Pith. sign in

REVIEW 4 major objections 4 minor 21 references

Personalised 3D Human Digital Twin with Soft-Body Feet for Walking Simulation

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read The paper claims that adding personalised soft-body feet to a skeletal human model lets a walking controller trained only on kinematic motion capture produce ground reaction forces comparable to measured data.

desk verdict Useful MuJoCo integration for personalised soft-body feet, but the main GRF result is partly an artifact of tuning and offsets applied to the same data used for evaluation. read the letter →

arxiv 2411.14701 v1 pith:CX4QHSTD submitted 2024-11-22 cs.RO

classification cs.RO
keywords 3Dhumandigitaltwinsoft-bodysimulationMuJoCoflexgroundreactionforcewalkingcontrolpolicymotioncapturejointanglesSMPL
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that a personalised 3D human digital twin, built from motion capture data, can produce dynamically realistic walking in simulation when soft-body feet are added to a skeletal model. The walking control policy is trained only on kinematic motion capture, yet the soft feet generate vertical ground reaction forces comparable to force-plate measurements, with the best model reaching an average experimental-match score of 0.628 versus 0.558 for the bare skeleton. The authors argue this shows that simplified deformable feet can carry the contact dynamics needed for faithful human-robot interaction simulations without requiring force data during training. If true, this opens a route to personalised digital twins for studying assistive robots and rehabilitation scenarios using only motion capture of individual subjects.

What carries the argument

The central object is the MuJoCo flex object: a deformable mesh simulated with hyperelastic material properties, here a 15-vertex simplified foot generated from an SMPL body shape and attached to the skeletal model by hard pins, chosen because welds and connects are soft contacts that caused oscillation. The mechanism is that the flex foot deforms under ground contact and generates the ground reaction force; tuning its stiffness (2000 to 12000) and damping (100) plus shifting foot position lets the deformation produce a smooth, M-shaped force curve without the skeleton itself touching the floor.

What would settle it

Run model E with its fixed stiffness, damping, and foot position on a second subject's motion capture without re-tuning any parameters; if the vertical ground-reaction-force experimental match falls below the bare-skeleton baseline, the claim that personalised soft feet generate realistic forces from kinematic-only training would be refuted. A simpler check is to re-evaluate model E on the same subject without applying the reported phase offset and record how much the experimental match drops.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a uniform, 15-vertex hyperelastic foot mesh pinned to a skeleton is sufficient for a kinematic-only trained walking policy to reproduce the measured double-bump vertical ground reaction force profile. By stiffening the foot material from model A through model E and adjusting foot position, the authors eliminate skeleton-floor contact spikes and bring joint angles closer to the reference, with model E outperforming the bare skeleton on the experimental-match metric. The paper attributes residual phase differences between simulated and measured forces to shape simplification of the foot, and applies offsets when comparing curves.

Load-bearing premise

The load-bearing premise is that a 15-vertex, uniformly elastic foot pinned to the skeleton behaves like a real human foot in contact, and that the stiffness increases and phase offsets used during evaluation are legitimate corrections rather than tuning to the measured force data.

Editorial extensions

If this is right

  • If the claim is correct, personalised digital twins can be generated from motion capture alone and used to predict ground reaction forces in scenarios where force plates are not available.
  • The same soft-foot approach could be transferred to other skeletal or musculoskeletal models in the same physics engine, since only the foot contact needs to be replaced.
  • Stiffness tuning of the soft foot becomes a practical lever for matching contact dynamics: too soft feet cause intermittent skeleton-floor contact and force spikes, while too stiff feet approximate the bare skeleton.
  • The reported improvement in experimental match from 0.558 to 0.628 suggests that soft-body contact can add dynamic realism on top of kinematic fidelity without changing the control objective.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves open whether the phase offsets applied when comparing simulated to measured ground reaction forces are legitimate gait-phase alignment or a post-hoc correction; removing the offsets and reporting the drop in experimental match would clarify how much of the result comes from the model itself.
  • The stiffness and foot position were tuned on the same subject's force data used for evaluation, so a held-out subject test would show whether the personalisation procedure generalises or merely fits.
  • Because only the left foot had enough complete recordings to form a reliable standard deviation, the force comparison is one-footed; extending the capture protocol to the right foot would verify the claim for both limbs.
  • The 15-vertex simplification is likely the main source of the phase offset and the initial gradual rise in force, so increasing mesh resolution while keeping the material model fixed could test whether the offset shrinks.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes a personalised 3D human digital twin in MuJoCo that augments an existing skeletal (SK) model with soft-body feet represented as simplified flex objects. The feet are generated from motion-captured SMPL body shapes, reduced to roughly 15 vertices, attached to the skeleton with hard pins, and assigned hyperelastic material properties. A walking control policy is trained for several variants: the original skeleton (SK), two softer-foot models (A and B), and a final stiffened/repositioned model (E). The models are evaluated against recorded ground reaction forces (GRF) using an experimental-match (EM) metric and against joint angles using a linear-fit method (LFM). The paper reports that model E achieves the best GRF match (average EM 0.628 vs 0.558 for SK) and joint-angle results close to the SK model and reference motion.

Significance. If the evaluation were unbiased, the paper would make a useful engineering contribution: it demonstrates a low-cost way to add personalised soft-body contact to an existing skeletal model while training the control policy only on kinematic motion capture. The pipeline is concrete and reproducible in principle, and the use of real subject data, multiple stiffness variants, EM, and LFM metrics are strengths. However, the central empirical claim is not currently supported because the material properties, foot position, and phase offsets are tuned against the same recordings used for evaluation, and no held-out validation is performed. The kinematic results also show that model E is generally worse than the bare skeleton, so the soft-foot benefit rests almost entirely on the confounded GRF comparison. The contribution is therefore promising but requires a substantially strengthened evaluation before the reported claims can be accepted.

major comments (4)
  1. [§3.2 and Table 4] The central GRF comparison is confounded because the EM values for models A, B, and E are computed only after hand-applied phase offsets were used to 'better match the GRF profiles,' and the offset values are not reported. Because these offsets are chosen on the four gait cycles from the same recordings on which EM is then evaluated, the EM values in Table 4 are fit statistics rather than unbiased predictions; consequently, the improvement of model E over model SK (0.628 vs 0.558) does not by itself support the abstract's claim that the soft-body feet generate GRF comparable to real measured data. Please report the offsets, justify them independently of the evaluation data (e.g., from a calibration gait cycle), and recompute EM without data-dependent alignment.
  2. [§2.3 and §3.1] The material stiffness and foot position were fine-tuned using the GRF and joint-angle results of the same recordings used for the evaluation, as stated in Section 3.1: 'the material properties of the flex and the position of the flex foot were fine-tuned to improve the GRF and joint angle results.' This creates a direct circularity for the main empirical claim: the reported match of model E may largely reflect fitting to the test data. Please describe the tuning procedure (search grid, number of configurations tried, stopping criterion), and validate the final model on held-out gait cycles or subjects not used during tuning.
  3. [§3.3 and Table 5] The joint-angle results do not support the claim that model E 'closely follow[s] joint angle results of the bare skeletal model.' In Table 5, the mean R² of model SK is higher than that of model E for five of the six joints (e.g., L Ankle 0.950 vs 0.838; R Knee 0.928 vs 0.918), and model E only exceeds SK on R Ankle (0.945 vs 0.893). The paper should either soften the kinematic claim or provide an analysis showing that the differences are within measurement noise and not systematically worse.
  4. [§3.2] The GRF evaluation is limited to one foot and a small number of recordings: six complete left-foot recordings, with the right foot excluded because only three complete recordings were available. With n=6, the mean and standard deviation of the reference GRF are noisy, and no statistical test is provided for the EM differences across models; the per-cycle EM ranges for SK (0.54–0.58) and E (0.60–0.67) do not overlap, but it is unclear whether this is significant given the small sample. At a minimum, report per-recording values, the number of subjects, and a confidence interval or test for the EM comparison.
minor comments (4)
  1. [Table 1] The connective tissue Young's modulus range is printed as '1.5 × 10^6 – 2.25 × 10^5 kPa', which is in descending order and uses units that appear inconsistent with the muscle and fat rows; please verify the values and units.
  2. [§3.2] The sentence 'The higher EM values of model A over model SK confirm this' contradicts Table 4, where model A has average EM 0.463 ± 0.013 and model SK 0.558 ± 0.022; please correct the text or the table.
  3. [§3.1] The statement that 'the model took on average 11.7 seconds to train and around 4200 iterations to converge' alongside 'the base skeletal model required 5.3 seconds per iteration' suggests a missing 'per iteration' in the first quantity; please clarify.
  4. [General] Please add a data/code availability statement or clarify whether the models, motion-capture recordings, and evaluation scripts are available for replication.

Circularity Check

2 steps flagged · score 6.0 of 10

Reported GRF match is a tuned fit: stiffness, foot position, and phase offsets are adjusted on the evaluation recordings before computing the EM metric that supports the abstract's claim.

  1. fitted input called prediction [Section 3.1 (and Section 2.3); Table 3 and Table 4]
    "From the initial results of the training, the material properties of the flex and the position of the flex foot were fine-tuned to improve the GRF and joint angle results. This was done by progressively increasing the stiffness of the flex material and evaluating the GRF results based on the approaches detailed below. Material properties that produced better GRF results were deemed to be more representative properties."

    The stiffness and foot position of the flex model are selected by comparing simulated GRF with the force-plate recordings, and then those same recordings are used to compute the EM metric in Table 4 that supports the claim 'comparable to real measured data.' Model E's higher EM (0.628 vs SK's 0.558) is therefore a measure of how well the model was tuned to the evaluation data, not an independent prediction of the measured GRF. The SK baseline received no such tuning.

  2. fitted input called prediction [Section 3.2, Figure 3 caption and EM calculation]
    "Note that for models A, B, and E, offsets were applied to better match the GRF profiles."

    The EM metric is defined as the percentage of the gait cycle in which the simulation GRF lies within the standard deviation of recorded data (Section 3.2). For models A, B, and E, a phase offset is applied 'to better match the GRF profiles' before these EM values are computed. Because the offset is chosen on the same recorded curves being evaluated and is not reported, the EM values in Table 4 are post-alignment fit statistics. The SK model was evaluated without an offset, making the comparison inconsistent; the reported advantage of the soft-foot model is confounded.

full rationale

The paper's derivation of the soft-foot model is self-contained: SMPL shape, MuJoCo flex, pin attachment, and RL policy are all described, and none of the results are obtained by definitional identity or by renaming. However, the central GRF evaluation is circular in the statistical sense: the quantities that determine the match—material stiffness, foot position, and per-model phase offsets—are chosen by comparing to the very force-plate recordings used to compute the headline EM metric. The paper is transparent about this tuning and even about the offsets, but transparency does not make the match predictive. The joint-angle analysis is less confounded; however, it does not support an improvement over the bare SK model (Table 5 shows SK has higher mean R2 on five of six joints), so the only benefit claimed for soft feet is the GRF match, which is partly a fit. No self-citation chain or uniqueness theorem is load-bearing: references [10]-[12] describe the prior SK/control pipeline that is an input here, not a justification for the soft-foot result. The score reflects partial circularity in the evaluation protocol rather than a definitional identity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central result rests on several unvalidated modelling assumptions: uniform hyperelastic foot representation, simplified 15-vertex geometry, hard pin attachment, and hand-tuned stiffness and position parameters. The tuning is performed against the same GRF and joint angle data used for evaluation, so the model's agreement with measured data is partly a fitting result rather than an independent prediction.

free parameters (4)
  • Flex foot material stiffness (stiffness/damping or Young's modulus) = Models A/B/E: stiffness 2000, 5000, 12000; damping 100; equivalent Young's modulus 57, 110, 500 kPa
    Stiffness was progressively increased and chosen to improve GRF match (Section 2.3, Table 3); damping 100 held constant, originally damping 0.1 chosen via simulation experiments.
  • Damping coefficient for elasticity plugin = 0.1
    Chosen after simulation experiments to reduce oscillations (Section 2.3), not from direct tissue measurement.
  • Flex foot position offset relative to skeletal foot = Adjusted for model E; exact distance not reported
    Position of left flex foot was adjusted to improve GRF and joint angle results (Section 3.1).
  • GRF phase offsets applied to models A, B, E = Not reported; described as 'a slight offset'
    Offsets were applied to align GRF profiles with measured data before computing Experimental Match (Section 3.2), inflating apparent agreement.
assumptions (4)
  • domain assumption MuJoCo flex hyperelastic model with uniform material properties adequately represents human foot soft tissue
    Used in Sections 2.1 to 2.3; supported only by a general reference [9] on hyperelastic soft tissue and tissue property ranges, not by validation against foot-specific tissue deformation.
  • domain assumption SMPL-based foot shape from motion capture of the subject is a sufficient geometric representation after simplification to 15 vertices
    Feet are decimated from about 350 to 15 vertices (Section 2.1); the paper acknowledges shape simplification may cause phase differences and offset (Section 3.2).
  • domain assumption Hard pin connections between flex feet and skeleton preserve walking dynamics without unmodelled compliance
    Pins are chosen over welds and connects to avoid oscillation (Section 2.2), but the connection stiffness and location are not independently validated.
  • domain assumption The walking control policy trained only on kinematic observations transfers to the flex-foot model without otherwise retuning dynamics
    The paper compares control policies trained on SK and flex models; it is a core architectural assumption that kinematic-only training yields dynamically valid GRF.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Personalised 3D Human Digital Twin with Soft-Body Feet for Walking Simulation." pith.science (2026). https://pith.science/paper/CX4QHSTD

@misc{pith2026241114701,
  author       = {Pith},
  title        = {Pith review of: Personalised 3D Human Digital Twin with Soft-Body Feet for Walking Simulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CX4QHSTD}},
  note         = {Machine review of arXiv:2411.14701}
}
read the original abstract

With the increasing use of assistive robots in rehabilitation and assisted mobility of human patients, there has been a need for a deeper understanding of human-robot interactions particularly through simulations, allowing an understanding of these interactions in a digital environment. There is an emphasis on accurately modelling personalised 3D human digital twins in these simulations, to glean more insights on human-robot interactions. In this paper, we propose to integrate personalised soft-body feet, generated using the motion capture data of real human subjects, into a skeletal model and train it with a walking control policy. Through evaluation using ground reaction force and joint angle results, the soft-body feet were able to generate ground reaction force results comparable to real measured data and closely follow joint angle results of the bare skeletal model and the reference motion. This presents an interesting avenue to produce a dynamically accurate human model in simulation driven by their own control policy while only seeing kinematic information during training.

Figures

Figures reproduced from arXiv: 2411.14701 by the authors.

Figure 1
Figure 1. details the overall pipeline for the paper. Prior to this paper, the RRIS team implemented a pipeline where motion capture data [11] is used to generate a personalised 3D human skeletal (SK) model [10] which is then used in simulations. With MuJoCo as the physics engine, this model is trained with a walking control policy [12]. These efforts are represented in the bottom part of the figure. The top part of the figur… view at source ↗
Figure 2
Figure 2. Top row: Full body flex model with reduced resolution (Left) and Pin connections between the flex feet and the SK model, as shown by the spheres (Right). Bottom row: From left to right, Original high-resolution foot shape, Foot shape of reduced resolution, Mesh model of the foot, flex object in MuJoCo of the Foot foot shapes to reduce computational cost. The original shapes were simplified in Blender, from around 35… view at source ↗
Figure 3
Figure 3. Plot of vertical GRF against gait percentage for models SK, A, B, and E (Top row left to right, bottom row left to right). Note that for models A, B, and E, offsets were applied to better match the GRF profiles. SK model (model SK), the simplified flex foot with a material property at around the mid-range of muscle material properties (model A), a model with flex feet of material property at the higher-range of musc… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Joint angles for models SK, A, B, and E the GRF plots, the slight difference in the flex feet shape may have affected the joint angle outputs, as the models may require slightly different joint angles to complete the same motion. In this case, the convex hulls of the S…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 10 canonical work pages

  1. [1]

    Wearable assistive robotics: A perspective on current challenges and future trends,

    U. Martinez-Hernandez, B. Metcalfe, T. Assaf, L. Jabban, J. Male and D. Zhang, “Wearable assistive robotics: A perspective on current challenges and future trends,” Sensors21(20), 6751 (2021).https://doi.org/10.3390/s21206751

  2. [2]

    Ethical issues in patient safety research: Interpreting existing guidance,

    World Health Organization, “Ethical issues in patient safety research: Interpreting existing guidance,”https://iris.who.int/ handle/10665/85371, last accessed 2023/9/13

  3. [3]

    Welcome to the AMMR documentation!,

    AMMR, “Welcome to the AMMR documentation!,”https://anyscript.org/ammr- doc/index.html#, last accessed 2023/9/14

  4. [4]

    OpenSim: Simulating musculoskeletal dynamics and neuromuscular control to study human and animal movement,

    A. Seth, J. L. Hicks, T. K. Uchida, A. Habib, C. L. Dembia, J. J. Dunne, C. F. Ong, M. S. DeMers, A. Rajagopal, M. Millard, S. R. Hammer, E. M. Arnold, J. R. Yong, S. K. Lakshmikanth, M. A. Sherman, J. P. Ku and S. L. Delp, “OpenSim: Simulating musculoskeletal dynamics and neuromuscular control to study human and animal movement,” PLOS Computational Biolo...

  5. [5]

    Modeling and analysis of physical human-robot interaction of an upper body exoskeleton in assistive applications,

    S. Christensen, X. Li and S. Bai, “Modeling and analysis of physical human-robot interaction of an upper body exoskeleton in assistive applications,” Modeling, Identification and Control: A Norwegian Research Bulletin42(4), 159–172 (2021).https: //doi.org/10.4173/mic.2021.4.2

  6. [6]

    RCare World: A human- centric simulation world for Caregiving Robots,

    R. Ye, W. Xu, H. Fu, K. R. Jenamani, V. Nguyen, C. Lu, K. Dimitropoulou and T. Bhattacharjee, “RCare World: A human- centric simulation world for Caregiving Robots,” 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 33-40 (2022).https://doi.org/10.1109/IROS47612.2022.9982244

  7. [7]

    Parametric modeling of 3D human body shape—A survey,

    Z. Q. Cheng, Y. Chen, R. R. Martin, T. Wu and Z. Song, “Parametric modeling of 3D human body shape—A survey,” Computers & Graphics71, 88-100 (2018).https://doi.org/10.1016/j.cag.2017.11.008

  8. [8]

    XPBD: Position-Based Simulation of Compliant Constrained Dynamics,

    M. Macklin, M. Müller and N. Chentanez, “XPBD: Position-Based Simulation of Compliant Constrained Dynamics,” Motion In Games, 49-54 (2016).https://doi.org/10.1145/2994258.2994272

Show all 21 references
  1. [9]

    A hyperelastic model to capture the mechanical behaviour and histological aspects of the soft tissues,

    K. K. Dwivedi, P. Lakhani, S. Kumar and N. Kumar, “A hyperelastic model to capture the mechanical behaviour and histological aspects of the soft tissues,” Journal of the Mechanical Behavior of Biomedical Materials126, 105013 (2022). https://doi.org/10.1016/j.jmbbm.2021.105013

  2. [10]

    Investigation of Modeling Differences between OpenSim and Visual3D for Gait Analysis of Healthy Gait,

    W. T. Ang, "Investigation of Modeling Differences between OpenSim and Visual3D for Gait Analysis of Healthy Gait," 16th International Convention on Rehabilitation Engineering and Assistive Technology (i-CREATe ’23), Pathum Thani, Thailand, pp. 42-46 (2023).https://doi.org/10.1...

  3. [11]

    An Asian-centric human movement database capturing activities of daily living,

    P. Liang, W. H. Kwong, A. Sidarta, C. K. Yap, W. K. Tan, L. S. Lim, P. Y. Chan, C. K. K. Wee, S. K. Wee, K. Chua, C. Quek and W. T. Ang, "An Asian-centric human movement database capturing activities of daily living," Scientific Data7, 290 (2020). https://doi.org/10.1038/s4159...

  4. [12]

    Creation and Evaluation of Human Models with Varied Walking Ability from Motion Capture for Assistive Device Development,

    S. S. Chan, M. Lei, H. Johan, and W. T. Ang, "Creation and Evaluation of Human Models with Varied Walking Ability from Motion Capture for Assistive Device Development," 2023 International Conference on Rehabilitation Robotics (ICORR), Singapore, pp. 1-6. IEEE (2023).https://do...

  5. [13]

    SMPL: A Skinned Multi-Person Linear Model,

    M. Loper, N. Mahmood, J. Romero, G. Pons-Moll and M. J. Black, “SMPL: A Skinned Multi-Person Linear Model,” ACM Transition on Graphic34(6), pp. 1-16 (2015).https://doi.org/10.1145/2816795.2818013

  6. [14]

    A finite element model of abdominal human tissue for improving the accuracy in insulin absorption assessment: A feasibility study,

    P. Arpaia, D. Cuneo, S. Grassini and F. Mancino, “A finite element model of abdominal human tissue for improving the accuracy in insulin absorption assessment: A feasibility study,” Measurement Sensors18(11), 100218 (2021).https://doi.org/ 10.1016/j.measen.2021.100218

  7. [15]

    MuJoCo Documentation: Overview,

    DeepMind, “MuJoCo Documentation: Overview,” https://mujoco.readthedocs.io/en/stable/overview.html, last accessed 2023/9/23

  8. [16]

    Mechanical properties of whole-body soft human tissues: a review,

    G. Singh and A. Chanda, “Mechanical properties of whole-body soft human tissues: a review,” Biomedical Materials16(6), (2021). https://doi.org/10.1088/1748- 605X/ac2b7a

  9. [17]

    A review of the characterizations of soft tissues used in human body modeling: Scope, limitations, and the path forward,

    N. Arnold, J. Scott and T. R. Bush, “A review of the characterizations of soft tissues used in human body modeling: Scope, limitations, and the path forward,” Journal of Tissue Viability,32(2), 286-304 (2023).https://doi.org/10.1016/j.jtv.2023.02.003

  10. [18]

    Parametric study of lumbar belts in the case of low back pain: effect of patients’ specific characteristics,

    R. Bonnaire, W.-S. Han, P. Calmels, R. Convert and J. Molimard, "Parametric study of lumbar belts in the case of low back pain: effect of patients’ specific characteristics," Computational Biomechanics for Medicine, 43-59 (2020).https://doi.org/10. 1007/978- 3- 030- 15923- 8_4

  11. [19]

    Poisson’s ratio and Young’s modulus of lipid bilayers in different phases,

    T. Jadidi, M. R. Tabar and A. Mashaghi, “Poisson’s ratio and Young’s modulus of lipid bilayers in different phases,” Frontiers in Bioengineering and Biotechnology2(22), (2014). https://doi.org/10.3389/fbioe.2014.00008

  12. [20]

    Natural and Robust Walking using Reinforcement Learning without Demonstrations in High-Dimensional Musculoskeletal Models,

    P. Schumacher, T. Geijtenbeek, V. Caggiano, V. Kumar, S. Schmitt, G. Martius and D. F. B. Haeufle, “Natural and Robust Walking using Reinforcement Learning without Demonstrations in High-Dimensional Musculoskeletal Models,” arXiv, (2023). https://doi.org/10.48550/arXiv.2309.02976

  13. [21]

    Assessment of Waveform Similarity in Clinical Gait Data: The Linear Fit Method,

    M. Iosa, A. Cereatti, A. Merlo, I. Campanini, S. Paolucci, and A. Cappozzo, "Assessment of Waveform Similarity in Clinical Gait Data: The Linear Fit Method," Biomed Res Int, 214156 (2014).https://doi.org/10.1155/2014/214156

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.