pith. sign in

arxiv: 2605.15949 · v1 · pith:ZVU2AEDInew · submitted 2026-05-15 · 💻 cs.RO

A Reproducible and Physically Feasible Dynamic Parameter Identification Framework for a Low-Cost Robot Arm

Junji Oaki , Koki Yamane , Koki Inami , Sho Sakaino This is my paper

Pith reviewed 2026-05-20 17:48 UTC · model grok-4.3

classification 💻 cs.RO
keywords dynamic parameter identificationrobot armphysical feasibilitylow-cost roboticssemidefinite programmingordinary least squaresinertia matrix
0
0 comments X

The pith

A multi-stage pipeline with symmetry reduction delivers physically feasible dynamic models for low-cost robot arms.

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

The paper sets out to show that a practical identification process can produce accurate and physically valid dynamic parameters for the CRANE-X7 low-cost arm. It starts by cutting the rigid-body model from 65 to 39 parameters through removal of products of inertia based on approximate link symmetry, then applies ordinary least squares fitting, followed by semidefinite programming to restore feasibility and closed-loop input error refinement for further improvement. Multiple hand-designed trajectories are run, a central model is chosen via PCA, and final checks confirm positive definiteness of the inertia matrix across poses while keeping high accuracy on unseen validation motions.

Core claim

The central claim is that preprocessing, inverse-dynamics-regressor OLS, conditional SDP projection for feasibility recovery, and CLIE refinement, applied after symmetry-based reduction to 39 base parameters, produce a statistically central, physically acceptable model that maintains strong predictive performance on held-out motions for the CRANE-X7 arm.

What carries the argument

The multi-stage pipeline of OLS fitting, conditional semidefinite-programming projection, and closed-loop input error refinement, used after reducing the model from 65 to 39 base parameters by removing products of inertia according to approximate link symmetry.

Load-bearing premise

Removing products of inertia according to approximate link symmetry reduces the model from 65 to 39 base parameters without materially harming predictive accuracy or identifiability.

What would settle it

A direct comparison showing that the 39-parameter reduced model yields substantially lower torque prediction accuracy on the same validation motions than the full 65-parameter model would disprove the claim.

Figures

Figures reproduced from arXiv: 2605.15949 by Junji Oaki, Koki Inami, Koki Yamane, Sho Sakaino.

Figure 1
Figure 1. Figure 1: CRANE-X7 with the replaced cross-structure hand [5] and its kine [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Joint-level modeling and identification for FF+PD controller design. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Closed-loop input error (CLIE) refinement. A candidate parameter [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: PA, PB, PC, AP, and AG form 40 identification candidates; V01–V03 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Held-out validation trajectory V03. Representative commands of J2, [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: PCA-based visualization of parameter clouds from OLS, SDP, and CLIE. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Validation torque prediction on V03 for J2, J4, and J6. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
read the original abstract

This paper presents a reproducible and physically feasible dynamic parameter identification framework for CRANE-X7, a low-cost robot arm driven by modular smart actuators. To improve practical identifiability, products of inertia are removed according to approximate link symmetry, reducing the rigid-body model from 65 to 39 base parameters. Identification motions are hand-designed from structured single-joint and adjacent-joint primitives under practical joint-range limits. The proposed pipeline combines preprocessing, inverse-dynamics-regressor-based ordinary least squares (OLS), conditional semidefinite-programming (SDP) projection for feasibility recovery, and closed-loop input error (CLIE) refinement. Candidate solutions from 40 structured trajectories are analyzed in a common PCA space to select a statistically central representative model. Because statistical centrality alone does not ensure physical acceptability, the selected model is finally screened by an all-pose positive-definiteness audit of the inertia matrix and, when necessary, corrected by a localized post-CLIE SDP rescue step. Experiments show that the parameter cloud becomes progressively more concentrated from OLS to SDP and CLIE, while the final accepted model preserves high predictive accuracy on held-out validation motions. These results demonstrate a practical route to statistically coherent and physically feasible dynamic models for low-cost robot platforms.

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 presents a reproducible pipeline for identifying physically feasible dynamic parameters of the CRANE-X7 low-cost robot arm. It reduces the rigid-body model from 65 to 39 base parameters by setting products of inertia to zero under an approximate link-symmetry assumption, then applies OLS fitting to torque/position data from 40 hand-designed single- and adjacent-joint trajectories, followed by SDP projection, CLIE refinement, PCA-based selection of a statistically central model, and a final all-pose positive-definiteness audit with optional post-CLIE SDP rescue. Experiments are reported to show progressive concentration of the parameter cloud and retention of high predictive accuracy on held-out validation motions.

Significance. If the symmetry reduction and validation results hold, the work supplies a practical, end-to-end route to statistically coherent and physically feasible inertial models for low-cost modular arms, where standard full-parameter identification is often ill-conditioned. The combination of structured excitation primitives, SDP/CLIE feasibility recovery, and explicit positive-definiteness screening addresses a common practical gap; the emphasis on reproducibility and the use of an external physical audit rather than purely statistical selection are notable strengths.

major comments (2)
  1. [model reduction and validation sections] The reduction from 65 to 39 base parameters by zeroing products of inertia under approximate link symmetry (described in the abstract and the model-reduction paragraph) is load-bearing for the claim of improved practical identifiability. For a modular low-cost platform, cabling, actuator mounting offsets, and manufacturing tolerances can violate the symmetry assumption; the manuscript should quantify the torque prediction error introduced by this reduction on the validation set or demonstrate that the chosen single-joint and adjacent-joint primitives sufficiently excite the omitted directions.
  2. [candidate selection and audit paragraph] The final model selection combines PCA centrality with an all-pose positive-definiteness audit and occasional post-CLIE SDP rescue. It is unclear how frequently the rescue step is invoked across the 40 trajectories and whether the corrected parameters remain within the statistically central cluster; a table reporting the fraction of candidates requiring rescue and the resulting change in validation RMSE would strengthen the claim that the pipeline reliably yields feasible yet accurate models.
minor comments (2)
  1. [methods] Notation for the base inertial parameters and the regressor matrix should be introduced once with explicit dimensions (e.g., number of base parameters after symmetry reduction) to avoid ambiguity when comparing OLS, SDP, and CLIE stages.
  2. [results figures] Figure captions for the parameter-cloud PCA plots should state the percentage of variance explained by the first two principal components and indicate which trajectories correspond to the selected central model.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive review. We address each major comment below and will revise the manuscript accordingly to strengthen the presentation of the model reduction and the model selection procedure.

read point-by-point responses

  1. Referee: [model reduction and validation sections] The reduction from 65 to 39 base parameters by zeroing products of inertia under approximate link symmetry (described in the abstract and the model-reduction paragraph) is load-bearing for the claim of improved practical identifiability. For a modular low-cost platform, cabling, actuator mounting offsets, and manufacturing tolerances can violate the symmetry assumption; the manuscript should quantify the torque prediction error introduced by this reduction on the validation set or demonstrate that the chosen single-joint and adjacent-joint primitives sufficiently excite the omitted directions.

    Authors: We agree that the symmetry reduction is a central modeling choice whose practical impact merits explicit quantification. The current validation results show that the reduced model retains high predictive accuracy, but a direct assessment of the error introduced by zeroing the products of inertia would be valuable. In the revised manuscript we will add a quantitative comparison on the held-out validation motions: we will compute the increase in torque RMSE when the zeroed inertia products are forced to zero versus when they are retained (where the full regressor remains well-conditioned). We will also report the relative excitation of the omitted regressor columns across the 40 trajectories by their column norms and condition-number contribution, thereby showing that the single- and adjacent-joint primitives primarily excite the retained 39 parameters. revision: yes

  2. Referee: [candidate selection and audit paragraph] The final model selection combines PCA centrality with an all-pose positive-definiteness audit and occasional post-CLIE SDP rescue. It is unclear how frequently the rescue step is invoked across the 40 trajectories and whether the corrected parameters remain within the statistically central cluster; a table reporting the fraction of candidates requiring rescue and the resulting change in validation RMSE would strengthen the claim that the pipeline reliably yields feasible yet accurate models.

    Authors: We appreciate the request for quantitative detail on the rescue step. Although the manuscript notes that rescue is applied 'when necessary,' it does not report frequency or effect size. In the revised version we will insert a table that lists, for the 40 trajectories: (i) the fraction requiring post-CLIE SDP rescue, (ii) the Euclidean distance of each rescued parameter vector from the PCA centroid before and after correction, and (iii) the change in validation RMSE attributable to the rescue. This will confirm that rescued models remain statistically central while satisfying the positive-definiteness audit. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the identification framework

full rationale

The paper presents an empirical identification pipeline that fits rigid-body parameters to measured torque and position data via OLS, applies SDP projection to enforce physical feasibility, and refines via CLIE. Parameter reduction from 65 to 39 is performed by an explicit modeling choice that sets products of inertia to zero under an approximate symmetry assumption; this is stated as a preprocessing step rather than derived from the equations. Model selection combines statistical centrality in PCA space with an independent all-pose positive-definiteness audit and held-out validation motions. No load-bearing step reduces by construction to its own inputs, no self-citations are invoked as uniqueness theorems, and the central claims rest on external experimental benchmarks rather than tautological re-labeling of fitted quantities.

Axiom & Free-Parameter Ledger

1 free parameters · 3 axioms · 0 invented entities

The framework rests on standard linear regression and convex optimization assumptions plus the domain-specific claim that link symmetry permits dropping products of inertia. No new physical entities are postulated.

free parameters (1)
  • 39 base inertial parameters
    The reduced set of inertial, friction, and actuator parameters obtained after symmetry-based removal of products of inertia; these are the quantities fitted to data.
axioms (3)
  • standard math The rigid-body dynamics can be expressed as a linear regressor in the base parameters.
    Invoked when the inverse-dynamics regressor is formed for OLS.
  • domain assumption Approximate link symmetry justifies setting products of inertia to zero.
    Stated as the step that reduces the model from 65 to 39 parameters.
  • domain assumption The SDP projection and subsequent positive-definiteness audit recover physically feasible inertia matrices.
    Used to guarantee that the final model satisfies inertia-matrix positive-definiteness in all poses.

pith-pipeline@v0.9.0 · 5763 in / 1525 out tokens · 53481 ms · 2026-05-20T17:48:50.030844+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

36 extracted references · 36 canonical work pages

  1. [1]

    Learning fine-grained bimanual manipulation with low-cost hardware,

    T. Z. Zhao, V . Kumar, S. Levine, and C. Finn, “Learning fine-grained bimanual manipulation with low-cost hardware,” inRobotics: Science and Systems, Daegu, Korea, Jul. 2023

    work page 2023

  2. [2]

    Aloha 2: An enhanced low-cost hardware for bimanual teleoperation,

    ALOHA 2 Team, J. Aldaco, T. Armstrong, R. Baruch, J. Bingham, S. Chan, K. Draper, D. Dwibedi, C. Finn, P. Florence, S. Goodrich, W. Gramlich, T. Haarnoja, A. Herzog, J. Hoech, T. Nguyen, I. Pierson, T. Wahrburg, L. Wang, S. Young, Y . Lu, and S. Levine, “Aloha 2: An enhanced low-cost hardware for bimanual teleoperation,”arXiv preprint, arXiv:2405.02292 [c...

    work page arXiv 2024

  3. [3]

    CRANE-X7,

    RT Corporation, “CRANE-X7,” https://rt-net.jp/products/crane-x7/

    work page

  4. [4]

    DYNAMIXEL X-Series e-Manual,

    ROBOTIS, “DYNAMIXEL X-Series e-Manual,” Online, 2026, avail- able: ROBOTIS e-Manual

    work page 2026

  5. [5]

    Soft and rigid object grasping with cross-structure hand using bilateral control-based imitation learning,

    K. Yamane, Y . Saigusa, S. Sakaino, and T. Tsuji, “Soft and rigid object grasping with cross-structure hand using bilateral control-based imitation learning,”IEEE Robotics and Automation Letters, vol. 9, no. 2, pp. 1198–1205, Feb. 2024

    work page 2024

  6. [6]

    Motion retouch: Motion modifi- cation using four-channel bilateral control,

    K. Inami, S. Sakaino, and T. Tsuji, “Motion retouch: Motion modifi- cation using four-channel bilateral control,” inProceedings of the 2025 IEEE International Conference on Mechatronics, Wollongong, Australia, Feb. 2025, pp. 1–6

    work page 2025

  7. [7]

    Design and experimental validation of sensorless 4-channel bilateral teleoperation for low-cost manipulators,

    K. Yamane, Y . Li, M. Konosu, K. Inami, J. Oaki, S. Sakaino, and T. Tsuji, “Design and experimental validation of sensorless 4-channel bilateral teleoperation for low-cost manipulators,”arXiv preprint, arXiv:2507.06174 [cs.RO], Jul. 2025

    work page arXiv 2025

  8. [8]

    A new identification method for serial manipulator arms,

    H. Mayeda, K. Osuka, and A. Kangawa, “A new identification method for serial manipulator arms,” inIFAC Proceedings Volumes, vol. 17, no. 2, Jul. 1984, pp. 2429–2434

    work page 1984

  9. [9]

    Base parameters of manipulator dynamic models,

    H. Mayeda, K. Yoshida, and K. Osuka, “Base parameters of manipulator dynamic models,”IEEE Transactions on Robotics and Automation, vol. 6, no. 3, pp. 312–321, Jun. 1990

    work page 1990

  10. [10]

    Direct calculation of minimum set of inertial parameters of serial robots,

    M. Gautier and W. Khalil, “Direct calculation of minimum set of inertial parameters of serial robots,”IEEE Transactions on Robotics and Automation, vol. 6, no. 3, pp. 368–373, Jun. 1990

    work page 1990

  11. [11]

    Verification of the positive definiteness of the inertial matrix of manipulators using base inertial parameters,

    K. Yoshida and W. Khalil, “Verification of the positive definiteness of the inertial matrix of manipulators using base inertial parameters,”The International Journal of Robotics Research, vol. 19, no. 5, pp. 498–510, May 2000

    work page 2000

  12. [12]

    Physical feasibility of robot base inertial parameter identification: A linear matrix inequality approach,

    C. D. Sousa and R. Cortes ˜ao, “Physical feasibility of robot base inertial parameter identification: A linear matrix inequality approach,”The International Journal of Robotics Research, vol. 33, no. 6, pp. 931– 944, Feb. 2014

    work page 2014

  13. [13]

    Linear matrix inequalities for physically consistent inertial parameter identification: A statistical perspective on the mass distribution,

    P. M. Wensing, S. Kim, and J.-J. E. Slotine, “Linear matrix inequalities for physically consistent inertial parameter identification: A statistical perspective on the mass distribution,”IEEE Robotics and Automation Letters, vol. 3, no. 1, pp. 60–67, Jan. 2018

    work page 2018

  14. [14]

    A new closed-loop output error method for parameter identification of robot dynamics,

    M. Gautier, A. Janot, and P.-O. Vandanjon, “A new closed-loop output error method for parameter identification of robot dynamics,”IEEE Transactions on Control Systems Technology, vol. 21, no. 2, pp. 428– 444, Mar. 2013

    work page 2013

  15. [15]

    Stable robot manipulator pa- rameter identification: A closed-loop input error approach,

    A. Perrusqu ´ıa, R. Garrido, and W. Yu, “Stable robot manipulator pa- rameter identification: A closed-loop input error approach,”Automatica, vol. 141, 110294, Jul. 2022

    work page 2022

  16. [16]

    Inertial parameter identification in robotics: A survey,

    Q. Leboutet, J. Roux, A. Janot, J. R. Guadarrama-Olvera, and G. Cheng, “Inertial parameter identification in robotics: A survey,”Applied Sci- ences, vol. 11, no. 9, 4303, May 2021

    work page 2021

  17. [17]

    Opensymoro: An open-source software package for symbolic modelling of robots,

    W. Khalil, A. Vijayalingam, B. Khomutenko, I. Mukhanov, P. Lemoine, and G. Ecorchard, “Opensymoro: An open-source software package for symbolic modelling of robots,” inProceedings of the 2014 IEEE/ASME International Conference on Advanced Intelligent Mechatronics, Besan- con, France, Jul. 2014, pp. 1206–1211

    work page 2014

  18. [18]

    Khalil and E

    W. Khalil and E. Dombre,Modeling, Identification and Control of Robots. London, UK: Hermes Penton Science, 2002

    work page 2002

  19. [19]

    Estimation of inertial parameters of manipulator loads and links,

    C. G. Atkeson, C. H. An, and J. M. Hollerbach, “Estimation of inertial parameters of manipulator loads and links,”The International Journal of Robotics Research, vol. 5, no. 3, pp. 101–119, Sep. 1986

    work page 1986

  20. [20]

    Ex- perimental robot identification using optimised periodic trajectories,

    J. Swevers, C. Ganseman, J. De Schutter, and H. Van Brussel, “Ex- perimental robot identification using optimised periodic trajectories,” Mechanical Systems and Signal Processing, vol. 10, no. 5, pp. 561– 577, Sep. 1996

    work page 1996

  21. [21]

    Optimal robot excitation and identification,

    J. Swevers, C. Ganseman, D. B. Tukel, J. De Schutter, and H. Van Brus- sel, “Optimal robot excitation and identification,”IEEE Transactions on Robotics and Automation, vol. 13, no. 5, pp. 730–740, Oct. 1997

    work page 1997

  22. [22]

    New criteria of exciting trajectories for robot identification,

    C. Presse and M. Gautier, “New criteria of exciting trajectories for robot identification,” inProceedings of the 1993 IEEE International Conference on Robotics and Automation, Atlanta, Georgia, May 1993, pp. 907–912

    work page 1993

  23. [23]

    Fourier-based optimal excitation trajectories for the dynamic identification of robots,

    K.-J. Park, “Fourier-based optimal excitation trajectories for the dynamic identification of robots,”Robotica, vol. 24, no. 5, pp. 625–633, Mar. 2006

    work page 2006

  24. [24]

    An overview of dynamic parameter iden- tification of robots,

    J. Wu, J. Wang, and Z. You, “An overview of dynamic parameter iden- tification of robots,”Robotics and Computer-Integrated Manufacturing, vol. 26, no. 5, pp. 414–419, Oct. 2010

    work page 2010

  25. [25]

    Dynamic parameter identification in industrial robots considering physical feasibility,

    V . Mata, F. Benimeli, N. Farhat, and A. Valera, “Dynamic parameter identification in industrial robots considering physical feasibility,”Ad- vanced Robotics, vol. 19, no. 1, pp. 101–119, Apr. 2005

    work page 2005

  26. [26]

    Parameter identification for industrial robots with a fast and robust trajectory design approach,

    J. Jin and N. Gans, “Parameter identification for industrial robots with a fast and robust trajectory design approach,”Robotics and Computer- Integrated Manufacturing, vol. 31, pp. 21–29, Feb. 2015

    work page 2015

  27. [27]

    Inertia tensor properties in robot dy- namics identification: A linear matrix inequality approach,

    C. D. Sousa and R. Cortes ˜ao, “Inertia tensor properties in robot dy- namics identification: A linear matrix inequality approach,”IEEE/ASME Transactions on Mechatronics, vol. 24, no. 1, pp. 406–411, Feb. 2019

    work page 2019

  28. [28]

    A lie-theory-based dynamic parameter identification methodology for serial manipulators,

    Z. Fu, J. Pan, E. Spyrakos-Papastavridis, Y .-H. Lin, X. Zhou, X. Chen, and J. S. Dai, “A lie-theory-based dynamic parameter identification methodology for serial manipulators,”IEEE/ASME Transactions on Mechatronics, vol. 26, no. 5, pp. 2688–2699, Oct. 2021

    work page 2021

  29. [29]

    An optimal information method for mobile manipulator dynamic parameter identification,

    V . A. Sujan and S. Dubowsky, “An optimal information method for mobile manipulator dynamic parameter identification,”IEEE/ASME Transactions on Mechatronics, vol. 8, no. 2, pp. 215–225, Jun. 2003

    work page 2003

  30. [30]

    Model-based control of a 3-dof parallel robot based on identified relevant parameters,

    M. D ´ıaz-Rodr´ıguez, ´A. Valera, V . Mata, and M. Vall ´es, “Model-based control of a 3-dof parallel robot based on identified relevant parameters,” IEEE/ASME Transactions on Mechatronics, vol. 18, no. 6, pp. 1737– 1744, Dec. 2013

    work page 2013

  31. [31]

    Dynamic model identifi- cation for industrial robots,

    J. Swevers, W. Verdonck, and J. De Schutter, “Dynamic model identifi- cation for industrial robots,”IEEE Control Systems Magazine, vol. 27, no. 5, pp. 58–71, Oct. 2007

    work page 2007

  32. [32]

    Sequential semidefinite optimization for physically and statistically consistent robot identification,

    A. Janot and P. M. Wensing, “Sequential semidefinite optimization for physically and statistically consistent robot identification,”Control Engineering Practice, vol. 107, 104699, Feb. 2021

    work page 2021

  33. [33]

    Dynamic identification of the franka emika panda robot with retrieval of feasible parameters using penalty-based optimization,

    C. Gaz, M. Cognetti, A. Oliva, P. R. Giordano, and A. D. Luca, “Dynamic identification of the franka emika panda robot with retrieval of feasible parameters using penalty-based optimization,”IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 4147–4154, Oct. 2019

    work page 2019

  34. [34]

    Dynamic identification of the kuka lbr iiwa robot with retrieval of physical parameters using global optimization,

    T. Xu, J. Fan, Y . Chen, X. Ng, J. Marcelo H. Ang, Q. Fang, Y . Zhu, and J. Zhao, “Dynamic identification of the kuka lbr iiwa robot with retrieval of physical parameters using global optimization,”IEEE Access, vol. 8, pp. 108 018 – 108 031, Jun. 2020

    work page 2020

  35. [35]

    Dynamic parameter identifi- cation of serial robots using a hybrid approach,

    Y . Huang, J. Ke, X. Zhang, and J. Ota, “Dynamic parameter identifi- cation of serial robots using a hybrid approach,”IEEE Transactions on Robotics, vol. 39, no. 2, pp. 1607–1621, Apr. 2023

    work page 2023

  36. [36]

    A convex optimization-based dynamic model identification package for the da vinci research kit,

    Y . Wang, R. Gondokaryono, A. Munawar, and G. S. Fischer, “A convex optimization-based dynamic model identification package for the da vinci research kit,”IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 3657–3664, Oct. 2019. 10 APPENDIX A. SUPPLEMENTARY TABLES FOR REPRODUCIBILITY This appendix gives the numerical reproducibility record for Sec. V...

    work page 2019