pith. sign in

arxiv: 2606.18861 · v1 · pith:WADY5RYFnew · submitted 2026-06-17 · 💻 cs.CV · cs.AI

URDF Synthesis from RGB-D Sequences via Differentiable Joint Inference and Energy-Consistent Verification

Xinze Zhang This is my paper

Pith reviewed 2026-06-26 21:35 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords URDF reconstructionRGB-D sequencesarticulated objectsdifferentiable joint inferenceenergy-consistent verificationkinematic parameterssimulation driftmanipulation tasks
0
0 comments X

The pith

KinemaForge reconstructs accurate, energy-consistent URDF models from RGB-D sequences by jointly inferring shape and kinematics with a differentiable verifier.

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

The paper aims to close two gaps in articulated object reconstruction from sensor data: the decoupling of part-level geometry from kinematic estimation and the production of models that violate energy conservation, causing drift in simulators. KinemaForge addresses this with a pipeline that uses a kinematic constraint graph, a differentiable screw-axis solver, and an energy residual loss to jointly infer and verify the model. A sympathetic reader would care because accurate digital twins of objects would allow robots to plan and execute tasks more reliably in simulation before real-world deployment. If the approach works, it would mean shorter RGB-D sequences suffice for creating stable URDFs that maintain physical properties over long rollouts.

Core claim

KinemaForge is a constraint-driven pipeline that jointly infers part-level shape, joint topology, and joint parameters from short RGB-D sequences and validates the result against an energy-consistent verifier built on differentiable rigid-body dynamics. The pipeline introduces three components: a kinematic constraint graph that encodes joint-part incidences as soft edges; a differentiable screw-axis solver that backpropagates from rendered observations through Featherstone's articulated-body algorithm to joint parameters; and an energy residual loss that penalises non-physical free responses of the reconstructed model. This produces URDFs with lower joint-axis error, less simulation drift, a

What carries the argument

the energy residual loss that penalises non-physical free responses of the reconstructed model

If this is right

  • Joint-axis error falls to 2.83 degrees, improving 37.4 percent over PARIS and 46.6 percent over Ditto.
  • Long-horizon simulation drift drops 64 percent versus PARIS in 50-second rollouts.
  • Closed-loop manipulation success rate rises 14.6 percentage points over Ditto.
  • The reconstructed models better satisfy energy conservation during free motion.

Where Pith is reading between the lines

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

  • The technique might apply to other reconstruction problems where physical invariants need enforcement.
  • It could support automatic dataset generation for training robot policies on articulated objects.
  • If energy consistency transfers to contact scenarios, it reduces the gap between simulation and reality in robotics.
  • Extensions to noisy or incomplete RGB-D data would test robustness in real settings.

Load-bearing premise

That minimizing the energy residual loss on free responses of the reconstructed model will produce URDFs that remain energy-consistent and drift-free under the contact and control conditions encountered in downstream manipulation tasks.

What would settle it

Observe whether the reconstructed URDF exhibits significant energy drift or increased simulation instability when used in a physics engine for contact-rich tasks, compared to models from other methods.

Figures

Figures reproduced from arXiv: 2606.18861 by Xinze Zhang.

Figure 1
Figure 1. Figure 1: Overview of the KinemaForge pipeline. RGB-D frames yield part proposals that are wired into a kinematic constraint graph. A [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Radar comparison across seven evaluation axes. KinemaForge dominates on accuracy, simulation fidelity and physical-consistency axes [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: Per-category performance on PartNet-Mobility. (a) Joint-axis error in degrees; (b) Part IoU. KinemaForge attains the lowest error in every [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Long-horizon drift in PyBullet. KinemaForge drifts roughly 6 [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Cross-dataset part IoU. KinemaForge generalises from the synthetic PartNet-Mobility regime to AKB-48 and to internal RGB-D captures [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Optimisation curves. The full pipeline converges in roughly half the iterations needed by the variant without the constraint graph, while [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Energy residual per (method, category) pair. KinemaForge’s worst case (0 [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
read the original abstract

Reconstructing simulation-ready digital twins of articulated objects from sensor observations remains constrained by two persistent gaps: (i) part-level geometric reconstruction is decoupled from kinematic-parameter estimation, and (ii) the recovered models often violate basic dynamic invariants such as energy conservation, leading to drift when the URDF is replayed in physics simulators. We present KinemaForge, a constraint-driven pipeline that jointly infers part-level shape, joint topology, and joint parameters from short RGB-D sequences and validates the result against an energy-consistent verifier built on differentiable rigid-body dynamics. The pipeline introduces three components: a kinematic constraint graph that encodes joint-part incidences as soft edges; a differentiable screw-axis solver that backpropagates from rendered observations through Featherstone's articulated-body algorithm to joint parameters; and an energy residual loss that penalises non-physical free responses of the reconstructed model. Across five PartNet-Mobility categories and an internal RGB-D benchmark, KinemaForge reduces the average joint-axis error from 4.52 degrees to 2.83 degrees (-37.4%) over the strongest geometric baseline (PARIS) and from 5.30 degrees to 2.83 degrees (-46.6%) over the interaction-based Ditto baseline, lowers long-horizon simulation drift by 64% (vs. PARIS) over 50 s rollouts, and yields URDFs whose closed-loop manipulation success rate improves by 14.6 percentage points over Ditto in our preliminary evaluation. Code and reconstruction data will be released upon acceptance.

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

1 major / 1 minor

Summary. The manuscript presents KinemaForge, a constraint-driven pipeline for reconstructing simulation-ready URDFs of articulated objects from short RGB-D sequences. It jointly infers part-level shape, joint topology, and parameters via a kinematic constraint graph encoding joint-part incidences as soft edges, a differentiable screw-axis solver that backpropagates from rendered observations through Featherstone's articulated-body algorithm, and an energy residual loss that penalizes non-physical free responses. On five PartNet-Mobility categories and an internal RGB-D benchmark, it reports joint-axis error reduced to 2.83° (–37.4% vs. PARIS, –46.6% vs. Ditto), 64% lower long-horizon drift over 50 s rollouts vs. PARIS, and +14.6 pp closed-loop manipulation success vs. Ditto, with code and data to be released.

Significance. If the energy-consistent verifier produces URDFs that remain drift-free under the contact and control conditions of downstream tasks, the work would meaningfully advance reconstruction of physically plausible digital twins by coupling differentiable dynamics with geometric inference. The quantitative gains in kinematic accuracy, simulation stability, and task performance, together with the planned code release, would strengthen reproducibility and practical utility in robotics and computer vision.

major comments (1)
  1. [Abstract] Abstract (energy residual loss component): The loss penalizes non-physical free responses, yet the reported 64% drift reduction and 14.6 pp manipulation success gains are measured on contact-rich, closed-loop tasks. No indication is given that the verifier is applied or regularized on trajectories containing contacts or external wrenches; if the learned parameters satisfy energy conservation only in the absence of contacts, the headline improvements may not generalize and the central claim of energy-consistent URDFs for manipulation would require additional support.
minor comments (1)
  1. The abstract refers to a 'preliminary evaluation' for the manipulation success metric; specifying the number of trials, task diversity, and whether the same energy loss weighting was used would improve clarity without altering the central claims.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the energy residual loss and its relation to the reported results. We respond point-by-point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract (energy residual loss component): The loss penalizes non-physical free responses, yet the reported 64% drift reduction and 14.6 pp manipulation success gains are measured on contact-rich, closed-loop tasks. No indication is given that the verifier is applied or regularized on trajectories containing contacts or external wrenches; if the learned parameters satisfy energy conservation only in the absence of contacts, the headline improvements may not generalize and the central claim of energy-consistent URDFs for manipulation would require additional support.

    Authors: We agree that the energy residual loss is formulated exclusively on free responses (no contacts or external wrenches). This isolates enforcement of energy conservation as a fundamental invariant, which is a prerequisite for stable long-term behavior even when contacts occur. The 50 s drift rollouts and closed-loop manipulation evaluations are performed inside a full physics simulator whose dynamics already incorporate contacts and gravity; the energy-consistent parameters reduce spurious dissipation that would otherwise amplify under those conditions. Nevertheless, the manuscript does not demonstrate explicit regularization on contact-containing trajectories. In revision we will (i) qualify the abstract and method sections to state the free-response scope of the loss, (ii) add a short discussion explaining why energy conservation remains beneficial under contacts, and (iii) include a brief ablation regularizing on a small set of simulated contact trajectories if space permits. These changes directly address the concern while preserving the core technical contribution. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation remains self-contained

full rationale

The abstract and described pipeline introduce a kinematic constraint graph, differentiable screw-axis solver via Featherstone's algorithm, and an energy residual loss applied to free responses. No equations, self-citations, or fitted parameters are shown that reduce any claimed output (joint-axis error, drift reduction, or manipulation success) to the inputs by construction. The energy loss is presented as an added physical constraint rather than a tautological fit or renamed known result, and the reported improvements are empirical comparisons against external baselines (PARIS, Ditto). The derivation chain therefore supplies independent content beyond its inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; full manuscript text was not supplied, so the ledger is necessarily incomplete and speculative.

free parameters (1)
  • energy residual loss weighting coefficients
    Abstract implies tunable weights to balance the energy term against other losses, but no values or fitting procedure are given.
axioms (1)
  • standard math Featherstone's articulated-body algorithm can be differentiated end-to-end for joint-parameter optimization
    Invoked by the differentiable screw-axis solver component.

pith-pipeline@v0.9.1-grok · 5804 in / 1350 out tokens · 23016 ms · 2026-06-26T21:35:06.680708+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

57 extracted references · 2 linked inside Pith

  1. [1]

    Coumans, Y

    E. Coumans, Y . Bai, PyBullet, a Python module for physics simulation for games, robotics and machine learning, http://pybullet.org(2016–2021)

    2016

  2. [2]

    H. Zhao, L. Yan, Z. Hou, J. Lin, Y . Zhao, Z. Ji, Y . Wang, Error analysis strategy for long-term correlated network systems: Generalized nonlinear stochastic processes and dual-layer filtering architecture, IEEE Internet of Things Journal (2025)

    2025

  3. [3]

    Z. Li, Y . Hu, Z. Chen, Q. Huang, G. Qiu, Z. Fu, M. Liu, Retrack: Evidence-driven dual-stream directional anchor calibration network for composed video retrieval, in: Proceedings of the AAAI Conference on Artificial Intelligence, 2026

    2026

  4. [4]

    Z. Yu, M. Y . I. Idris, P. Wang, R. Qureshi, Dinov3-powered multi-task foundation model for quantitative remote sensing estimation (student abstract), in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 40, 2026, pp. 41455–41456

    2026

  5. [5]

    S. Jia, N. Zhu, J. Zhong, J. Zhou, H. Zhang, J.-N. Hwang, L. Li, Ram: Recover any 3d human motion in-the-wild (2026).arXiv:2603.19929. URLhttps://arxiv.org/abs/2603.19929

    Pith/arXiv arXiv 2026

  6. [6]

    Z. Chen, Y . Hu, Z. Fu, Z. Li, J. Huang, Q. Huang, Y . Wei, Intent: Invariance and discrimination-aware noise mit- igation for robust composed image retrieval, in: Proceedings of the AAAI Conference on Artificial Intelligence, 2026

    2026

  7. [7]

    Z. Yu, H. Jiang, P. Wang, Z. Lin, Y . Xiang, Spatiotemporal alignment for remote sensing image recovery via terrain-aware diffusion, in: ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2026, pp. 11257–11261

    2026

  8. [8]

    L. Li, S. Jia, J. Wang, Z. Jiang, F. Zhou, J. Dai, T. Zhang, Z. Wu, J.-N. Hwang, Human Motion Instruction Tuning, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

    2025

  9. [9]

    Z. Li, Y . Hu, Z. Chen, S. Zhang, Q. Huang, Z. Fu, Y . Wei, Habit: Chrono-synergia robust progressive learning framework for composed image retrieval, in: Proceedings of the AAAI Conference on Artificial Intelligence, 2026. 14

    2026

  10. [10]

    Sarkar, M

    A. Sarkar, M. Y . I. Idris, Z. Yu, Reasoning in computer vision: Taxonomy, models, tasks, and methodologies, arXiv preprint arXiv:2508.10523 (2025)

    arXiv 2025

  11. [11]

    L. Li, S. Jia, J.-N. Hwang, Multiple human motion understanding, in: Proceedings of the AAAI Conference on Artificial Intelligence, V ol. 40, 2026, pp. 6297–6305

    2026

  12. [12]

    Z. Fu, Y . Hu, Q. Yang, S. Zhang, Z. Chen, Z. Li, Air-know: Arbiter-calibrated knowledge-internalizing robust network for composed image retrieval, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

    2026

  13. [13]

    R. Gu, S. Jia, Y . Ma, J. Zhong, J.-N. Hwang, L. Li, Mocount: Motion-based repetitive action counting, in: Proceedings of the 33rd ACM International Conference on Multimedia, 2025, pp. 9026–9034

    2025

  14. [14]

    L. Li, S. Jia, J. Wang, Z. An, J. Li, J.-N. Hwang, S. Belongie, Chatmotion: A multimodal multi-agent for human motion analysis, arXiv preprint arXiv:2502.18180 (2025)

    arXiv 2025

  15. [15]

    Z. Li, Y . Hu, Z. Chen, M. Zhang, Z. Fu, L. Nie, Conesep: Cone-based robust noise-unlearning compositional network for composed image retrieval, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

    2026

  16. [16]

    S. Jia, L. Li, Adaptive masking enhances visual grounding, arXiv preprint arXiv:2410.03161 (2024)

    arXiv 2024

  17. [17]

    B. Li, H. Dong, D. Zhang, Z. Zhao, J. Gao, X. Li, Exploring efficient open-vocabulary segmentation in the remote sensing, arXiv preprint arXiv:2509.12040 (2025)

    arXiv 2025

  18. [18]

    B. Li, D. Zhang, Z. Zhao, J. Gao, X. Li, Stitchfusion: Weaving any visual modalities to enhance multimodal semantic segmentation, in: Proceedings of the 33rd ACM International Conference on Multimedia, 2025, pp. 1308–1317

    2025

  19. [19]

    Z. Chen, A. Walsman, M. Memmel, K. Mo, A. Fang, K. Vemuri, A. Wu, D. Fox, A. Gupta, URDFormer: A pipeline for constructing articulated simulation environments from real-world images, in: Robotics: Science and Systems (RSS), 2024

    2024

  20. [20]

    L. Le, J. Xie, W. Liang, H.-J. Wang, Y . Yang, Y . J. Ma, K. Vedder, A. Krishna, D. Jayaraman, E. Eaton, Articulate-anything: Automatic modeling of articulated objects via a vision-language foundation model, in: International Conference on Learning Representations (ICLR), 2025

    2025

  21. [21]

    Z. Yu, M. Y . I. Idris, H. Wang, P. Wang, J. Chen, K. Wang, From physics to foundation models: A review of ai-driven quantitative remote sensing inversion, arXiv preprint arXiv:2507.09081 (2025). 15

    arXiv 2025

  22. [22]

    J. Liu, A. Mahdavi-Amiri, M. Savva, PARIS: Part-level reconstruction and motion analysis for articulated ob- jects, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 352– 363

    2023

  23. [23]

    B. Deng, J. P. Lewis, T. Jeruzalski, G. Pons-Moll, G. Hinton, M. Norouzi, A. Tagliasacchi, NASA: Neural articulated shape approximation, in: European Conference on Computer Vision (ECCV), 2020, pp. 612–628

    2020

  24. [24]

    Jiang, Y

    H. Jiang, Y . Mao, M. Savva, A. X. Chang, OPD: Single-view 3D openable part detection, in: European Confer- ence on Computer Vision (ECCV), 2022, pp. 410–426

    2022

  25. [25]

    B. Li, F. Wang, D. Zhang, Z. Zhao, J. Gao, X. Li, Maris: Marine open-vocabulary instance segmentation with geometric enhancement and semantic alignment, arXiv preprint arXiv:2510.15398 (2025)

    arXiv 2025

  26. [26]

    B. Li, T. Huo, D. Zhang, Z. Zhao, J. Gao, X. Li, Exploring the underwater world segmentation without extra training, arXiv preprint arXiv:2511.07923 (2025)

    arXiv 2025

  27. [27]

    B. Li, D. Zhang, Z. Zhao, J. Gao, X. Li, U3m: Unbiased multiscale modal fusion model for multimodal semantic segmentation, Pattern Recognition 168 (2025) 111801

    2025

  28. [28]

    Y . Chen, Z. Cao, H. Ren, C. Yang, W. Li, S. Wang, Y . Wang, L. Zhang, Y . Shao, Z. Zhao, et al., Roborouter: Training-free policy routing for robotic manipulation, arXiv preprint arXiv:2603.07892 (2026)

    Pith/arXiv arXiv 2026

  29. [29]

    Jiang, T

    G. Jiang, T. Zhang, D. Li, Z. Zhao, H. Li, M. Li, H. Wang, Stg-avatar: Animatable human avatars via spacetime gaussian, in: 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, 2025, pp. 20058–20065

    2025

  30. [30]

    S. Yan, P. Shi, Z. Zhao, K. Wang, K. Cao, J. Wu, J. Li, Turboreg: Turboclique for robust and efficient point cloud registration, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 26371–26381

    2025

  31. [31]

    Z. Zhao, H. Yang, B. Liao, Y . Zeng, S. Yan, Y . Gu, P. Liu, Y . Zhou, H. Li, J. Civera, Advances in global solvers for 3d vision, arXiv preprint arXiv:2602.14662 (2026)

    arXiv 2026

  32. [32]

    J. Chan, Z. Zhao, Y .-L. Liu, Adagar: Adaptive gabor representation for dynamic scene reconstruction, arXiv preprint arXiv:2601.00796 (2026)

    arXiv 2026

  33. [33]

    Z. Yu, J. Wang, H. Chen, M. Y . I. Idris, Qrs-trs: Style transfer-based image-to-image translation for carbon stock estimation in quantitative remote sensing, IEEE Access (2025)

    2025

  34. [34]

    Featherstone, Rigid Body Dynamics Algorithms, Springer, 2008

    R. Featherstone, Rigid Body Dynamics Algorithms, Springer, 2008. 16

    2008

  35. [35]

    Xiang, Y

    F. Xiang, Y . Qin, K. Mo, Y . Xia, H. Zhu, F. Liu, M. Liu, H. Jiang, Y . Yuan, H. Wang, L. Yi, A. X. Chang, L. J. Guibas, H. Su, SAPIEN: A simulated part-based interactive environment, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 11094–11104

    2020

  36. [36]

    Jiang, C.-C

    Z. Jiang, C.-C. Hsu, Y . Zhu, Ditto: Building digital twins of articulated objects from interaction, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 5616–5626

    2022

  37. [37]

    A. Jain, R. Lioutikov, C. Chuck, S. Niekum, ScrewNet: Category-independent articulation model estimation from depth images using screw theory, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 13670–13677

    2021

  38. [38]

    L. Liu, W. Xu, H. Fu, S. Qian, Q. Yu, Y . Han, C. Lu, AKB-48: A real-world articulated object knowledge base, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 14809–14818

    2022

  39. [39]

    Martín-Martín, C

    R. Martín-Martín, C. Eppner, O. Brock, The RBO dataset of articulated objects and interactions, The Interna- tional Journal of Robotics Research 38 (9) (2019) 1013–1019

    2019

  40. [40]

    X. Wang, B. Zhou, Y . Shi, X. Chen, Q. Zhao, K. Xu, Shape2Motion: Joint analysis of motion parts and attributes from 3D shapes, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 8876–8884

    2019

  41. [41]

    X. Li, H. Wang, L. Yi, L. J. Guibas, A. L. Abbott, S. Song, Category-level articulated object pose estimation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 3706–3715

    2020

  42. [42]

    L. Liu, H. Xue, W. Xu, H. Fu, C. Lu, Toward real-world category-level articulation pose estimation, IEEE Transactions on Image Processing 31 (2022) 1072–1083

    2022

  43. [43]

    K. Mo, L. J. Guibas, M. Mukadam, A. Gupta, S. Tulsiani, Where2Act: From pixels to actions for articulated 3D objects, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 6813–6823

    2021

  44. [44]

    Abbatematteo, S

    B. Abbatematteo, S. Tellex, G. Konidaris, Learning to generalize kinematic models to novel objects, in: Confer- ence on Robot Learning (CoRL), 2019, pp. 1289–1299

    2019

  45. [45]

    Mildenhall, P

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, R. Ng, NeRF: Representing scenes as neural radiance fields for view synthesis, in: European Conference on Computer Vision (ECCV), 2020, pp. 405–421. 17

    2020

  46. [46]

    Kerbl, G

    B. Kerbl, G. Kopanas, T. Leimkühler, G. Drettakis, 3D Gaussian Splatting for real-time radiance field rendering, ACM Transactions on Graphics 42 (4) (2023)

    2023

  47. [47]

    Noguchi, X

    A. Noguchi, X. Sun, S. Lin, T. Harada, Neural articulated radiance field, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 5762–5772

    2021

  48. [48]

    Heiden, Z

    E. Heiden, Z. Liu, V . Vineet, E. Coumans, G. S. Sukhatme, Inferring articulated rigid body dynamics from RGBD video, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022

    2022

  49. [49]

    Khalil, Dynamic modeling of robots using recursive Newton-Euler techniques, in: International Conference on Informatics in Control, Automation and Robotics (ICINCO), 2010

    W. Khalil, Dynamic modeling of robots using recursive Newton-Euler techniques, in: International Conference on Informatics in Control, Automation and Robotics (ICINCO), 2010

    2010

  50. [50]

    C. D. Freeman, E. Frey, A. Raichuk, S. Girgin, I. Mordatch, O. Bachem, Brax – a differentiable physics en- gine for large scale rigid body simulation, in: Advances in Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks Track, 2021

    2021

  51. [51]

    Werling, D

    K. Werling, D. Omens, J. Lee, I. Exarchos, C. K. Liu, Fast and feature-complete differentiable physics for articulated rigid bodies with contact, in: Robotics: Science and Systems (RSS), 2021

    2021

  52. [52]

    Y . Hu, L. Anderson, T.-M. Li, Q. Sun, N. Carr, J. Ragan-Kelley, F. Durand, DiffTaichi: Differentiable program- ming for physical simulation, in: International Conference on Learning Representations (ICLR), 2020

    2020

  53. [53]

    T. A. Howell, S. Le Cleac’h, J. Brüdigam, J. Z. Kolter, M. Schwager, Z. Manchester, Dojo: A differentiable physics engine for robotics, arXiv preprint arXiv:2203.00806 (2022)

    arXiv 2022

  54. [54]

    Strecke, J

    M. Strecke, J. Stückler, DiffSDFSim: Differentiable rigid-body dynamics with implicit shapes, in: International Conference on 3D Vision (3DV), 2021

    2021

  55. [55]

    K. M. Jatavallabhula, M. Macklin, F. Golemo, V . V oleti, L. Petrini, M. Weiss, B. Considine, J. Parent-Lévesque, K. Xie, K. Erleben, L. Paull, F. Shkurti, D. Nowrouzezahrai, S. Fidler, gradsim: Differentiable simulation for system identification and visuomotor control, in: International Conference on Learning Representations (ICLR), 2021

    2021

  56. [56]

    T. Mu, Z. Ling, F. Xiang, D. Yang, X. Li, S. Tao, Z. Huang, Z. Jia, H. Su, ManiSkill: Generalizable manipula- tion skill benchmark with large-scale demonstrations, in: Advances in Neural Information Processing Systems (NeurIPS) Datasets and Benchmarks Track, 2021

    2021

  57. [57]

    H. Geng, H. Xu, C. Zhao, C. Xu, L. Yi, S. Huang, H. Wang, GAPartNet: Cross-category domain-generalizable object perception and manipulation via generalizable and actionable parts, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 7081–7091. 18

    2023