arxiv: 2604.10199 · v1 · submitted 2026-04-11 · 💻 cs.GR · cs.LG

Recognition: unknown

FatigueFusion: Latent Space Fusion for Fatigue-Driven Motion Synthesis

Iliana Loi , Konstantinos Moustakas

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:58 UTC · model grok-4.3

classification 💻 cs.GR cs.LG

keywords fatigue-driven motion synthesislatent space fusionhuman motion generationphysics-informed neural networks3D animationbiomechanicsfatigue modeling

0 comments

The pith

A latent space fusion architecture imposes subject-specific fatigue features on non-fatigued motion sequences without requiring any fatigue input data.

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

The paper develops FatigueFusion to generate fatigued human movements by combining fatigue characteristics directly into a latent representation of motion. It uses algorithmic and learned modules to add temporal and spatial fatigue patterns to normal joint sequences, while physics-informed networks model how fatigue intensity increases over time. This setup works end-to-end on standard non-fatigued inputs and control parameters, so it slots into existing animation or simulation systems. If the approach holds, it would support new fatigued motion variants, gradual fatigue buildups, and blended fatigue states for biomechanics, ergonomics, and 3D character animation.

Core claim

FatigueFusion fuses fatigue features in latent space through algorithmic and data-driven modules that impose subject-specific temporal and spatial characteristics on nonfatigued motions, combined with PINN-based simulation of fatigue intensity. This produces novel fatigued movements, intermediate fatigued states, and progressively fatigued motions. Because all modulation occurs in latent space, the system runs directly on non-fatigued joint angle sequences and control parameters without any fatigue input data, enabling seamless use in any motion synthesis pipeline and accurate rendering of fatigue states in animation and simulation.

What carries the argument

The FatigueFusion latent space fusion mechanism, which combines algorithmic rules, data-driven learning, and physics-informed neural network simulation to modulate motion representations with subject-specific fatigue features.

Load-bearing premise

Fatigue features can be accurately imposed and simulated in latent space without any fatigue input data while keeping the resulting motions physically plausible for all synthesis tasks.

What would settle it

A side-by-side comparison where the generated fatigued joint trajectories or timing patterns diverge measurably from real motion-capture recordings of fatigued human subjects would show the method does not produce accurate fatigue effects.

Figures

Figures reproduced from arXiv: 2604.10199 by Iliana Loi, Konstantinos Moustakas.

**Figure 1.** Figure 1: A general overview of our FatigueFusion pipeline consisting of Fatigue Tempo, Fatigue Features and Fatigue Intensity modules. The Figure for 3CC- [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

**Figure 2.** Figure 2: A closer look at the architecture of CVAE and FusionAE comprising the Fatigue Features module. During training the CVAE encodes fatigue profiles in [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: Non-fatigued A: The ground-truth non-fatigued gait motion of subject [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 5.** Figure 5: Figure A: Ground-truth non-fatigued gait of subject A; Figure B: S [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗

**Figure 4.** Figure 4: Non-fatigued C: The ground-truth non-fatigued gait motion of subject [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 6.** Figure 6: The transition from the non-fatigued motion of subject A, [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

**Figure 7.** Figure 7: Comparison of hip adduction, hip, knee and ankle flexion angles, across A) synthesized (output of FatigueFusion model - red line), ground-truth [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗

**Figure 8.** Figure 8: Comparison of Fatigue-PINN results with the ones of the proposed FatigueFusion. Both frameworks are fed with the non-fatigued motion of subject [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗

**Figure 9.** Figure 9: Qualitative comparison of the output of the FatigueFusion framework [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗

read the original abstract

Investigating the impact of fatigue on human physiological function and motor behavior is crucial for developing biomechanics and medical applications aimed at mitigating fatigue, reducing injury risk, and creating sophisticated ergonomic designs, as well as for producing physically-plausible 3D animation sequences. While the former has a prominent position in state-of-the-art literature, fatigue-driven motion generation is still an underexplored area. In this study, we present FatigueFusion, a deep-learning architecture for the fusion of fatigue features within a latent representation space, enabling the creation of a variation of novel fatigued movements, intermediate fatigued states, and progressively fatigued motions. Unlike existing approaches that focus on imitating the effects of fatigue accumulation in motion patterns, our framework incorporates algorithmic and data-driven modules to impose subject-specific temporal and spatial fatigue features on nonfatigued motions, while leveraging PINN-based techniques to simulate fatigue intensity. Since all motion modulation tasks are taking place in latent space, FatigueFusion offers an end-to-end architecture that operates directly on non-fatigued joint angle sequences and control parameters, allowing seamless integration into any motion synthesis pipeline, without relying on fatigue input data. Overall, our framework can be employed for various fatigue-driven synthesis tasks, such as fatigue profile transfer and fusion, while it also provides a solution for accurate rendering of the human fatigue state in both animation and simulation pipelines.

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

FatigueFusion describes a latent-space fusion method with PINNs to generate fatigued motions from non-fatigued inputs without fatigue data, but the abstract gives no numbers to check if it works.

read the letter

The main thing here is that the authors introduce FatigueFusion, a named architecture that fuses fatigue features in latent space using algorithmic modules, data-driven components, and PINN-based intensity simulation to turn non-fatigued joint-angle sequences into fatigued ones, including variations and progressive states, all without requiring fatigue input data at runtime. It is designed as an end-to-end system that slots directly into existing motion synthesis pipelines via control parameters and joint angles. This targets an area the abstract correctly flags as underexplored compared to other fatigue work in biomechanics and ergonomics. The practical framing around subject-specific temporal and spatial imposition and seamless integration is a reasonable strength for graphics and animation use cases. The stress-test note finds no internal contradictions in the description, which aligns with what is visible. The soft spots are straightforward: the abstract contains zero quantitative results, error metrics, ablations, or comparisons, so the central claims about physical plausibility and accurate fatigue imposition rest on unshown evidence. If the full paper has only descriptive claims or weak validation, the assumption that the latent fusion plus PINN avoids artifacts or circular fitting becomes the load-bearing part. This is for readers in computer graphics and biomechanics who build or extend motion tools and want ideas for adding fatigue effects. Someone working on character animation or ergonomic simulation could extract the high-level architecture as a starting point. It deserves a serious referee because the problem is real and the technique combination is specific enough to warrant review, even if the authors will need to supply solid experiments and comparisons in revision. I would send it out for peer review.

Referee Report

1 major / 1 minor

Summary. The paper proposes FatigueFusion, a deep-learning architecture that performs latent-space fusion of fatigue features to synthesize novel fatigued motions, intermediate fatigue states, and progressively fatigued motions from non-fatigued joint-angle sequences. It combines algorithmic and data-driven modules to impose subject-specific temporal and spatial fatigue features on non-fatigued motions and employs PINN-based techniques to simulate fatigue intensity, all without requiring any fatigue input data, thereby enabling end-to-end integration into arbitrary motion synthesis pipelines for tasks such as fatigue profile transfer.

Significance. If the central claims hold, the work would constitute a meaningful contribution to the underexplored domain of fatigue-driven 3D motion synthesis in computer graphics and biomechanics. It offers a practical route to generating physically plausible fatigued animations and simulations directly from standard non-fatigued inputs, with potential downstream value for injury-risk modeling, ergonomic design, and medical applications.

major comments (1)

[Abstract] Abstract: the central assertion that the framework can accurately impose subject-specific fatigue features and simulate intensity via PINNs while preserving physical plausibility, all without any fatigue input data, is unsupported by quantitative results, error metrics, ablation studies, or validation experiments. This absence is load-bearing because the paper's primary contribution rests on the empirical effectiveness of the latent-space fusion and PINN modules.

minor comments (1)

The description of the latent-space fusion mechanism and the precise manner in which PINN-based simulation is integrated could benefit from additional architectural diagrams or pseudocode to clarify data flow between the algorithmic, data-driven, and physics-informed components.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thoughtful review and for highlighting the need to strengthen the empirical grounding of our claims. We address the major comment below and will revise the manuscript accordingly to provide clearer quantitative support for the effectiveness of the latent-space fusion and PINN modules.

read point-by-point responses

Referee: [Abstract] Abstract: the central assertion that the framework can accurately impose subject-specific fatigue features and simulate intensity via PINNs while preserving physical plausibility, all without any fatigue input data, is unsupported by quantitative results, error metrics, ablation studies, or validation experiments. This absence is load-bearing because the paper's primary contribution rests on the empirical effectiveness of the latent-space fusion and PINN modules.

Authors: We agree that the abstract and main text would be strengthened by explicit quantitative validation. The current manuscript primarily demonstrates the approach through qualitative motion visualizations, intermediate fatigue state synthesis, and progressive fatigue sequences, along with architectural descriptions of the algorithmic, data-driven, and PINN components. To directly address this point, the revised version will include: (1) quantitative error metrics such as joint-angle MSE and velocity deviation between synthesized fatigued motions and available reference data; (2) ablation studies isolating the contributions of the latent fusion module and the PINN-based intensity simulation; and (3) additional validation experiments assessing physical plausibility via biomechanical proxies (e.g., center-of-mass stability and joint torque consistency) and user studies for perceptual realism. These additions will be placed in a new experimental section and will support the claims of subject-specific feature imposition without requiring fatigue input data during inference. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The provided abstract and context describe a deep-learning architecture for latent-space fusion of fatigue features using algorithmic modules and PINN-based intensity simulation. No equations, loss formulations, training procedures, or derivation chains are visible in the text. Without any quotable steps that reduce a claimed prediction or first-principles result to fitted inputs or self-citations by construction, no circularity patterns (self-definitional, fitted-input-called-prediction, etc.) can be exhibited. The framework is presented as operating directly on non-fatigued inputs, but this is an architectural claim rather than a mathematical derivation that collapses to its own data. Per hard rules, absence of specific reducible steps requires score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only abstract available; no explicit free parameters, axioms, or invented entities are stated. The approach implicitly assumes that fatigue can be represented as additive features in latent space and that PINNs can simulate intensity without direct supervision.

pith-pipeline@v0.9.0 · 5539 in / 1195 out tokens · 39721 ms · 2026-05-10T15:58:49.807800+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

57 extracted references · 5 canonical work pages · 2 internal anchors

[1]

The psychological and physiological health effects of fatigue,

A. M. Lock, D. L. Bonetti, and A. Campbell, “The psychological and physiological health effects of fatigue,”Occupational medicine, vol. 68, no. 8, pp. 502–511, 2018

2018
[2]

Unraveling the interplay between mental workload, occupational fatigue, physiological responses and cognitive performance in office workers,

N. Mahdavi, L. Tapak, E. Darvishi, A. Doosti-Irani, and M. Shafiee Mot- lagh, “Unraveling the interplay between mental workload, occupational fatigue, physiological responses and cognitive performance in office workers,”Scientific Reports, vol. 14, no. 1, p. 17866, 2024

2024
[3]

Match-related fatigue in basketball: A systematic review,

M. Pernigoni, D. Ferioli, J. Calleja-Gonz ´alez, P. Sansone, A. Tessitore, A. T. Scanlan, and D. Conte, “Match-related fatigue in basketball: A systematic review,”Journal of sports sciences, vol. 42, no. 18, pp. 1727– 1758, 2024

2024
[4]

Effect of fatigue on kinematics, kinetics and muscle activities of lower limbs during gait,

L. Zhang, Y . Yan, G. Liu, B. Han, J. Fei, and Y . Zhang, “Effect of fatigue on kinematics, kinetics and muscle activities of lower limbs during gait,” Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine, vol. 236, no. 9, pp. 1365–1374, 2022

2022
[5]

Differential effects of fatigue on movement variability,

N. Cortes, J. Onate, and S. Morrison, “Differential effects of fatigue on movement variability,”Gait & posture, vol. 39, no. 3, pp. 888–893, 2014

2014
[6]

Fatigue risk management by active fatigue monitoring system: A layer of defense in fatigue mitigation,

P. K. Dhawan, A. Yugendar, A. K. Singh, and C. Shekhar, “Fatigue risk management by active fatigue monitoring system: A layer of defense in fatigue mitigation,” inAsian Mining Congress. Springer, 2025, pp. 721–736

2025
[7]

Fatigue risk management systems,

P. H. Gander, L. J. Wu, M. van den Berg, A. Lamp, L. Hoeg, and G. Belenky, “Fatigue risk management systems,”Principles and practice of sleep medicine, pp. 697–707, 2017

2017
[8]

Monitoring fatigue status in elite team-sport athletes: implications for practice,

R. T. Thorpe, G. Atkinson, B. Drust, and W. Gregson, “Monitoring fatigue status in elite team-sport athletes: implications for practice,” International journal of sports physiology and performance, vol. 12, no. s2, pp. S2–27, 2017

2017
[9]

Investigating virtual reality for al- leviating human-computer interaction fatigue: A multimodal assessment and comparison with flat video,

X. Wang, J. Qu, L. Bu, and S. Zhu, “Investigating virtual reality for al- leviating human-computer interaction fatigue: A multimodal assessment and comparison with flat video,”IEEE Transactions on Visualization and Computer Graphics, 2025

2025
[10]

A new over- loading fatigue model for ergonomic risk assessment with application to human-robot collaboration,

M. Lorenzini, W. Kim, E. De Momi, and A. Ajoudani, “A new over- loading fatigue model for ergonomic risk assessment with application to human-robot collaboration,” in2019 International Conference on Robotics and Automation (ICRA). IEEE, 2019, pp. 1962–1968

2019
[11]

A multimodal fatigue detection system using semg and imu signals with a hybrid cnn-lstm-attention model,

S. Hwang, N. Kwon, D. Lee, J. Kim, S. Yang, I. Youn, H.-J. Moon, J.-K. Sung, and S. Han, “A multimodal fatigue detection system using semg and imu signals with a hybrid cnn-lstm-attention model,”Sensors, vol. 25, no. 11, p. 3309, 2025

2025
[12]

Physical fatigue detection from gait cycles via a multi-task recurrent neural network,

K. Aoki, H. Nishikawa, Y . Makihara, D. Muramatsu, N. Takemura, and Y . Yagi, “Physical fatigue detection from gait cycles via a multi-task recurrent neural network,”IEEE Access, vol. 9, pp. 127 565–127 575, 2021

2021
[13]

Assessment of cognitive fatigue from gait cycle analysis,

H. R. Pavel, E. Karim, A. Jaiswal, S. Acharya, G. Nale, M. Theofanidis, and F. Makedon, “Assessment of cognitive fatigue from gait cycle analysis,”Technologies, vol. 11, no. 1, p. 18, 2023

2023
[14]

Synthetic 3d full-body skeletal motion from 2d paths using rnn with lstm cells and linear networks,

D. Carneros-Prado, C. C. Dobrescu, L. Caba ˜nero, L. Villa, Y . V . Altamirano-Flores, I. H. Lopez-Nava, I. Gonz ´alez, J. Fontecha, and R. Herv´as, “Synthetic 3d full-body skeletal motion from 2d paths using rnn with lstm cells and linear networks,”Computers in Biology and Medicine, vol. 180, p. 108943, 2024

2024
[15]

A two-part transformer network for controllable motion synthesis,

S. Hou, H. Tao, H. Bao, and W. Xu, “A two-part transformer network for controllable motion synthesis,”IEEE Transactions on Visualization and Computer Graphics, vol. 30, no. 8, pp. 5047–5062, 2023

2023
[16]

Tl- control: Trajectory and language control for human motion synthesis,

W. Wan, Z. Dou, T. Komura, W. Wang, D. Jayaraman, and L. Liu, “Tl- control: Trajectory and language control for human motion synthesis,” in European Conference on Computer Vision. Springer, 2024, pp. 37–54

2024
[17]

Motiondiffuse: Text-driven human motion generation with diffusion model,

M. Zhang, Z. Cai, L. Pan, F. Hong, X. Guo, L. Yang, and Z. Liu, “Motiondiffuse: Text-driven human motion generation with diffusion model,”IEEE transactions on pattern analysis and machine intelligence, vol. 46, no. 6, pp. 4115–4128, 2024

2024
[18]

Guess: Gradually enriching synthesis for text-driven human motion generation,

X. Gao, Y . Yang, Z. Xie, S. Du, Z. Sun, and Y . Wu, “Guess: Gradually enriching synthesis for text-driven human motion generation,”IEEE Transactions on Visualization and Computer Graphics, vol. 30, no. 12, pp. 7518–7530, 2024

2024
[19]

Biomechanics-guided residual approach to generalizable human motion generation and estimation,

Z. Kang, X. Wang, and Y . Mu, “Biomechanics-guided residual approach to generalizable human motion generation and estimation,”
[20]

Available: https://arxiv.org/abs/2503.06151

[Online]. Available: https://arxiv.org/abs/2503.06151

work page arXiv
[21]

Morph: A motion-free physics optimization framework for human motion generation,

Z. Li, M. Luo, R. Hou, X. Zhao, H. Liu, H. Chang, Z. Liu, and C. Li, “Morph: A motion-free physics optimization framework for human motion generation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 14 580–14 589

2025
[22]

Fatigue-pinn: Physics-informed fatigue-driven motion modulation and synthesis,

I. Loi and K. Moustakas, “Fatigue-pinn: Physics-informed fatigue-driven motion modulation and synthesis,”IEEE Access, vol. 13, pp. 109 378– 109 398, 2025

2025
[23]

Discovering fatigued movements for virtual character animation,

N. Cheema, R. Xu, N. H. Kim, P. H ¨am¨al¨ainen, V . Golyanik, M. Haber- mann, C. Theobalt, and P. Slusallek, “Discovering fatigued movements for virtual character animation,” inSIGGRAPH Asia 2023 Conference Papers. New York, NY , USA: Association for Computing Machinery, 2023

2023
[24]

Duo- gait: A gait dataset for walking under dual-task and fatigue conditions with inertial measurement units,

L. Zhou, E. Fischer, C. M. Brahms, U. Granacher, and B. Arnrich, “Duo- gait: A gait dataset for walking under dual-task and fatigue conditions with inertial measurement units,”Scientific Data, vol. 10, no. 1, p. 543, 2023

2023
[25]

A dynamical model of muscle activation, fatigue, and recovery,

J. Z. Liu, R. W. Brown, and G. H. Yue, “A dynamical model of muscle activation, fatigue, and recovery,”Biophysical journal, vol. 82, no. 5, pp. 2344–2359, 2002

2002
[26]

A theoretical approach for modeling peripheral muscle fatigue and recovery,

T. Xia and L. A. Frey Law, “A theoretical approach for modeling peripheral muscle fatigue and recovery,”Journal of Biomechanics, vol. 41, no. 14, pp. 3046–3052, 2008

2008
[27]

A three-compartment muscle fatigue model accurately predicts joint-specific maximum en- durance times for sustained isometric tasks,

L. A. Frey Law, J. M. Looft, and J. Heitsman, “A three-compartment muscle fatigue model accurately predicts joint-specific maximum en- durance times for sustained isometric tasks,”Journal of Biomechanics, vol. 45, no. 10, pp. 1803–8, 2012

2012
[28]

Machine learning approaches for 3d motion synthesis and musculoskeletal dynamics estimation: A survey,

I. Loi, E. I. Zacharaki, and K. Moustakas, “Machine learning approaches for 3d motion synthesis and musculoskeletal dynamics estimation: A survey,”IEEE transactions on visualization and computer graphics, vol. 30, no. 8, 08 2023

2023
[29]

A deep learning framework for character motion synthesis and editing,

D. Holden, J. Saito, and T. Komura, “A deep learning framework for character motion synthesis and editing,”ACM Transactions on Graphics (ToG), vol. 35, no. 4, pp. 1–11, 2016

2016
[30]

Hi- erarchical learning recurrent neural networks for 3d motion synthesis,

D. Zhou, C. Guo, R. Liu, C. Che, D. Yang, Q. Zhang, and X. Wei, “Hi- erarchical learning recurrent neural networks for 3d motion synthesis,” International Journal of Machine Learning and Cybernetics, vol. 12, no. 8, pp. 2255–2267, 2021

2021
[31]

Action-conditioned 3d human motion synthesis with transformer vae,

M. Petrovich, M. J. Black, and G. Varol, “Action-conditioned 3d human motion synthesis with transformer vae,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 985–10 995

2021
[32]

A unified 3d human motion synthesis model via conditional variational auto-encoder,

Y . Cai, Y . Wang, Y . Zhu, T.-J. Cham, J. Cai, J. Yuan, J. Liu, C. Zheng, S. Yan, H. Dinget al., “A unified 3d human motion synthesis model via conditional variational auto-encoder,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 11 645–11 655

2021
[33]

Ganimator: Neural motion synthesis from a single sequence,

P. Li, K. Aberman, Z. Zhang, R. Hanocka, and O. Sorkine-Hornung, “Ganimator: Neural motion synthesis from a single sequence,”ACM Transactions on Graphics (TOG), vol. 41, no. 4, pp. 1–12, 2022

2022
[34]

Motiondreamer: One-to-many motion synthesis with local- ized generative masked transformer,

Y . Wang, C. Guo, Y . Mu, M. G. Javed, X. Zuo, J. Lu, H. Jiang, and L. Cheng, “Motiondreamer: One-to-many motion synthesis with local- ized generative masked transformer,”arXiv preprint arXiv:2504.08959, 2025. JOURNAL OF LATEX CLASS FILES, VOL. .., NO. .., .. 2026 13

work page arXiv 2025
[35]

Generating distance-aware human-to-human interaction motions from text guid- ance,

J.-Q. Zhang, J.-J. Wang, F.-L. Zhang, and M. Wang, “Generating distance-aware human-to-human interaction motions from text guid- ance,”IEEE Transactions on Visualization and Computer Graphics, 2026

2026
[36]

Difusion: Flexible stylized motion generation using digest-and-fusion scheme,

Y . Wang, H. Mo, and C. Gao, “Difusion: Flexible stylized motion generation using digest-and-fusion scheme,”IEEE Transactions on Vi- sualization and Computer Graphics, 2025

2025
[37]

Ls-gan: Human motion synthesis with latent-space gans,

A. Amballa, G. Akkinapalli, and V . Muralikrishnan, “Ls-gan: Human motion synthesis with latent-space gans,” inProceedings of the Winter Conference on Applications of Computer Vision, 2025, pp. 326–335

2025
[38]

Unimotion: Unifying 3d human motion synthesis and understanding,

C. Li, J. Chibane, Y . He, N. Pearl, A. Geiger, and G. Pons-Moll, “Unimotion: Unifying 3d human motion synthesis and understanding,” in2025 International Conference on 3D Vision (3DV). IEEE, 2025, pp. 240–249

2025
[39]

Tedi: Temporally- entangled diffusion for long-term motion synthesis,

Z. Zhang, R. Liu, R. Hanocka, and K. Aberman, “Tedi: Temporally- entangled diffusion for long-term motion synthesis,” inACM SIG- GRAPH 2024 Conference Papers, 2024, pp. 1–11

2024
[40]

Mofusion: A framework for denoising-diffusion-based motion synthesis,

R. Dabral, M. H. Mughal, V . Golyanik, and C. Theobalt, “Mofusion: A framework for denoising-diffusion-based motion synthesis,” inPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 9760–9770

2023
[41]

Parc: Physics-based aug- mentation with reinforcement learning for character controllers,

M. Xu, Y . Shi, K. Yin, and X. B. Peng, “Parc: Physics-based aug- mentation with reinforcement learning for character controllers,” in Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers, 2025, pp. 1–11

2025
[42]

Pdp: Physics-based character animation via diffusion policy,

T. E. Truong, M. Piseno, Z. Xie, and K. Liu, “Pdp: Physics-based character animation via diffusion policy,” inSIGGRAPH Asia 2024 Conference Papers, 2024, pp. 1–10

2024
[43]

Moconvq: Uni- fied physics-based motion control via scalable discrete representations,

H. Yao, Z. Song, Y . Zhou, T. Ao, B. Chen, and L. Liu, “Moconvq: Uni- fied physics-based motion control via scalable discrete representations,” ACM Transactions on Graphics (TOG), vol. 43, no. 4, pp. 1–21, 2024

2024
[44]

Physics-based motion imitation with adversarial differential discrimina- tors,

Z. Zhang, S. Bashkirov, D. Yang, Y . Shi, M. Taylor, and X. B. Peng, “Physics-based motion imitation with adversarial differential discrimina- tors,” inProceedings of the SIGGRAPH Asia 2025 Conference Papers, 2025, pp. 1–12

2025
[45]

Flex- motion: Lightweight, physics-aware, and controllable human motion generation,

A. Tashakori, A. Tashakori, G. Yang, Z. J. Wang, and P. Servati, “Flex- motion: Lightweight, physics-aware, and controllable human motion generation,”arXiv preprint arXiv:2501.16778, 2025

work page arXiv 2025
[46]

Full-body musculoskeletal model for muscle-driven sim- ulation of human gait,

A. Rajagopal, C. L. Dembia, M. S. DeMers, D. D. Delp, J. L. Hicks, and S. L. Delp, “Full-body musculoskeletal model for muscle-driven sim- ulation of human gait,”IEEE transactions on biomedical engineering, vol. 63, no. 10, pp. 2068–2079, 2016

2068
[47]

Cholletet al.(2015) Keras

F. Cholletet al.(2015) Keras. [Online]. Available: https://github.com/ fchollet/keras

2015
[48]

Adam: A method for stochastic optimization,

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,”
[49]

Adam: A Method for Stochastic Optimization

[Online]. Available: https://arxiv.org/abs/1412.6980

work page internal anchor Pith review Pith/arXiv arXiv
[50]

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks

A. Radford, L. Metz, and S. Chintala, “Unsupervised representation learning with deep convolutional generative adversarial networks,”arXiv preprint arXiv:1511.06434, 2015

work page internal anchor Pith review arXiv 2015
[51]

Fatigue reduces the complexity of knee extensor torque fluctuations during maximal and submaximal intermittent isometric contractions in man,

J. Pethick, S. L. Winter, and M. Burnley, “Fatigue reduces the complexity of knee extensor torque fluctuations during maximal and submaximal intermittent isometric contractions in man,”The Journal of Physiology, vol. 593, no. 8, pp. 2085–2096, 2015

2085
[52]

Muscle weakness, fatigue, and torque variability: effects of age and mobility status,

J. A. Kent-Braun, D. M. Callahan, J. L. Fay, S. A. Foulis, and J. P. Buonaccorsi, “Muscle weakness, fatigue, and torque variability: effects of age and mobility status,”Muscle & nerve, vol. 49, no. 2, pp. 209–217, 2014

2014
[53]

Opensense: An open-source toolbox for inertial-measurement-unit-based measurement of lower extremity kinematics over long durations,

M. Al Borno, J. O’Day, V . Ibarra, J. Dunne, A. Seth, A. Habib, C. Ong, J. Hicks, S. Uhlrich, and S. Delp, “Opensense: An open-source toolbox for inertial-measurement-unit-based measurement of lower extremity kinematics over long durations,”Journal of neuroengineering and re- habilitation, vol. 19, no. 1, p. 22, 2022

2022
[54]

A history of the unity game engine,

J. K. Haas, “A history of the unity game engine,” 2014

2014
[55]

Action2motion: Conditioned generation of 3d human motions,

C. Guo, X. Zuo, S. Wang, S. Zou, Q. Sun, A. Deng, M. Gong, and L. Cheng, “Action2motion: Conditioned generation of 3d human motions,” inProceedings of the 28th ACM international conference on multimedia, 2020, pp. 2021–2029

2020
[56]

Health & gait: a dataset for gait-based analysis,

J. Zafra-Palma, N. Mar ´ın-Jim´enez, J. Castro-Pi ˜nero, M. Cuenca-Garc´ıa, R. Mu ˜noz-Salinas, and M. J. Mar ´ın-Jim´enez, “Health & gait: a dataset for gait-based analysis,”Scientific Data, vol. 12, no. 1, p. 44, 2025

2025
[57]

A multi-sensor gait dataset collected under non-standardized dual-task conditions,

Y . Liao, J. Cao, L. Yu, J. Xiang, and Y . Zhao, “A multi-sensor gait dataset collected under non-standardized dual-task conditions,”Scientific data, vol. 12, no. 1, p. 1121, 2025

2025