Pith. sign in

REVIEW 3 major objections 5 minor 22 references

PB&J: Peanut Butter and Joints for Damped Articulation

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A passive elastic element on each finger is necessary for the three joints of a tendon-driven robot hand to flex together, and peanut-butter-filled rotary dampers cut joint settling time from about 4.9 s to 0.48 s.

desk verdict A creative, honest platform paper: PB&J dampers are new and the open hand is useful, but the 'elastic element is necessary' claim needs repeated trials before it is load-bearing. read the letter →

arxiv 2505.24860 v1 pith:N7UKW55R submitted 2025-05-30 cs.RO

Avery S. Williamson , Michael J. Bennington , Ravesh Sukhnandan , Mrinali Nakhre , Yuemin Mao , Victoria A. Webster-Wood This is my paper
classification cs.RO
keywords biomimetichandunderactuatedtendon-drivenfingerconcurrentflexionpassiveelasticelementrotaryviscoelasticdamperpeanutbutterrheologyballcatchingopen-sourcehardware
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's central finding is that a passive elastic strip running along the back of each finger is necessary for the three finger joints to flex together: pairwise joint-angle correlation rises from 0.350 (range [0.220, 0.443]) without the element to 0.959 (range [0.945, 0.972]) with it. This concurrent flexion is what lets the underactuated hand wrap around spherical objects instead of curling its fingertip first. The paper also models, fabricates, and tests concentric-cylinder rotary dampers filled with peanut butter, reducing pendulum settling time from 4.9 ± 0.5 s to 0.48 ± 0.04 s. A simple position-based controller using ball height lets the hand catch a lightweight falling ball in 59% of 22 trials. The broader aim is an open, low-cost, bioinspired platform for studying joint viscoelasticity in physical robots.

What carries the argument

Two load-bearing mechanisms carry the argument. First is a silicone ligament analog with an estimated effective rotational stiffness of $6.6 \times 10^{-2}$ N m/rad, fixed along the back of each finger; it supplies the parallel elasticity that synchronizes joint motion. Second is a concentric-fin rotary damper whose torque model is $T = -\mu G \dot{\theta}$, where $\mu$ is working-fluid viscosity, $\dot{\theta}$ is joint angular velocity, and $G$ is a geometry factor derived from a Couette-flow analysis of five interdigitated fins with 0.4 mm channels. Peanut butter, with a reported apparent viscosity of 150,000 to 250,000 cP, is the working fluid, and pendulum drop tests with bootstrapped parameter estimation identify the damping coefficient. The paper's key quantitative comparison is the pairwise correlation of normalized joint-angle time series with versus without the elastic element.

What would settle it

Replace the silicone strip with a non-elastic string under the same tension: if the pairwise joint-angle correlation stays near 0.959, then the elastic property itself is not the active ingredient. Separately, measure damper torque with a controlled, continuous peanut butter film at known shear rates: if the coefficient stays near $0.759 \times 10^{-3}$ N m s/rad, then the Newtonian fluid model, not fabrication voids, explains the shortfall.

Watch

Extended reading notes

Core claim

For a tendon-driven finger with two articulated phalanges plus a fingertip, concurrent flexion of all three joints does not emerge from tendon actuation alone; it requires a passive elastic element in parallel with the joints. With the silicone ligament analog, pairwise joint-angle correlations reach 0.959 (range [0.945, 0.972]), while without it the finger articulates distal-first and the correlation drops to 0.350 (range [0.220, 0.443]). The paper's damper work shows that interdigitated concentric fins moving through peanut butter, modeled with the Couette-flow torque law $T = -\mu G \dot{\theta}$, reduce joint oscillation from 8.0 ± 1.0 swings over 4.9 ± 0.5 s to one swing over 0.48 ± 0.04 s. The measured damper coefficient is $0.759 \times 10^{-3}$ N m s/rad (95% credible interval [0.671, 0.920] × 10⁻³), about 12 to 20 times below the predicted 9 to 15 × 10⁻³ N m s/rad; the authors attribute the gap to peanut butter being a thixotropic Bingham plastic rather than a Newtonian fluid and to the 400-micron film being discontinuous.

Load-bearing premise

The load-bearing premise is that the damper's working fluid, peanut butter, behaves as a Newtonian fluid with an apparent viscosity of 150,000 to 250,000 cP and that the 400-micron film between the damper fins is continuous; the paper itself notes that peanut butter is actually a thixotropic Bingham plastic and that the film was likely discontinuous.

Editorial extensions

If this is right

  • Robotic hand designers can obtain synchronized multi-joint flexion from a single tendon actuator by adding passive parallel elasticity, without additional motors or sensors.
  • The modular rotary damper inserts reduce joint settling time by roughly a factor of ten, making underactuated grippers more stable against impact-induced oscillation at low cost.
  • If the damper's film-continuity and fluid-modeling issues are fixed, the same concentric-fin geometry should reach the human joint-damping range of $[8.1, 14.2] \times 10^{-3}$ N m s/rad.
  • A hand with these passive viscoelastic elements can catch a falling ball with a simple proportional position controller, showing that passive mechanics can compensate for limited control bandwidth.
  • The open-source design gives experimenters independent control of joint stiffness and damping to test neuromechanical hypotheses in a physical platform.

Reading between the lines

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

  • The necessity of parallel elasticity for concurrent flexion likely extends beyond this hand: any underactuated, serially routed tendon drive with multiple joints may show the same distal-first cascade unless a passive restoring element is present.
  • The damper shortfall suggests peanut butter's shear-thinning behavior could be exploited deliberately: choosing a bioderived fluid with a flatter viscosity-versus-shear curve, or engineering thicker channels, might tune damping without synthetic oils.
  • The pairwise-correlation metric could serve as a standard benchmark for comparing passive versus active strategies for joint coordination in biomechanical models.
  • If concurrent flexion is genuinely ligament-mediated, then emulating human ligament placement may offer a lower-power alternative to impedance control for precision grasping.
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

3 major / 5 minor

Summary. The paper presents the design, fabrication, and characterization of a low-cost, open-source, tendon-driven five-finger robotic hand with passive silicone ligament analogs and rotary dampers filled with peanut butter. The main empirical claims are: (i) a parallel elastic element is necessary for concurrent flexion of all three finger joints, supported by average pairwise joint-angle correlations of 0.959 with the element versus 0.350 without it (Sec. 4.1); (ii) the fabricated peanut-butter dampers significantly reduce joint oscillation, with pendulum settling time dropping from 4.9 +/- 0.5 s to 0.48 +/- 0.04 s (Sec. 4.2); and (iii) a simple position-based controller catches a falling ball in 59% of 22 trials (Sec. 4.3). The manuscript includes a Couette-flow model of the dampers, a Simulink simulation used for motor selection, and a bootstrapped parameter-estimation routine for the pendulum dynamics.

Significance. If the concurrent-flexion result is repeatable, it provides a simple and transferable design principle for underactuated hands: a parallel passive elastic element can synchronize joint motion without additional sensing or control. The paper's strengths include the open-source release of the design and code, the use of accessible and bioderived materials, and the direct quantitative comparison of correlated joint motion with and without the elastic element. The damper section is honest in reporting a large (12-20x) discrepancy between the predicted and measured damping coefficient, but that discrepancy weakens the predictive value of the model as presented. The overall significance is moderate: the platform is useful for exploratory studies, but the load-bearing necessity claim needs stronger evidence before the design rationale is fully established.

major comments (3)
  1. [§3.3 / §4.1] The central claim that the parallel elastic element is 'vital' and 'critical' for concurrent flexion rests on a single motor sweep per condition. Section 3.3 describes one sweep from 0 to 270 degrees with 10-degree increments for each condition, and the reported correlation ranges [0.945, 0.972] and [0.220, 0.443] are the spread across the three joint pairs within a single trial, not across repeated trials or fingers. No standard deviation, confidence interval, or significance test is reported for the difference. Because the manuscript states the element is necessary for grasping, I request repeated measurements (at least 3-5 trials and ideally multiple fingers) with mean, spread, and a statistical comparison; otherwise, the observed separation could reflect drift, noise, or a single atypical assembly rather than a repeatable design property.
  2. [§2.4 / §4.2 / Appendix A] The damper model predicts a damping coefficient of approximately 9-15 x 10^-3 N m s/rad, but the measured value is 0.759 x 10^-3 N m s/rad, a 12-20x overprediction. The authors correctly attribute this to non-Newtonian, thixotropic peanut-butter behavior, non-steady flow, and a discontinuous 400-micron film, but the consequence is that the model as presented cannot be used to select working-fluid viscosity or damper geometry for a target damping level. Please recalibrate the model to the measured data, present the analytical prediction explicitly as an upper bound or design heuristic with the stated limitations, or provide an independent validation measurement so that the model has predictive value.
  3. [§4.2] The statement that the pendulum data 'fit well' with the model (Fig. 4(e)) is weak evidence because the damping coefficient b and the friction coefficients mu_k and mu_d are fitted to the same data used to judge the fit. A good fit of a multi-parameter model to calibration data does not validate the model. Please provide an identifiability analysis, cross-validation, or a comparison of the fitted b against an independent measurement (for example, a constant-velocity torque test of the damper).
minor comments (5)
  1. [Appendix A, Eq. (4)] Equation (4) appears to contain a typographical error: the second term in the braced expression uses (rho - w/2)^2 where the notation elsewhere and the context indicate (rho_i - w/2)^2.
  2. [§4.2] The reported measured damping coefficient is given inconsistently: '0.759 [0.671, 0.920] x 10^-3' appears earlier, and '0.758 [0.670, 0.939] x 10^-3' appears later in the same section. Please harmonize the value and the credible interval.
  3. [Abstract / Conclusions] The phrase 'in roboto' is used without definition; please define it on first use or replace it with clearer wording (for example, 'in robot experiments').
  4. [References] Reference [1] is incomplete (no author, year, or venue), and references [19] and [20] lack full bibliographic details; please complete the bibliography.
  5. [§3.3] There is a minor typesetting error in 'The mapping offingertip position'; also, the tracking description would benefit from stating how marker occlusion or crossing was handled during the motor sweep.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central claims rest on direct experimental comparison and first-principles models, with discrepancies explicitly acknowledged.

full rationale

The paper's two main quantitative claims are empirical comparisons and first-principles derivations, not reductions to their inputs. The concurrent-flexion claim is a direct experimental contrast: the paper measures joint-angle time series under two conditions (with and without a passive elastic element) and computes pairwise correlations, reporting 0.959 with the element versus 0.350 without it. The metric is operationally defined by the measurement, and the causal claim 'elastic element necessary' is an interpretation of that controlled comparison, not a quantity fitted from the conclusion. The damper model is derived from Couette-flow geometry and a literature-reported viscosity range (150,000-250,000 cP), giving a predicted damping coefficient of roughly 9-15 x 10^-3 N m s/rad; the measured value (0.759 x 10^-3 N m s/rad) is obtained from pendulum identification and is explicitly reported as much lower than the model prediction. Thus no fitted parameter is renamed as a prediction; the paper instead acknowledges the model-data gap and attributes it to non-Newtonian fluid behavior and fabrication discontinuities. Human joint parameters cited in the design are taken from external physiology literature (Kamper et al.), not from the authors' own prior work, and no uniqueness theorem or self-citation is invoked to force a choice. The limitations noted in the text (single-sweep flexion comparison without repeatability statistics, Newtonian-fluid assumption, thin-film discontinuities) are correctness and robustness concerns, not circularity. The derivation chain is therefore self-contained against external benchmarks and shows no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central experimental results (concurrent flexion, damping, catching) rely on a handful of fitted or assumed parameters, but no new physical entities are introduced. The damper model rests on a Newtonian-fluid assumption that the paper itself later rejects, which is the most consequential assumption in the ledger.

free parameters (6)
  • Damping coefficient b (pendulum fit) = 0.759 [0.671, 0.920] x 10^-3 N m s/rad (mean [95% CI])
    Fit to damped pendulum drop data via bootstrapped parameter estimation (Sec 4.2).
  • Kinetic friction coefficient mu_k = 2.88 [2.66, 3.10] x 10^-3
    Fit to damper-free pendulum data (Sec 4.2).
  • Viscous friction coefficient mu_d = 0.0013 [0.0000, 0.0118] x 10^-3 (not significant)
    Fit to damper-free pendulum data; right-skewed and near zero (Sec 4.2).
  • Controller parameters (threshold y_t, proportionality constant, capture height y_c, final distance du) = Experimentally tuned, values not reported
    Tuned to improve catching performance (Sec 2.3).
  • Effective rotational stiffness of silicone ligament = 6.6 x 10^-2 N m/rad
    Estimated from shear modulus ~2.4 kPa from tensile test, data not shown (Sec 2.1).
  • Equivalent stiffness of Nylon tendon threads = 9.52 N/mm
    Designed to mimic human tendon stiffness using 24 parallel Nylon threads (Sec 2.1).
assumptions (5)
  • domain assumption Peanut butter behaves as a Newtonian fluid with viscosity 150,000-250,000 cP
    Used to select working fluid and predict damper torque in Sec 2.4; later acknowledged as thixotropic Bingham plastic in Sec 4.2.
  • domain assumption Steady-state circumferential Couette flow in the damper with rotational symmetry
    Basis of Appendix A model derivation.
  • domain assumption Pendulum weight modeled as a point mass, with Coulomb plus viscous friction at the joint
    Basis of Appendix B pendulum dynamics model.
  • domain assumption Human joint damping and stiffness values from Kamper et al. [11] apply at this hand's scale
    Used in the Simulink model to size the motor (Sec 2.2).
  • domain assumption Finger lengths scaled from measurements of one author are representative of human hand proportions
    Guides the CAD design in Sec 2.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PB&J: Peanut Butter and Joints for Damped Articulation." pith.science (2026). https://pith.science/paper/N7UKW55R

@misc{pith2026250524860,
  author       = {Pith},
  title        = {Pith review of: PB&J: Peanut Butter and Joints for Damped Articulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N7UKW55R}},
  note         = {Machine review of arXiv:2505.24860}
}
read the original abstract

Many bioinspired robots mimic the rigid articulated joint structure of the human hand for grasping tasks, but experience high-frequency mechanical perturbations that can destabilize the system and negatively affect precision without a high-frequency controller. Despite having bandwidth-limited controllers that experience time delays between sensing and actuation, biological systems can respond successfully to and mitigate these high-frequency perturbations. Human joints include damping and stiffness that many rigid articulated bioinspired hand robots lack. To enable researchers to explore the effects of joint viscoelasticity in joint control, we developed a human-hand-inspired grasping robot with viscoelastic structures that utilizes accessible and bioderived materials to reduce the economic and environmental impact of prototyping novel robotic systems. We demonstrate that an elastic element at the finger joints is necessary to achieve concurrent flexion, which enables secure grasping of spherical objects. To significantly damp the manufactured finger joints, we modeled, manufactured, and characterized rotary dampers using peanut butter as an organic analog joint working fluid. Finally, we demonstrated that a real-time position-based controller could be used to successfully catch a lightweight falling ball. We developed this open-source, low-cost grasping platform that abstracts the morphological and mechanical properties of the human hand to enable researchers to explore questions about biomechanics in roboto that would otherwise be difficult to test in simulation or modeling.

Figures

Figures reproduced from arXiv: 2505.24860 by the authors.

Figure 1
Figure 1. System Design Overview: (a) Isometric view of CAD assembly depicting 3D printed components. (b) CAD model of (b1) finger tip and (b2) finger digit. (c) Simulink Simscape model of hand and tendon system showing angular position θ of articulated finger segments. (d) Mapping of articulated finger segment angle θ with respect to ball height. All scale bars: 10mm. inertial dynamics and viscoelastic forces at the joints. … view at source ↗
Figure 2
Figure 2. (a) Mechatronics system schematic. A Teensy 4.1 microcontroller was used to integrate signals from an IR distance sensor and motor encoders and to drive the tendon motor. (b) The Teensy implemented a closed-looped controller based on the height of the ball above the palm [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Viscous Damping Component Overview (a) CAD model and cross section of final concentric fin damper design. (b) Parametric sweep of normalized damping coef￾ficient (G, reported in units 10−6 m3 ). The G factor is impacted by the channel width and wall width for a given number of internal fins (N). The region denoted (1) contains parameters below the printing tolerances of the printers used to fabricate the dampers, an… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Experimental characterization of robotic subsystems. (a) Robot finger behavior during contraction without (a1) and with (a2) an elastic element. Concurrent flexion is only observed with a parallel elastic element (Scale bar: 30mm). (b) Range-of-motion normalized time s…
Figure 5
Figure 5. Figure 5: Sequential frames demonstrating a successful ball catch by the robot hand. The ball is highlighted by a red dashed circle in the first frame. in time to intercept. Concurrent flexion of the joints facilitated the wrapping of the fingers around the ball once to arrest i…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [11]

    Kamper, D.G., Hornby, T.G., Rymer, W.Z.: Extrinsic flexor muscles generate con- current flexion of all three finger joints (2002)

    work page 2002

  2. [1]

    By leaps and bounds an exclusive look at how boston dynamics is redefining robot agility by erico guizzo• photography by bob o’connor,

    work page

  3. [2]

    Springer Berlin Heidelberg (1981)

    Fundamentals of Sensory Physiology. Springer Berlin Heidelberg (1981)

    work page 1981

  4. [3]

    Burridge, R.R., Rizzi, A.A., Koditschek, D.E.: Sequential composition of dynami- cally dexterous robot behaviors

    work page

  5. [4]

    PLoS ONE11 (7 2016)

    Cesqui, B., Russo, M., Lacquaniti, F., D’Avella, A.: Grasping in one-handed catch- ing in relation to performance. PLoS ONE11 (7 2016)

    work page 2016

  6. [5]

    Rheologica Acta 40(1), 86–96 (2001)

    Citerne, G.P., Carreau, P.J., Moan, M.: Rheological properties of peanut butter. Rheologica Acta 40(1), 86–96 (2001)

    work page 2001

  7. [6]

    Vibration4, 482–528 (6 2021)

    Dong, R.G., Wu, J.Z., Xu, X.S., Welcome, D.E., Krajnak, K.: A review of hand–arm vibration studies conducted by us niosh since 2000. Vibration4, 482–528 (6 2021)

    work page 2000

  8. [7]

    Guizzo, E., Ackerman, E.: DARPA Robotics Challenge: A Compilation of Robots Falling Down - IEEE Spectrum,

    work page

Show all 22 references
  1. [8]

    Developmental Neuroscience 40(5-6), 586–600 (2018)

    Holme Nielsen, C., Bladt Brandt, A., Thymann, T., Obelitz-Ryom, K., Jiang, P., Vanden Hole, C., van Ginneken, C., Pankratova, S., Sangild, P.T.: Rapid Postna- tal Adaptation of Neurodevelopment in Pigs Born Late Preterm. Developmental Neuroscience 40(5-6), 586–600 (2018)

    2018

  2. [9]

    Mathematical Proceedings of the Cambridge Philosophical Society46, 469– 478 (1950)

    Illingworth, C.R.: Some solutions of the equations of flow of a viscous compressible fluid. Mathematical Proceedings of the Cambridge Philosophical Society46, 469– 478 (1950)

    1950

  3. [10]

    Experimental Brain Research188, 223–236 (6 2008)

    Ingram, J.N., Körding, K.P., Howard, I.S., Wolpert, D.M.: The statistics of natural hand movements. Experimental Brain Research188, 223–236 (6 2008)

    2008

  4. [12]

    Journal of Physiology590, 2471–2483 (5 2012)

    Lakie, M., Vernooij, C.A., Osborne, T.M., Reynolds, R.F.: The resonant compo- nent of human physiological hand tremor is altered by slow voluntary movements. Journal of Physiology590, 2471–2483 (5 2012)

    2012

  5. [13]

    Journal of Neuroscience27, 9354–9368 (8 2007)

    Liu, D., Todorov, E.: Evidence for the flexible sensorimotor strategies predicted by optimal feedback control. Journal of Neuroscience27, 9354–9368 (8 2007)

    2007

  6. [14]

    CRC Press (1994)

    Murray, R.M., Li, Z., Sastry, S.: A mathematical introduction to robotic manipu- lation. CRC Press (1994)

    1994

  7. [15]

    Namiki, A., Imai, Y., Ishikawa, M., Kaneko, M.: Development of a high-speed multifingered hand system and its application to catching. vol. 3, pp. 2666–2671 (2003)

    2003

  8. [16]

    Nguyen-Tuong, D., Peters, J.: Model learning for robot control: A survey (11 2011)

    2011

  9. [17]

    Integrative and Comparative Biology47(1), 16– 54 (Jul 2007)

    Nishikawa, K., Biewener, A.A., Aerts, P., Ahn, A.N., Chiel, H.J., Daley, M.A., Daniel, T.L., Full, R.J., Hale, M.E., Hedrick, T.L., Lappin, A.K., Nichols, T.R., Quinn, R.D., Satterlie, R.A., Szymik, B.: Neuromechanics: an integrative approach for understanding motor control. I...

    2007

  10. [18]

    Medicine & Science in Sports & Exercise35(8), 1397 (Aug 2003)

    Owings, T.M., Lancianese, S.L., Lampe, E.M., Grabiner, M.D.: Influence of Ball Velocity, Attention, and Age on Response Time for a Simulated Catch. Medicine & Science in Sports & Exercise35(8), 1397 (Aug 2003)

    2003

  11. [19]

    Scientific, P.: Viscosity of peanut butter cream (2018), , accessed: 2023-04-17

    2018

  12. [20]

    Smooth-On, I.: Viscosity scale reference guide, , accessed: 2023-04-17 16 Authors Suppressed Due to Excessive Length

    2023

  13. [21]

    Sutton, G., Szczecinski, N., Quinn, R., Chiel, H.: Neural control of rhythmic limb motion is shaped by size and speed

  14. [22]

    Journal of Neurophysiology113, 647–656 (1 2015)

    Vernooij, C.A., Lakie, M., Reynolds, R.F.: The complete frequency spectrum of physiological tremor can be recreated by broadband mechanical or electrical drive. Journal of Neurophysiology113, 647–656 (1 2015)

    2015

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.