pith. machine review for the scientific record. sign in

arxiv: 2605.06593 · v1 · submitted 2026-05-07 · 💻 cs.RO · cs.GR· cs.LG

Recognition: unknown

ReActor: Reinforcement Learning for Physics-Aware Motion Retargeting

David M\"uller , Agon Serifi , Sammy Christen , Ruben Grandia , Espen Knoop , Moritz B\"acher
Authors on Pith no claims yet

Pith reviewed 2026-05-08 08:44 UTC · model grok-4.3

classification 💻 cs.RO cs.GRcs.LG
keywords motion retargetingreinforcement learningbilevel optimizationphysics simulationimitation learningrobot morphologyhuman motion transferquadruped robots
0
0 comments X

The pith

A bilevel optimization framework retargets human motions to robots by jointly training a tracking policy with reinforcement learning inside physics simulation.

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

The paper shows how to adapt reference motions from a human body to a robot body by solving a bilevel problem. The upper level tunes a small set of retargeting parameters while the lower level trains a policy that must track the retargeted motion under realistic physics. Because the retargeting parameters are optimized automatically, the method needs only a handful of corresponding rigid bodies and avoids the usual trial-and-error adjustments. The resulting motions contain no foot sliding, self-collisions, or impossible forces, so the learned policies transfer more reliably to the real robot. This matters for anyone who wants to turn human demonstrations into working robot skills without building a new retargeting pipeline for every new morphology.

Core claim

By casting retargeting as the upper level of a bilevel program whose lower level trains a reinforcement-learning policy to track the retargeted trajectory inside a physics simulator, and by supplying an approximate gradient for the upper-level objective, the framework discovers retargeting parameters that preserve the original motion's character while guaranteeing physical feasibility from only a sparse set of semantic rigid-body matches.

What carries the argument

Bilevel optimization that couples an upper-level retargeting parameterization with a lower-level reinforcement-learning policy trained to track the motion under full physics simulation, using an approximate gradient to make the outer optimization tractable.

If this is right

  • Only a sparse set of semantic rigid-body correspondences is required instead of dense or manual feature matching.
  • Retargeting parameters are identified automatically, removing the need for per-motion manual tuning.
  • The adapted motions are free of physical inconsistencies such as foot sliding or dynamically infeasible forces.
  • The motions support robust imitation learning that transfers from simulation to hardware.
  • The same pipeline works across morphologies that differ substantially from the human source, including quadrupeds.

Where Pith is reading between the lines

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

  • Embedding physics simulation inside the retargeting loop may shrink the usual sim-to-real gap for imitation policies.
  • The bilevel structure could be reused for other motion-adaptation tasks where control and morphology changes must be solved together.
  • Once trained, the retargeting parameters might serve as a fast feed-forward adapter for new reference motions without re-optimizing.
  • Extending the approach to online retargeting during execution could let robots adjust human-like motions on the fly.

Load-bearing premise

The approximate gradient computed for the upper-level loss is accurate enough to drive the bilevel optimizer to useful parameter values without trapping it in poor local minima or demanding prohibitive computation.

What would settle it

Running the method on a new robot morphology and observing that the output motion still exhibits foot sliding, self-collisions, or cannot be tracked by the learned policy without further manual fixes would show the central claim is false.

Figures

Figures reproduced from arXiv: 2605.06593 by Agon Serifi, David M\"uller, Espen Knoop, Moritz B\"acher, Ruben Grandia, Sammy Christen.

Figure 1
Figure 1. Figure 1: Physics-aware retargeting of human motion (left) onto two humanoid robots (middle) and a quadruped (right) with varying degrees of freedom and vastly different shapes, sizes, and proportions. Retargeting human kinematic reference motion onto a robot’s morphology remains a formidable challenge. Existing methods often produce physical inconsistencies, such as foot sliding, self-collisions, or dynamically inf… view at source ↗
Figure 2
Figure 2. Figure 2: Bilevel Optimization for Motion Retargeting. As illustrated in view at source ↗
Figure 3
Figure 3. Figure 3: Retargeting Parameterization. A user provides the source and target morphologies in a nominal configuration and defines corresponding rigid-body pairs. A global scale and nominal transformations (TFs) are automatically extracted from the input such that the frames in nominal coordinates align with the corresponding target frames. After this nominal calibration, we introduce parameters in nominal coordinate… view at source ↗
Figure 4
Figure 4. Figure 4: Semantic correspondences and nominal configurations. for the SMPL body model and our target robots: Unitree G1, Lima, and ANYmal D. Implementation Details. We retarget motions onto two humanoids of different scale: Unitree G1 (1.27 m, 35 kg, 29 DoF), and Lima (0.84 m, 16.2 kg, 20 DoF), a custom small-scale robot. For the Unitree G1, we apply the baseline methods using their provided hyperparameter sets. Fo… view at source ↗
Figure 6
Figure 6. Figure 6: Training Curves with and without the Bilevel Optimiza￾tion. Tracking reward, upper-level loss, and parameter update rate during training with and without bilevel optimization. The update rate shows outer-loop convergence and is zero without bilevel opti￾mization, as the parameters remain static. RL Only ReActor view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative Comparison With and Without Bilevel Opti￾mization. With bi-level optimization, results have fewer retargeting artifacts. correspondences with a sparse set of only the root and end effectors. The result deviates further from the source, as would be expected, but the method remains stable. As an additional robustness test against mapping mismatches in the input, we assign the left robot hand to t… view at source ↗
Figure 8
Figure 8. Figure 8: Effect of the External Force Penalty. Mean over motions of the maximum applied forces and torques, the mean upper-level loss L, and the total number of failures as a function of the force penalty weight. ANYmal D (50 kg, 12 DoF). See view at source ↗
read the original abstract

Retargeting human kinematic reference motion onto a robot's morphology remains a formidable challenge. Existing methods often produce physical inconsistencies, such as foot sliding, self-collisions, or dynamically infeasible motions, which hinder downstream imitation learning. We propose a bilevel optimization framework that jointly adapts reference motions to a robot's morphology while training a tracking policy using reinforcement learning. To make the optimization tractable, we derive an approximate gradient for the upper-level loss. Our framework requires only a sparse set of semantic rigid-body correspondences and eliminates the need for manual tuning by identifying optimal values for a parameterization expressive enough to preserve characteristic motion across different embodiments. Moreover, by integrating retargeting directly with physics simulation, we produce physically plausible motions that facilitate robust imitation learning. We validate our method in simulation and on hardware, demonstrating challenging motions for morphologies that differ significantly from a human, including retargeting onto a quadruped.

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 / 2 minor

Summary. The paper proposes ReActor, a bilevel optimization framework for physics-aware motion retargeting that jointly adapts a human reference motion to a target robot morphology (via a parameterization) while training a tracking policy with reinforcement learning. An approximate gradient is derived for the upper-level loss to make the optimization tractable; the method requires only sparse semantic rigid-body correspondences, claims to eliminate manual tuning by automatically identifying optimal parameterization values that preserve characteristic motion, and integrates retargeting directly with physics simulation to produce plausible motions suitable for imitation learning. Validation is reported in simulation and on hardware for morphologies including quadrupeds that differ significantly from human form.

Significance. If the central claims hold, the work would be significant for robotics motion retargeting and imitation learning: it automates what is typically a manual, embodiment-specific process, directly enforces physical plausibility via simulation, and demonstrates results on challenging cross-embodiment cases (e.g., human-to-quadruped). The bilevel formulation that couples retargeting and policy training is a constructive idea that could reduce downstream sim-to-real gaps.

major comments (2)
  1. [§3.2] §3.2 (Bilevel Optimization and Approximate Gradient): The derivation of the approximate gradient for the upper-level loss is load-bearing for the tractability and automation claims, yet the manuscript provides no quantitative validation of its accuracy (e.g., comparison to finite differences on a simplified case, error bounds, or ablation against exact gradients where feasible). Without this, it is unclear whether the optimizer reliably reaches solutions that preserve characteristic motion or merely converges to local minima that still require post-hoc tuning.
  2. [§5] §5 (Experiments, quadruped and hardware results): The reported success on morphologies differing significantly from human form rests on qualitative and aggregate metrics, but lacks direct quantitative comparison to hand-tuned baselines or alternative retargeting methods on load-bearing physical-plausibility measures (foot sliding, self-collision rate, torque feasibility). This weakens the claim that the automatic parameterization eliminates manual intervention.
minor comments (2)
  1. [§3] The notation for the retargeting parameterization (free parameters, bounds, and how semantic correspondences map to them) is introduced piecemeal; a consolidated table or diagram in §3 would improve readability.
  2. [Figures 4-6] Figure captions for qualitative motion results could explicitly state the source human motion, target robot, and any post-processing applied, to allow readers to assess preservation of characteristic features.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and describe the revisions we will incorporate to strengthen the work.

read point-by-point responses

  1. Referee: [§3.2] §3.2 (Bilevel Optimization and Approximate Gradient): The derivation of the approximate gradient for the upper-level loss is load-bearing for the tractability and automation claims, yet the manuscript provides no quantitative validation of its accuracy (e.g., comparison to finite differences on a simplified case, error bounds, or ablation against exact gradients where feasible). Without this, it is unclear whether the optimizer reliably reaches solutions that preserve characteristic motion or merely converges to local minima that still require post-hoc tuning.

    Authors: We agree that explicit quantitative validation of the approximate gradient would strengthen the tractability claims. The derivation enables the bilevel optimization to remain computationally feasible, and the successful retargeting results across embodiments provide indirect empirical support. However, we acknowledge the absence of direct comparisons such as finite-difference checks or error bounds in the current manuscript. In the revised version, we will add an ablation study on simplified cases that compares the approximate gradient to finite differences and exact gradients (where feasible), including quantitative error metrics and analysis of convergence to characteristic-motion-preserving solutions. revision: yes

  2. Referee: [§5] §5 (Experiments, quadruped and hardware results): The reported success on morphologies differing significantly from human form rests on qualitative and aggregate metrics, but lacks direct quantitative comparison to hand-tuned baselines or alternative retargeting methods on load-bearing physical-plausibility measures (foot sliding, self-collision rate, torque feasibility). This weakens the claim that the automatic parameterization eliminates manual intervention.

    Authors: We appreciate the call for stronger quantitative evidence on physical plausibility. The current experiments emphasize feasibility on challenging cross-embodiment cases (including human-to-quadruped) using qualitative demonstrations and aggregate metrics to highlight the automation benefit. We agree that direct comparisons on specific measures would better support the elimination of manual tuning. In the revision, we will add quantitative evaluations against hand-tuned baselines and alternative retargeting methods, reporting foot-sliding distances, self-collision rates, and torque feasibility for the quadruped simulation and hardware results. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external physics simulation and standard bilevel RL optimization.

full rationale

The paper's core contribution is a bilevel optimization that jointly performs motion retargeting and RL policy training, using an approximate gradient derived for tractability. This does not reduce any claimed prediction or result to its inputs by construction, nor does it rely on self-definitional parameters, fitted inputs renamed as predictions, or load-bearing self-citations. The framework explicitly integrates external physics simulation and requires only sparse rigid-body correspondences as input, with the optimization discovering parameterization values rather than presupposing them. No ansatz is smuggled via citation, and no uniqueness theorem from prior author work is invoked to force the method. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Only abstract available, so ledger is necessarily incomplete; main unstated elements are the precise form of the retargeting parameterization and the quality of the approximate gradient.

free parameters (1)
  • retargeting parameterization
    Abstract states that optimal values are identified for an expressive parameterization, implying several tunable or optimized parameters whose count and fitting procedure are unspecified.
axioms (1)
  • domain assumption Physics simulator provides sufficiently accurate dynamics for the target robot morphology
    Central to producing physically plausible motions and facilitating imitation learning.

pith-pipeline@v0.9.0 · 5473 in / 1232 out tokens · 35952 ms · 2026-05-08T08:44:38.634374+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

102 extracted references · 76 canonical work pages · 1 internal anchor

  1. [1]

    and Pons-Moll, Gerard and Black, Michael J

    Mahmood, Naureen and Ghorbani, Nima and Troje, Nikolaus F. and Pons-Moll, Gerard and Black, Michael J. , year =. doi:10.48550/arXiv.1904.03278 , abstract =

    work page doi:10.48550/arxiv.1904.03278 1904

  2. [2]

    IEEE Robot

    A. IEEE Robot. Autom. Lett. , author =. 2021 , file =. doi:10.1109/LRA.2021.3056030 , abstract =

    work page doi:10.1109/lra.2021.3056030 2021

  3. [3]

    OmniRetarget: Interaction- preserving data generation for humanoid whole-body loco-manipulation and scene interaction.arXiv preprint arXiv:2509.26633, 2025

    Yang, Lujie and Huang, Xiaoyu and Wu, Zhen and Kanazawa, Angjoo and Abbeel, Pieter and Sferrazza, Carmelo and Liu, C. Karen and Duan, Rocky and Shi, Guanya , year =. doi:10.48550/arXiv.2509.26633 , abstract =

    work page doi:10.48550/arxiv.2509.26633

  4. [4]

    ACM Trans. Graph. , author =. 2018 , file =. doi:10.1145/3197517.3201311 , abstract =

    work page doi:10.1145/3197517.3201311 2018

  5. [5]

    Proximal Policy Optimization Algorithms

    Schulman, John and Wolski, Filip and Dhariwal, Prafulla and Radford, Alec and Klimov, Oleg , year =. Proximal. doi:10.48550/arXiv.1707.06347 , abstract =

    work page internal anchor Pith review doi:10.48550/arxiv.1707.06347

  6. [6]

    Design and

    Grandia, Ruben and Knoop, Espen and Hopkins, Michael and Wiedebach, Georg and Bishop, Jared and Pickles, Steven and Müller, David and Bächer, Moritz , year =. Design and. Robotics:. doi:10.15607/RSS.2024.XX.103 , abstract =

    work page doi:10.15607/rss.2024.xx.103 2024

  7. [7]

    Hopkins AND Georg Wiedebach AND Jared Bishop AND Steven Pickles AND David Müller AND Moritz Bächer , TITLE =

    Ruben Grandia AND Espen Knoop AND Michael A. Hopkins AND Georg Wiedebach AND Jared Bishop AND Steven Pickles AND David Müller AND Moritz Bächer , TITLE =. Proceedings of Robotics: Science and Systems , YEAR =

  8. [8]

    ACM Trans. Graph. , author =. 2022 , file =. doi:10.1145/3528223.3530110 , abstract =

    work page doi:10.1145/3528223.3530110 2022

  9. [9]

    ACM Trans. Graph. , author =. 2021 , file =. doi:10.1145/3450626.3459670 , abstract =

    work page doi:10.1145/3450626.3459670 2021

  10. [10]

    Beyondmimic: From motion tracking to versatile humanoid control via guided diffusion.arXiv preprint arXiv:2508.08241, 2025

    Liao, Qiayuan and Truong, Takara E. and Huang, Xiaoyu and Gao, Yuman and Tevet, Guy and Sreenath, Koushil and Liu, C. Karen , year =. doi:10.48550/arXiv.2508.08241 , abstract =

    work page doi:10.48550/arxiv.2508.08241

  11. [11]

    Symp. Comput. Anim. , author =. 2024 , file =. doi:10.1111/cgf.15175 , abstract =

    work page doi:10.1111/cgf.15175 2024

  12. [12]

    Cheng, Xuxin and Shi, Kexin and Agarwal, Ananye and Pathak, Deepak , year =. Extreme. doi:10.1109/ICRA57147.2024.10610200 , abstract =

    work page doi:10.1109/icra57147.2024.10610200 2024

  13. [13]

    doi:10.1145/3721238.3730621 , abstract =

    Gat, Inbar and Raab, Sigal and Tevet, Guy and Reshef, Yuval and Bermano, Amit Haim and Cohen-Or, Daniel , year =. doi:10.1145/3721238.3730621 , abstract =

    work page doi:10.1145/3721238.3730621

  14. [14]

    Robust motion in-betweening , year =

    Robust motion in-betweening , volume =. ACM Trans. Graph. , author =. 2020 , file =. doi:10.1145/3386569.3392480 , abstract =

    work page doi:10.1145/3386569.3392480 2020

  15. [15]

    Learning agile and dynamic motor skills for legged robots , volume =. Sci. Robot. , author =. 2019 , file =. doi:10.1126/scirobotics.aau5872 , abstract =

    work page doi:10.1126/scirobotics.aau5872 2019

  16. [16]

    ACM Trans

    Character controllers using motion. ACM Trans. Graph. , author =. 2020 , file =. doi:10.1145/3386569.3392422 , abstract =

    work page doi:10.1145/3386569.3392422 2020

  17. [17]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp

    Luo, Zhengyi and Cao, Jinkun and Winkler, Alexander and Kitani, Kris and Xu, Weipeng , year =. Perpetual. doi:10.1109/ICCV51070.2023.01000 , abstract =

    work page doi:10.1109/iccv51070.2023.01000 2023

  18. [18]

    Serifi, Agon and Grandia, Ruben and Knoop, Espen and Gross, Markus and Bächer, Moritz , year =. Robot. doi:10.1145/3680528.3687626 , abstract =

    work page doi:10.1145/3680528.3687626

  19. [19]

    ACM Trans

    A scalable approach to control diverse behaviors for physically simulated characters , volume =. ACM Trans. Graph. , author =. 2020 , file =. doi:10.1145/3386569.3392381 , abstract =

    work page doi:10.1145/3386569.3392381 2020

  20. [20]

    ACM Trans. Graph. , author =. 2023 , file =. doi:10.1145/3592454 , abstract =

    work page doi:10.1145/3592454 2023

  21. [21]

    Retargeting Matters: General motion retargeting for humanoid motion tracking.arXiv preprint arXiv:2510.02252, 2025

    Araujo, Joao Pedro and Ze, Yanjie and Xu, Pei and Wu, Jiajun and Liu, C. Karen , year =. Retargeting. doi:10.48550/arXiv.2510.02252 , abstract =

    work page doi:10.48550/arxiv.2510.02252

  22. [22]

    doi:10.1145/3610548.3618206 , abstract =

    Lee, Sunmin and Kang, Taeho and Park, Jungnam and Lee, Jehee and Won, Jungdam , year =. doi:10.1145/3610548.3618206 , abstract =

    work page doi:10.1145/3610548.3618206

  23. [23]

    doi:10.1145/3610548.3618255 , abstract =

    Li, Tianyu and Won, Jungdam and Clegg, Alexander and Kim, Jeonghwan and Rai, Akshara and Ha, Sehoon , year =. doi:10.1145/3610548.3618255 , abstract =

    work page doi:10.1145/3610548.3618255

  24. [24]

    IEEE Trans

    Motion. IEEE Trans. Robot. , author =. 2017 , file =. doi:10.1109/TRO.2017.2752711 , abstract =

    work page doi:10.1109/tro.2017.2752711 2017

  25. [25]

    Online and markerless motion retargeting with kinematic constraints , doi =

    Dariush, Behzad and Gienger, Michael and Arumbakkam, Arjun and Goerick, Christian and Zhu, Youding and Fujimura, Kikuo , year =. Online and markerless motion retargeting with kinematic constraints , doi =

  26. [26]

    Darvish, Kourosh and Tirupachuri, Yeshasvi and Romualdi, Giulio and Rapetti, Lorenzo and Ferigo, Diego and Chavez, Francisco Javier Andrade and Pucci, Daniele , year =. Whole-. doi:10.1109/Humanoids43949.2019.9035059 , abstract =

    work page doi:10.1109/humanoids43949.2019.9035059 2019

  27. [27]

    and Hodgins, J.K

    Pollard, N.S. and Hodgins, J.K. and Riley, M.J. and Atkeson, C.G. , year =. Adapting human motion for the control of a humanoid robot , doi =

  28. [28]

    ASME Int

    A. ASME Int. Mech. Eng. Congr. Expo. , author =. 2015 , file =. doi:10.1115/IMECE2014-37700 , abstract =

    work page doi:10.1115/imece2014-37700 2015

  29. [29]

    ACM Trans

    Vibration-minimizing motion retargeting for robotic characters , volume =. ACM Trans. Graph. , author =. 2019 , file =. doi:10.1145/3306346.3323034 , abstract =

    work page doi:10.1145/3306346.3323034 2019

  30. [30]

    doi:10.1145/3550471.3564762 , abstract =

    Kim, Sunwoo and Sorokin, Maks and Lee, Jehee and Ha, Sehoon , year =. doi:10.1145/3550471.3564762 , abstract =

    work page doi:10.1145/3550471.3564762

  31. [31]

    Implicit

    Chen, Xingyu and Wu, Hanyu and Wu, Sikai and Zhou, Mingliang and Xiang, Diyun and Zhang, Haodong , year =. Implicit. doi:10.48550/arXiv.2509.15443 , abstract =

    work page doi:10.48550/arxiv.2509.15443

  32. [32]

    IEEE Trans

    Spatio-. IEEE Trans. Robot. , author =. 2025 , file =. doi:10.1109/TRO.2025.3600123 , abstract =

    work page doi:10.1109/tro.2025.3600123 2025

  33. [33]

    and Serifi, Agon and Grandia, Ruben and Müller, David and Knoop, Espen and Bächer, Moritz , year =

    Alegre, Lucas N. and Serifi, Agon and Grandia, Ruben and Müller, David and Knoop, Espen and Bächer, Moritz , year =. doi:10.1145/3721238.3730656 , abstract =

    work page doi:10.1145/3721238.3730656

  34. [34]

    doi:10.1109/Humanoids57100.2023.10375150 , abstract =

    Yan, Yashuai and Mascaro, Esteve Valls and Lee, Dongheui , year =. doi:10.1109/Humanoids57100.2023.10375150 , abstract =

    work page doi:10.1109/humanoids57100.2023.10375150 2023

  35. [35]

    Villegas, Ruben and Yang, Jimei and Ceylan, Duygu and Lee, Honglak , year =. Neural. doi:10.1109/CVPR.2018.00901 , abstract =

    work page doi:10.1109/cvpr.2018.00901 2018

  36. [36]

    Brit. Mach. Vis. Conf. , author =. 2019 , file =

    2019

  37. [37]

    ACM Trans

    Skeleton-aware networks for deep motion retargeting , volume =. ACM Trans. Graph. , author =. 2020 , file =. doi:10.1145/3386569.3392462 , abstract =

    work page doi:10.1145/3386569.3392462 2020

  38. [38]

    AAAI Conf

    Proc. AAAI Conf. Artif. Intell. , author =. 2022 , file =. doi:10.1609/aaai.v36i3.20274 , abstract =

    work page doi:10.1609/aaai.v36i3.20274 2022

  39. [39]

    IEEE Trans

    Pose-. IEEE Trans. Vis. Comput. Graph. , author =. 2024 , file =. doi:10.1109/TVCG.2023.3277918 , abstract =

    work page doi:10.1109/tvcg.2023.3277918 2024

  40. [40]

    Zhang, Jiaxu and Weng, Junwu and Kang, Di and Zhao, Fang and Huang, Shaoli and Zhe, Xuefei and Bao, Linchao and Shan, Ying and Wang, Jue and Tu, Zhigang , year =. Skinned. doi:10.1109/CVPR52729.2023.01332 , abstract =

    work page doi:10.1109/cvpr52729.2023.01332 2023

  41. [41]

    IEEE Trans

    Skinned. IEEE Trans. Vis. Comput. Graph. , author =. 2025 , file =. doi:10.1109/TVCG.2024.3423426 , abstract =

    work page doi:10.1109/tvcg.2024.3423426 2025

  42. [42]

    IEEE/ASME Trans

    Multicontact. IEEE/ASME Trans. Mechatron. , author =. 2022 , file =. doi:10.1109/TMECH.2022.3152844 , abstract =

    work page doi:10.1109/tmech.2022.3152844 2022

  43. [43]

    doi:10.1145/3757377.3763811 , abstract =

    Chen, Ling-Hao and Zhang, Yuhong and Yin, Zixin and Dou, Zhiyang and Chen, Xin and Wang, Jingbo and Komura, Taku and Zhang, Lei , year =. doi:10.1145/3757377.3763811 , abstract =

    work page doi:10.1145/3757377.3763811

  44. [44]

    Residual

    Yuan, Ye and Kitani, Kris , year =. Residual. Advances in

  45. [45]

    ACM Trans

    Automatic rigging and animation of. ACM Trans. Graph. , author =. 2007 , file =. doi:10.1145/1276377.1276467 , abstract =

    work page doi:10.1145/1276377.1276467 2007

  46. [46]

    Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)

    Clevert, Djork-Arné and Unterthiner, Thomas and Hochreiter, Sepp , year =. Fast and. doi:10.48550/arXiv.1511.07289 , abstract =

    work page Pith review doi:10.48550/arxiv.1511.07289

  47. [47]

    Learning to

    Rudin, Nikita and Hoeller, David and Reist, Philipp and Hutter, Marco , year =. Learning to

  48. [48]

    In: 2024 IEEE/RSJ Int

    He, Tairan and Luo, Zhengyi and Xiao, Wenli and Zhang, Chong and Kitani, Kris and Liu, Changliu and Shi, Guanya , year =. Learning. doi:10.1109/IROS58592.2024.10801984 , abstract =

    work page doi:10.1109/iros58592.2024.10801984 2024

  49. [49]

    Advances in

    Xie, Weiji and Han, Jinrui and Zheng, Jiakun and Li, Huanyu and Liu, Xinzhe and Shi, Jiyuan and Zhang, Weinan and Bai, Chenjia and Li, Xuelong , year =. Advances in

  50. [50]

    Fu, Zipeng and Zhao, Qingqing and Wu, Qi and Wetzstein, Gordon and Finn, Chelsea , year =

  51. [51]

    Allshire, Arthur and Choi, Hongsuk and Zhang, Junyi and McAllister, David and Zhang, Anthony and Kim, Chung Min and Darrell, Trevor and Abbeel, Pieter and Malik, Jitendra and Kanazawa, Angjoo , year =. Visual

  52. [52]

    Retargetting motion to new characters , doi =

    Gleicher, Michael , year =. Retargetting motion to new characters , doi =

  53. [53]

    A hierarchical approach to interactive motion editing for human-like figures , doi =

    Lee, Jehee and Shin, Sung Yong , year =. A hierarchical approach to interactive motion editing for human-like figures , doi =

  54. [54]

    ACM Trans

    A physically-based motion retargeting filter , volume =. ACM Trans. Graph. , author =. 2005 , file =. doi:10.1145/1037957.1037963 , abstract =

    work page doi:10.1145/1037957.1037963 2005

  55. [55]

    Physically based motion transformation , doi =

    Popović, Zoran and Witkin, Andrew , year =. Physically based motion transformation , doi =

  56. [56]

    Motion adaptation based on character shape , volume =. Comput. Animat. Virtual Worlds , author =. 2008 , file =. doi:10.1002/cav.233 , abstract =

    work page doi:10.1002/cav.233 2008

  57. [57]

    ACM Trans

    Spatial relationship preserving character motion adaptation , volume =. ACM Trans. Graph. , author =. 2010 , file =. doi:10.1145/1778765.1778770 , abstract =

    work page doi:10.1145/1778765.1778770 2010

  58. [58]

    j Terry and Zhao, Tong and Graesdal, Bernhard Paus and Kelestemur, Tarik and Wang, Jiuguang and Pang, Tao and Tedrake, Russ , year =

    Yang, Lujie and Suh, H. j Terry and Zhao, Tong and Graesdal, Bernhard Paus and Kelestemur, Tarik and Wang, Jiuguang and Pang, Tao and Tedrake, Russ , year =. Physics-. Robotics:

  59. [59]

    ACM Trans

    Learning body shape variation in physics-based characters , volume =. ACM Trans. Graph. , author =. 2019 , file =. doi:10.1145/3355089.3356499 , abstract =

    work page doi:10.1145/3355089.3356499 2019

  60. [60]

    and DeWeese, John and Maynard, Jordan and van Prooijen, Kees , year =

    Hecker, Chris and Raabe, Bernd and Enslow, Ryan W. and DeWeese, John and Maynard, Jordan and van Prooijen, Kees , year =. Real-time motion retargeting to highly varied user-created morphologies , doi =

  61. [61]

    Simulation of. Comput. Graph. Forum. , author =. 2008 , file =. doi:10.1111/j.1467-8659.2008.01134.x , abstract =

    work page doi:10.1111/j.1467-8659.2008.01134.x 2008

  62. [62]

    Motion capture-driven simulations that hit and react , doi =. Symp. Comput. Anim. , author =. 2002 , file =

    2002

  63. [63]

    Functionality-. Comput. Graph. Forum. , author =. 2021 , file =. doi:10.1111/cgf.14191 , abstract =

    work page doi:10.1111/cgf.14191 2021

  64. [64]

    Simulation and

    Zhang, Yunbo and Gopinath, Deepak and Ye, Yuting and Hodgins, Jessica and Turk, Greg and Won, Jungdam , year =. Simulation and. doi:10.1145/3588432.3591491 , abstract =

    work page doi:10.1145/3588432.3591491

  65. [65]

    Surface based motion retargeting by preserving spatial relationship , doi =

    Liu, Zhiguang and Mucherino, Antonio and Hoyet, Ludovic and Multon, Franck , year =. Surface based motion retargeting by preserving spatial relationship , doi =

  66. [66]

    In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV)

    Villegas, Ruben and Ceylan, Duygu and Hertzmann, Aaron and Yang, Jimei and Saito, Jun , year =. Contact-. doi:10.1109/ICCV48922.2021.00958 , abstract =

    work page doi:10.1109/iccv48922.2021.00958 2021

  67. [67]

    GitHub repository , howpublished =

    ProtoMotions3: An Open-source Framework for Humanoid Simulation and Control , author =. GitHub repository , howpublished =. 2025 , publisher =

    2025

  68. [68]

    ACM Trans

    Generalized biped walking control , volume =. ACM Trans. Graph. , author =. 2010 , file =. doi:10.1145/1778765.1781156 , abstract =

    work page doi:10.1145/1778765.1781156 2010

  69. [69]

    ACM Trans

    Discovery of complex behaviors through contact-invariant optimization , volume =. ACM Trans. Graph. , author =. 2012 , file =. doi:10.1145/2185520.2185539 , abstract =

    work page doi:10.1145/2185520.2185539 2012

  70. [70]

    and Wooten, Wayne L

    Hodgins, Jessica K. and Wooten, Wayne L. and Brogan, David C. and O'Brien, James F. , year =. Animating human athletics , doi =

  71. [71]

    ACM Trans

    Online control of simulated humanoids using particle belief propagation , volume =. ACM Trans. Graph. , author =. 2015 , file =. doi:10.1145/2767002 , abstract =

    work page doi:10.1145/2767002 2015

  72. [72]

    ACM Trans. Graph. , author =. 2007 , file =. doi:10.1145/1276377.1276509 , abstract =

    work page doi:10.1145/1276377.1276509 2007

  73. [73]

    ACM Trans. Graph. , author =. 2022 , file =. doi:10.1145/3550454.3555434 , abstract =

    work page doi:10.1145/3550454.3555434 2022

  74. [74]

    Versatile. Comput. Graph. Forum. , author =. 2025 , file =. doi:10.1111/cgf.70018 , abstract =

    work page doi:10.1111/cgf.70018 2025

  75. [75]

    Dou, Zhiyang and Chen, Xuelin and Fan, Qingnan and Komura, Taku and Wang, Wenping , year =. C·. doi:10.1145/3610548.3618205 , abstract =

    work page doi:10.1145/3610548.3618205

  76. [76]

    A. SIAM J. Optim. , author =. 2023 , file =. doi:10.1137/20M1387341 , abstract =

    work page doi:10.1137/20m1387341 2023

  77. [77]

    Tan, Jie and Zhang, Tingnan and Coumans, Erwin and Iscen, Atil and Bai, Yunfei and Hafner, Danijar and Bohez, Steven and Vanhoucke, Vincent , year =. Sim-to-. Robotics:. doi:10.15607/RSS.2018.XIV.010 , abstract =

    work page doi:10.15607/rss.2018.xiv.010 2018

  78. [78]

    Physics-based. Proc. ACM Comput. Graph. Interact. Tech. , author =. 2023 , file =. doi:10.1145/3606928 , abstract =

    work page doi:10.1145/3606928 2023

  79. [79]

    , year =

    Luo, Zhengyi and Hachiuma, Ryo and Yuan, Ye and Kitani, Kris M. , year =. Dynamics-regulated kinematic policy for egocentric pose estimation , abstract =. Advances in

  80. [80]

    ACM Trans. Graph. , author =. 2020 , file =. doi:10.1145/3414685.3417877 , abstract =

    work page doi:10.1145/3414685.3417877 2020

Showing first 80 references.