechanism are fully specified in Section 3 (Method), while stability and generalization results appear in Section 4. We will revise the abstract to add a brief clause indicating the core technical approach (e.g., “via a fluid-aware imitation objective and domain-randomized policy”) and a pointer to the detailed sections. revision: yes

  2. Referee: [Abstract] Abstract: the statement that SWIM 'outperforms alternative methods across multiple classes of tasks and metrics' is made without reference to any specific baselines, quantitative scores, error bars, or statistical tests. This directly underpins the superiority claim and cannot be evaluated.

    Authors: The superiority claim summarizes the quantitative comparisons presented in Section 4, which include specific baselines (standard RL, prior imitation methods, and ablations), success rates, stability metrics, and generalization errors with error bars across multiple random seeds and statistical significance tests. We will revise the abstract to reference these evaluations more explicitly (e.g., “outperforming baselines in Section 4 across …”) while respecting length constraints. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The provided abstract and context describe a proposed imitation-learning method for swimming animation that learns from one motion clip and generalizes via evaluation. No equations, fitted parameters renamed as predictions, self-citation chains, or ansatzes are present that would reduce any claim to its own inputs by construction. The central assertions rest on empirical comparisons rather than definitional or fitted reductions, satisfying the default expectation for a non-circular method paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no information on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5732 in / 955 out tokens · 27278 ms · 2026-06-28T20:12:02.017845+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

75 extracted references · 6 canonical work pages · 2 internal anchors

  1. [1]

    An introduction to physics-based animation

    Adam W Bargteil, Tamar Shinar, and Paul G Kry. An introduction to physics-based animation. InSIGGRAPH Asia 2020 Courses, pages 1–57. 2020

    2020

  2. [2]

    DeepMimic: example-guided deep reinforcement learning of physics-based character skills.ACM Transactions on Graphics, 37(4):1–14, 2018

    Xue Bin Peng, Pieter Abbeel, Sergey Levine, and Michiel Van De Panne. DeepMimic: example-guided deep reinforcement learning of physics-based character skills.ACM Transactions on Graphics, 37(4):1–14, 2018

    2018

  3. [3]

    Charles River Media Hingham, 2005

    Kenny Erleben, Jon Sporring, Knud Henriksen, and Henrik Dohlmann.Physics-based animation, volume 79. Charles River Media Hingham, 2005

    2005

  4. [4]

    Physics-based character animation and human motor control.Physics of Life Reviews, 46:190–219, 2023

    Joan Llobera and Caecilia Charbonnier. Physics-based character animation and human motor control.Physics of Life Reviews, 46:190–219, 2023

    2023

  5. [5]

    Masmitja, M

    I. Masmitja, M. Martin, T. O’Reilly, B. Kieft, N. Palomeras, J. Navarro, and K. Katija. Dynamic robotic tracking of under- water targets using reinforcement learning.Science Robotics, 8(80):eade7811, 2023

    2023

  6. [6]

    Learning coordinated badminton skills for legged manipu- lators.Science Robotics, 10(102):eadu3922, 2025

    Yuntao Ma, Andrei Cramariuc, Farbod Farshidian, and Marco Hutter. Learning coordinated badminton skills for legged manipu- lators.Science Robotics, 10(102):eadu3922, 2025

    2025

  7. [7]

    Robie, Carmen Morrow, Guido Novati, Zinovia Stefanidi, Gert-Jan Both, Gwyneth M

    Roman Vaxenburg, Igor Siwanowicz, Josh Merel, Alice A. Robie, Carmen Morrow, Guido Novati, Zinovia Stefanidi, Gert-Jan Both, Gwyneth M. Card, Michael B. Reiser, Matthew M. Botvinick, Kristin M. Branson, Yuval Tassa, and Srinivas C. Turaga. Whole- body physics simulation of fruit fly locomotion.Nature, 2025

    2025

  8. [8]

    Spacetime constraints.ACM Siggraph Computer Graphics, 22(4):159–168, 1988

    Andrew Witkin and Michael Kass. Spacetime constraints.ACM Siggraph Computer Graphics, 22(4):159–168, 1988

    1988

  9. [9]

    University of California, Los Angeles, 2013

    Weiguang Si.Realistic simulation and control of human swimming and underwater movement. University of California, Los Angeles, 2013

    2013

  10. [10]

    Sim- bicon: Simple biped locomotion control.ACM Transactions on Graphics (TOG), 26(3):105–es, 2007

    KangKang Yin, Kevin Loken, and Michiel Van de Panne. Sim- bicon: Simple biped locomotion control.ACM Transactions on Graphics (TOG), 26(3):105–es, 2007

    2007

  11. [11]

    Mimickit: A reinforcement learning framework for motion imitation and control.arXiv preprint arXiv:2510.13794, 2025

    Xue Bin Peng. Mimickit: A reinforcement learning framework for motion imitation and control.arXiv preprint arXiv:2510.13794, 2025

    work page arXiv 2025

  12. [12]

    1000 layer networks for self- supervised rl: Scaling depth can enable new goal-reaching ca- pabilities.Advances in Neural Information Processing Systems, 38:157643–157670, 2026

    Kevin Wang, Ishaan Javali, Michał Bortkiewicz, Tomasz Trzcin- ski, and Benjamin Eysenbach. 1000 layer networks for self- supervised rl: Scaling depth can enable new goal-reaching ca- pabilities.Advances in Neural Information Processing Systems, 38:157643–157670, 2026

    2026

  13. [13]

    Composite motion learning with task control.ACM Transactions on Graphics, 42(4):1–16, 2023

    Pei Xu, Xiumin Shang, Victor Zordan, and Ioannis Karamouzas. Composite motion learning with task control.ACM Transactions on Graphics, 42(4):1–16, 2023

    2023

  14. [14]

    Karen Liu, Julien Pettré, Michiel van de Panne, and Marie- Paule Cani

    Ariel Kwiatkowski, Eduardo Alvarado, Vicky Kalogeiton, C. Karen Liu, Julien Pettré, Michiel van de Panne, and Marie- Paule Cani. A survey on reinforcement learning methods in charac- ter animation.Computer Graphics Forum, 41(2):613–639, 2022

    2022

  15. [15]

    Learning to schedule control fragments for physics-based characters using deep q-learning

    Libin Liu and Jessica Hodgins. Learning to schedule control fragments for physics-based characters using deep q-learning. ACM Transactions on Graphics, 36(3):1–14, 2017

    2017

  16. [16]

    Drecon: data-driven responsive control of physics-based characters.ACM Transactions On Graphics (TOG), 38(6):1–11, 2019

    Kevin Bergamin, Simon Clavet, Daniel Holden, and James Richard Forbes. Drecon: data-driven responsive control of physics-based characters.ACM Transactions On Graphics (TOG), 38(6):1–11, 2019

    2019

  17. [17]

    ASE: large-scale reusable adversarial skill embed- dings for physically simulated characters.ACM Transactions on Graphics, 41(4):1–17, 2022

    Xue Bin Peng, Yunrong Guo, Lina Halper, Sergey Levine, and Sanja Fidler. ASE: large-scale reusable adversarial skill embed- dings for physically simulated characters.ACM Transactions on Graphics, 41(4):1–17, 2022

    2022

  18. [18]

    Deeploco: Dynamic locomotion skills using hierarchi- cal deep reinforcement learning.Acm transactions on graphics (tog), 36(4):1–13, 2017

    Xue Bin Peng, Glen Berseth, KangKang Yin, and Michiel Van De Panne. Deeploco: Dynamic locomotion skills using hierarchi- cal deep reinforcement learning.Acm transactions on graphics (tog), 36(4):1–13, 2017

    2017

  19. [19]

    Sfv: Reinforcement learning of physical skills from videos.ACM Transactions On Graphics (TOG), 37(6):1–14, 2018

    Xue Bin Peng, Angjoo Kanazawa, Jitendra Malik, Pieter Abbeel, and Sergey Levine. Sfv: Reinforcement learning of physical skills from videos.ACM Transactions On Graphics (TOG), 37(6):1–14, 2018

    2018

  20. [20]

    AMP: Adversarial motion priors for stylized physics- based character control.ACM Transactions on Graphics, 40(4), 2021

    Xue Bin Peng, Ze Ma, Pieter Abbeel, Sergey Levine, and Angjoo Kanazawa. AMP: Adversarial motion priors for stylized physics- based character control.ACM Transactions on Graphics, 40(4), 2021

    2021

  21. [21]

    Physics-based motion imitation with adversarial differential discriminators

    Ziyu Zhang, Sergey Bashkirov, Dun Yang, Yi Shi, Michael Tay- lor, and Xue Bin Peng. Physics-based motion imitation with adversarial differential discriminators. InSIGGRAPH Asia 2025 Conference Papers (SIGGRAPH Asia ’25 Conference Papers), 2025

    2025

  22. [22]

    Hierarchical visuomotor control of humanoids

    Josh Merel, Arun Ahuja, Vu Pham, Saran Tunyasuvunakool, Siqi Liu, Dhruva Tirumala, Nicolas Heess, and Greg Wayne. Hierarchical visuomotor control of humanoids.arXiv preprint arXiv:1811.09656, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  23. [23]

    MaskedMimic: Unified physics-based charac- ter control through masked motion inpainting.ACM Transactions on Graphics, 43(6):1–21, 2024

    Chen Tessler, Yunrong Guo, Ofir Nabati, Gal Chechik, and Xue Bin Peng. MaskedMimic: Unified physics-based charac- ter control through masked motion inpainting.ACM Transactions on Graphics, 43(6):1–21, 2024

    2024

  24. [24]

    Allsteps: curriculum-driven learning of stepping stone skills

    Zhaoming Xie, Hung Yu Ling, Nam Hee Kim, and Michiel Van De Panne. Allsteps: curriculum-driven learning of stepping stone skills. InComputer Graphics Forum, volume 39, pages 213–224. Wiley Online Library, 2020

    2020

  25. [25]

    Blind bipedal stair traversal via sim-to-real reinforcement learning,

    Jonah Siekmann, Kevin Green, John Warila, Alan Fern, and Jonathan Hurst. Blind bipedal stair traversal via sim-to-real rein- forcement learning.arXiv preprint arXiv:2105.08328, 2021

    work page arXiv 2021

  26. [26]

    Learning to use chop- sticks in diverse gripping styles.ACM Transactions on Graphics, 41(4):1–17, 2022

    Zeshi Yang, Kangkang Yin, and Libin Liu. Learning to use chop- sticks in diverse gripping styles.ACM Transactions on Graphics, 41(4):1–17, 2022

    2022

  27. [27]

    SkillMimic: Learning basketball interaction skills from demonstrations, 2025

    Yinhuai Wang, Qihan Zhao, Runyi Yu, Hok Wai Tsui, Ailing Zeng, Jing Lin, Zhengyi Luo, Jiwen Yu, Xiu Li, Qifeng Chen, Jian Zhang, Lei Zhang, and Ping Tan. SkillMimic: Learning basketball interaction skills from demonstrations, 2025

    2025

  28. [28]

    Learning basketball dribbling skills using trajectory optimization and deep reinforcement learn- ing.ACM Transactions on Graphics, 37(4):1–14, 2018

    Libin Liu and Jessica Hodgins. Learning basketball dribbling skills using trajectory optimization and deep reinforcement learn- ing.ACM Transactions on Graphics, 37(4):1–14, 2018

    2018

  29. [29]

    Learning physically simulated tennis skills from broadcast videos.ACM Transactions on Graphics, 42(4):1–14, 2023

    Haotian Zhang, Ye Yuan, Viktor Makoviychuk, Yunrong Guo, Sanja Fidler, Xue Bin Peng, and Kayvon Fatahalian. Learning physically simulated tennis skills from broadcast videos.ACM Transactions on Graphics, 42(4):1–14, 2023. 10

    2023

  30. [30]

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

    Viktor Makoviychuk, Lukasz Wawrzyniak, Yunrong Guo, Michelle Lu, Kier Storey, Miles Macklin, David Hoeller, Nikita Rudin, Arthur Allshire, Ankur Handa, et al. Isaac gym: High per- formance gpu-based physics simulation for robot learning.arXiv preprint arXiv:2108.10470, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  31. [31]

    Brax–a differentiable physics engine for large scale rigid body simulation.arXiv preprint arXiv:2106.13281, 2021

    C Daniel Freeman, Erik Frey, Anton Raichuk, Sertan Girgin, Igor Mordatch, and Olivier Bachem. Brax–a differentiable physics engine for large scale rigid body simulation.arXiv preprint arXiv:2106.13281, 2021

    work page arXiv 2021

  32. [32]

    Neural state machine for character-scene interactions.ACM Transactions on Graphics, 38(6):178, 2019

    Sebastian Starke, He Zhang, Taku Komura, and Jun Saito. Neural state machine for character-scene interactions.ACM Transactions on Graphics, 38(6):178, 2019

    2019

  33. [33]

    Lo- cal motion phases for learning multi-contact character movements

    Sebastian Starke, Yiwei Zhao, Taku Komura, and Kazi Zaman. Lo- cal motion phases for learning multi-contact character movements. ACM Transactions on Graphics, 39(4), 2020

    2020

  34. [34]

    Elastomonolith: A monolithic optimization-based liquid solver for contact-aware elastic-solid coupling.ACM Transactions on Graphics (TOG), 41(6):1–19, 2022

    Tetsuya Takahashi and Christopher Batty. Elastomonolith: A monolithic optimization-based liquid solver for contact-aware elastic-solid coupling.ACM Transactions on Graphics (TOG), 41(6):1–19, 2022

    2022

  35. [35]

    Eulerian solid- fluid coupling.ACM Transactions on Graphics (TOG), 35(6):1–8, 2016

    Yun Teng, David IW Levin, and Theodore Kim. Eulerian solid- fluid coupling.ACM Transactions on Graphics (TOG), 35(6):1–8, 2016

    2016

  36. [36]

    Interlinked SPH pressure solvers for strong fluid-rigid coupling.ACM Transactions on Graphics, 38(1):1–13, 2019

    Christoph Gissler, Andreas Peer, Stefan Band, Jan Bender, and Matthias Teschner. Interlinked SPH pressure solvers for strong fluid-rigid coupling.ACM Transactions on Graphics, 38(1):1–13, 2019

    2019

  37. [37]

    IQ-MPM: an interface quadra- ture material point method for non-sticky strongly two-way cou- pled nonlinear solids and fluids.ACM Transactions on Graphics, 39(4), 2020

    Yu Fang, Ziyin Qu, Minchen Li, Xinxin Zhang, Yixin Zhu, Mridul Aanjaneya, and Chenfanfu Jiang. IQ-MPM: an interface quadra- ture material point method for non-sticky strongly two-way cou- pled nonlinear solids and fluids.ACM Transactions on Graphics, 39(4), 2020

    2020

  38. [38]

    Monolith: a monolithic pressure-viscosity-contact solver for strong two-way rigid-rigid rigid-fluid coupling.ACM Transactions on Graphics, 39(6):1–16, 2020

    Tetsuya Takahashi and Christopher Batty. Monolith: a monolithic pressure-viscosity-contact solver for strong two-way rigid-rigid rigid-fluid coupling.ACM Transactions on Graphics, 39(6):1–16, 2020

    2020

  39. [39]

    Learning to simulate complex physics with graph networks

    Alvaro Sanchez-Gonzalez, Jonathan Godwin, Tobias Pfaff, Rex Ying, Jure Leskovec, and Peter Battaglia. Learning to simulate complex physics with graph networks. InInternational conference on machine learning, pages 8459–8468. PMLR, 2020

    2020

  40. [40]

    CFC: Simulating character-fluid coupling using a two- level world model.ACM Transactions on Graphics, 44(6):1–17, 2025

    Zhiyang Dou, Chen Peng, Xinyu Lu, Xiaohan Ye, Lixing Fang, Yuan Liu, Wenping Wang, Chuang Gan, Lingjie Liu, and Taku Komura. CFC: Simulating character-fluid coupling using a two- level world model.ACM Transactions on Graphics, 44(6):1–17, 2025

    2025

  41. [41]

    Proximal policy optimization algorithms, 2017

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms, 2017

    2017

  42. [42]

    Pybullet gymperium

    Benjamin Ellenberger. Pybullet gymperium. https:// github.com/benelot/pybullet-gym, 2018–2019

    2018

  43. [43]

    SPlisHSPlasH Library

    Jan Bender et al. SPlisHSPlasH Library

  44. [44]

    Layered dynamic control for interactive character swimming

    Po-Feng Yang, Joe Laszlo, and Karan Singh. Layered dynamic control for interactive character swimming. InProceedings of the 2004 ACM SIGGRAPH/Eurographics symposium on Com- puter animation - SCA ’04, page 39. ACM Press, 2004. ISSN: 17275288

    2004

  45. [45]

    Karen Liu

    Jie Tan, Yuting Gu, Greg Turk, and C. Karen Liu. Articulated swimming creatures. InACM SIGGRAPH 2011 papers, pages 1–12. ACM, 2011

    2011

  46. [46]

    Kwatra, C

    N. Kwatra, C. Wojtan, M. Carlson, I.A. Essa, P.J. Mucha, and G. Turk. Fluid simulation with articulated bodies.IEEE Trans- actions on Visualization and Computer Graphics, 16(1):70–80, 2010

    2010

  47. [47]

    Realistic biomechanical simulation and control of hu- man swimming.ACM Transactions on Graphics, 34(1):1–15, 2014

    Weiguang Si, Sung-Hee Lee, Eftychios Sifakis, and Demetri Ter- zopoulos. Realistic biomechanical simulation and control of hu- man swimming.ACM Transactions on Graphics, 34(1):1–15, 2014

    2014

  48. [48]

    Creat- ing fluid-interactive virtual agents by an efficient simulator with local-domain control.ACM Transactions on Graphics (SIG- GRAPH 2025), 2025

    Wenbin Song, Heng Zhang, Yang Wang, and Xiaopei Liu. Creat- ing fluid-interactive virtual agents by an efficient simulator with local-domain control.ACM Transactions on Graphics (SIG- GRAPH 2025), 2025. Accepted for publication. A video clip was selected for the Technical Papers Trailer

    2025

  49. [49]

    Sethuraman, Yiting Zhang, and Katherine A

    Jingyu Song, Haoyu Ma, Onur Bagoren, Advaith V . Sethuraman, Yiting Zhang, and Katherine A. Skinner. OceanSim: A GPU- accelerated underwater robot perception simulation framework, 2025

    2025

  50. [50]

    Petillot, and Canjun Yang

    Shuguang Chu, Zebin Huang, Yutong Li, Mingwei Lin, Ignacio Carlucho, Yvan R. Petillot, and Canjun Yang. MarineGym: A high-performance reinforcement learning platform for underwater robotics, 2025

    2025

  51. [51]

    Karen Liu

    Pei Xu, Yufei Ye, Shuchun Sun, Yu Ding, Elizabeth Schumann, and C. Karen Liu. MUSIC: Learning muscle-driven dexterous hand control. volume 45. ACM New York, NY , USA, 2026

    2026

  52. [52]

    ControlV AE: Model-based learning of generative controllers for physics-based characters.ACM Transactions on Graphics, 41(6):1–16, 2022

    Heyuan Yao, Zhenhua Song, Baoquan Chen, and Libin Liu. ControlV AE: Model-based learning of generative controllers for physics-based characters.ACM Transactions on Graphics, 41(6):1–16, 2022

    2022

  53. [53]

    MoConVQ: Unified physics-based motion control via scalable discrete representations, 2023

    Heyuan Yao, Zhenhua Song, Yuyang Zhou, Tenglong Ao, Bao- quan Chen, and Libin Liu. MoConVQ: Unified physics-based motion control via scalable discrete representations, 2023

    2023

  54. [54]

    Next generation computer animation techniques, 2017

    2017

  55. [55]

    Creature control in a fluid environment.IEEE Transactions on Visualization and Computer Graphics, 17(5):682– 693, 2011

    M Lentine, J T Gretarsson, C Schroeder, A Robinson-Mosher, and R Fedkiw. Creature control in a fluid environment.IEEE Transactions on Visualization and Computer Graphics, 17(5):682– 693, 2011

    2011

  56. [56]

    Physics-based fluid simulation in computer graphics: Survey, research trends, and challenges.Computational Visual Media, 10(5):803–858, 2024

    Xiaokun Wang, Yanrui Xu, Sinuo Liu, Bo Ren, Jirí Kosinka, Alexandru C Telea, Jiamin Wang, Chongming Song, Jian Chang, Chenfeng Li, et al. Physics-based fluid simulation in computer graphics: Survey, research trends, and challenges.Computational Visual Media, 10(5):803–858, 2024

    2024

  57. [57]

    Smoothed particle hydrodynamics techniques for the physics based simulation of fluids and solids.Eurographics 2019 - Tutorials, page 41 pages, 2019

    Dan Koschier, Jan Bender, Barbara Solenthaler, and Matthias Teschner. Smoothed particle hydrodynamics techniques for the physics based simulation of fluids and solids.Eurographics 2019 - Tutorials, page 41 pages, 2019

    2019

  58. [58]

    Neu- ral monte carlo fluid simulation

    Pranav Jain, Ziyin Qu, Peter Yichen Chen, and Oded Stein. Neu- ral monte carlo fluid simulation. InSpecial Interest Group on Computer Graphics and Interactive Techniques Conference Con- ference Papers, pages 1–11. ACM, 2024. 11

    2024

  59. [59]

    Raissi, P

    M. Raissi, P. Perdikaris, and G.E. Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equa- tions.Journal of Computational Physics, 378:686–707, 2019. Publisher: Elsevier BV

    2019

  60. [60]

    Physics-based deep learning.arXiv preprint arXiv:2109.05237, 2021

    Nils Thuerey, Philipp Holl, Maximilian Mueller, Patrick Schnell, Felix Trost, and Kiwon Um. Physics-based deep learning.arXiv preprint arXiv:2109.05237, 2021

    work page arXiv 2021

  61. [61]

    Weak baselines and report- ing biases lead to overoptimism in machine learning for fluid- related partial differential equations.Nature machine intelligence, 6(10):1256–1269, 2024

    Nick McGreivy and Ammar Hakim. Weak baselines and report- ing biases lead to overoptimism in machine learning for fluid- related partial differential equations.Nature machine intelligence, 6(10):1256–1269, 2024

    2024

  62. [62]

    Divergence-free smoothed particle hydrodynamics

    Jan Bender and Dan Koschier. Divergence-free smoothed particle hydrodynamics. InProceedings of the 14th ACM SIGGRAPH / Eurographics Symposium on Computer Animation, pages 147–

  63. [63]

    Versatile rigid-fluid coupling for incompressible SPH.ACM Transactions on Graphics, 31(4):1–8, 2012

    Nadir Akinci, Markus Ihmsen, Gizem Akinci, Barbara Solen- thaler, and Matthias Teschner. Versatile rigid-fluid coupling for incompressible SPH.ACM Transactions on Graphics, 31(4):1–8, 2012

    2012

  64. [64]

    Underwater soft robot modeling and control with differentiable simulation.IEEE Robotics and Automation Letters, 6(3):4994–5001, 2021

    Tao Du, Josie Hughes, Sebastien Wah, Wojciech Matusik, and Daniela Rus. Underwater soft robot modeling and control with differentiable simulation.IEEE Robotics and Automation Letters, 6(3):4994–5001, 2021

    2021

  65. [65]

    Reinforce- ment learning control for the swimming motions of a beaver-like, single-legged robot based on biological inspiration.Robotics and Autonomous Systems, 154:104116, 2022

    Gang Chen, Yuwang Lu, Xin Yang, and Huosheng Hu. Reinforce- ment learning control for the swimming motions of a beaver-like, single-legged robot based on biological inspiration.Robotics and Autonomous Systems, 154:104116, 2022

    2022

  66. [66]

    Unified motion planner for fishes with various swimming styles.ACM Transactions on Graphics (TOG), 35(4):1–15, 2016

    Daiki Satoi, Mikihiro Hagiwara, Akira Uemoto, Hisanao Nakadai, and Junichi Hoshino. Unified motion planner for fishes with various swimming styles.ACM Transactions on Graphics (TOG), 35(4):1–15, 2016

    2016

  67. [67]

    DiffFR: Differentiable SPH-based fluid-rigid coupling for rigid body control.ACM Transactions on Graphics, 42(6):1–17, 2023

    Zhehao Li, Qingyu Xu, Xiaohan Ye, Bo Ren, and Ligang Liu. DiffFR: Differentiable SPH-based fluid-rigid coupling for rigid body control.ACM Transactions on Graphics, 42(6):1–17, 2023

    2023

  68. [68]

    Binglun Wang, Niladri Shekhar Dutt, and Niloy J. Mitra. Pro- teusNeRF: Fast lightweight NeRF editing using 3d-aware image context.Proceedings of the ACM on Computer Graphics and Interactive Techniques, 7(1):1–17, 2024

    2024

  69. [69]

    Erwin Coumans.PyBullet Physics SDK, 2021

    2021

  70. [70]

    Kry, Michael Neff, Morgan McGuire, Ioannis Karamouzas, and Victor Zordan

    Pei Xu, Kaixiang Xie, Sheldon Andrews, Paul G. Kry, Michael Neff, Morgan McGuire, Ioannis Karamouzas, and Victor Zordan. AdaptNet: Policy adaptation for physics-based character control. ACM Transactions on Graphics, 42(6):1–17, 2023

    2023

  71. [71]

    Siddharth Mysore, Bassel Mabsout, Kate Saenko, and Renato Mancuso. How to train your quadrotor: A framework for con- sistently smooth and responsive flight control via reinforcement learning.ACM Transactions on Cyber-Physical Systems (TCPS), 5(4):1–24, 2021

    2021

  72. [72]

    Body roll in swim- ming: A review.Journal of Sports Sciences, 28(3):229–236, 2010

    Stelios G Psycharakis and Ross H Sanders. Body roll in swim- ming: A review.Journal of Sports Sciences, 28(3):229–236, 2010

    2010

  73. [73]

    Fernandes, and João Paulo Vilas-Boas

    Phornpot Chainok, Karla De Jesus, Luis Mourão, Pedro Fil- ipe Pereira Fonseca, Rodrigo Zacca, Ricardo J. Fernandes, and João Paulo Vilas-Boas. Biomechanical features of backstroke to breaststroke transition techniques in age-group swimmers.Fron- tiers in Sports and Active Living, 4:802967, 2022

    2022

  74. [74]

    Addressing function approximation error in actor-critic methods

    Scott Fujimoto, Herke van Hoof, and David Meger. Addressing function approximation error in actor-critic methods. InProceed- ings of the 35th International Conference on Machine Learning (ICML), pages 1587–1596, 2018

    2018

  75. [75]

    Current edition: 2017; last stand-alone edition, consolidated into NFPA 2500

    National Fire Protection Association, Quincy, MA.NFPA 1670: Standard on Operations and Training for Technical Search and Rescue Incidents, 2017. Current edition: 2017; last stand-alone edition, consolidated into NFPA 2500. 12

    2017