Pith. sign in

REVIEW 4 major objections 6 minor 46 references

This paper claims that embedding an analytical stiffness mapping into a cascade neural network lets an underactuated robotic hand predict grasp stability and in-hand tool displacement under load with near-zero energy-consistency violations.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 17:41 UTC pith:OF26YRAE

load-bearing objection Clever hybrid architecture for underactuated-hand compliance, but every empirical claim rests on a simulator built from the same model family the network learns—worth a referee's time, yet the numbers are proof-of-concept, not validation. the 4 major comments →

arxiv 2607.17541 v1 pith:OF26YRAE submitted 2026-07-20 cs.RO

Predicting Grasping Compliance in Robotic Hands through Analytical-Model-Informed Neural Networks

Qianwen Zhao , Long Wang This is my paper
classification cs.RO
keywords grasp stabilitygrasping complianceunderactuated handsanalytical-model-informed neural networkin-hand displacementCartesian stiffnesspassivityforceful tool use
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that for forceful tool use, grasping should be evaluated by how much the tool moves in the hand under load, not just by binary success or failure. It introduces AMINN, a hybrid network that embeds a parallel-mechanism stiffness model into a two-stage cascade: first predicting whether the grasp remains stable, then predicting in-hand tool displacement for stable grasps. In simulation experiments on a three-finger tendon-driven hand, AMINN matches a black-box MLP on stability F1 while improving displacement accuracy and reducing a passivity-violation metric from 0.65 to 0.00, meaning its predicted motions almost never imply net energy generation opposite to the applied force. The authors claim this demonstrates that analytically informed learning can keep predictive performance while making outputs physically plausible and more trustworthy for safety-critical manipulation.

Core claim

The paper's central claim is that a neural network can predict how a compliant underactuated hand and a grasped tool respond to external loading—both whether the grasp stays stable and how far the tool moves in the hand—by building the analytical Cartesian stiffness relation of the hand–tool system directly into the network. The key math is the parallel-mechanism mapping K_x = J_x^T (J_q K_q^{-1} J_q^T)^{-1} J_x, which turns learned joint-space stiffness into a wrench-to-displacement map; displacement is then Δx = K_x^{-1} w_e. Reported on the IID test split, the resulting AMINN cascade matches a black-box MLP on stability F1 (0.96 vs 0.95), improves displacement RMSE from 0.08 to 0.05, and

What carries the argument

The central object is the analytical Cartesian stiffness mapping for the hand–tool system treated as a parallel mechanism: the grasped tool is the moving platform, the palm is the base, each finger a branch. The identity K_x = J_x^T (J_q K_q^{-1} J_q^T)^{-1} J_x turns joint-space stiffness into a wrench-to-displacement map. In AMINN this runs as a differentiable two-pass layer: pass 1 computes a spring-only stiffness and baseline deflection; pass 2 estimates tendon engagement from how well that deflection aligns with each tendon direction, assembles effective joint stiffness, and recomputes K_x and displacement Δx = K_x^{-1} w_e. This forces predicted motions to respect the sign of the appli

Load-bearing premise

Every reported number depends on the simulator reproducing the real hand's behavior—no physical experiment is included—so if friction, hysteresis, or other unmodeled effects are significant in hardware, the accuracy and passivity results may not transfer.

What would settle it

Set up the instrumented hand to grasp a tool with a motion-capture marker, apply a known wrench through the wrist force/torque sensor, and measure the tool's actual in-hand displacement; if any stable-grasp case shows displacement opposite to the applied force, or if measured displacement errors greatly exceed the simulated RMSE, the paper's central claims are contradicted.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • A near-zero passivity-violation rate means predicted displacement directions can be trusted for stable grasps, enabling safety checks before execution during forceful tool use.
  • The model answers not only 'will the tool stay in the hand?' but 'how far and in which direction will it move under this load?'—the quantity needed for precision tasks.
  • The two-stage cascade aligns with deployment logic: first filter out unstable grasps, then estimate displacement only where in-hand deformation is physically meaningful.
  • The reported improvements (RMSE 0.05 vs 0.08, passivity 0.00 vs 0.65) suggest that adding mechanics-informed structure does not cost accuracy in the simulated setting.

Where Pith is reading between the lines

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

  • A stronger physics test than the paper's translational signed-work proxy would include rotational work; if zero violations persist there too, the passivity claim is more convincing.
  • If the approach transfers, grasp planners could treat predicted displacement magnitude as a task constraint, rejecting grasps or load directions that move the tool beyond tolerance.
  • The same analytical-layer recipe could be reused for other hand architectures by swapping the Jacobians, but whether the learned tendon-gating generalizes outside this hand's kinematics is untested.
  • Because the simulator shares the same spring-plus-tendon model family that the network learns, a hardware experiment with the instrumented hand is the decisive test the paper does not perform.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The manuscript proposes AMINN, a two-stage cascade predictor for underactuated robotic-hand grasping under external wrench. Stage 1 classifies grasp stability; Stage 2 regresses six-DOF in-hand tool displacement for stable predictions. The network embeds an analytical compliance layer derived from a parallel-mechanism model: branch-wise SPD spring blocks plus tendon-engagement terms are assembled into a joint-space stiffness K_eff^q, mapped to Cartesian K_x via Eq. (7), and used to generate displacement as K_x^{-1} w_e (Eq. 19). The architecture is evaluated in simulation against a cascade MLP baseline using three metrics: F1, RMSEpose, and a signed-work passivity-violation rate. Reported results (Table 1) favor AMINN on all three metrics at a selected operating point. The analytical derivation is standard and the model is mechanically interpretable; however, all quantitative results are generated by a simulator described as 'consistent with' the same analytical model family, and no hardware experiments are reported.

Significance. If the predictive performance were demonstrated on a real underactuated hand, AMINN would be a useful contribution: it combines interpretable analytical structure with learned nonlinear behavior and makes a concrete claim of improved physical consistency over a black-box baseline. The kinematic derivation and the network design are sound, and the notion of using a differentiable analytical stiffness layer is appealing. The paper does not ship code or data, so reproducibility cannot be assessed. The main value is conditional on evidence that the simulator represents hardware; the current evidence does not establish that.

major comments (4)
  1. [Sec. 6.1 / 7.2 / Table 1] The evaluation data are produced by a simulator that is 'consistent with the tendon-driven underactuated architecture and the parallel-robot kinematic model in Sec. 5.2' and uses the same spring-plus-tendon compliance representation embedded in AMINN's analytical layer (Eqs. 7, 10, 16-18). The targets themselves (Δx_HT, s) come from this simulator. Consequently, the reported F1/RMSE/passivity numbers may reflect in-family self-consistency rather than predictive skill on the physical hand described in Sec. 5.3. No hardware validation or simulator calibration is reported. Since every empirical claim in the abstract and conclusion depends on Table 1, this is a load-bearing gap. Please add hardware experiments, or at minimum evaluate on an independently implemented simulator with different physics (e.g., finite element contacts, friction, hysteresis) and report calibration against the real h
  2. [Sec. 4.2.4 / Eq. 19 / Sec. 7.1] The near-zero passivity-violation rate is a structural consequence of the model, not an empirical achievement. With Δx̂ = K_x^{-1} w_e and K_x constructed symmetric positive definite (Eqs. 10, 16-18), w_e^T Δx̂ is always nonnegative. Thus the Table 1 passivity comparison is biased: the MLP baseline has no such constraint. To make the comparison meaningful, either enforce the same positive-definite stiffness on the baseline (e.g., via a PSD layer) or evaluate a metric that includes full wrench/displacement pairs and off-diagonal terms. As written, the 'energy-based physical consistency' claim is overstated.
  3. [Table 1 / Sec. 7.2] The quantitative comparison consists of three point estimates. No error bars, number of runs, dataset size, or standard deviation are reported. It is therefore impossible to judge whether F1 0.96 vs 0.95 and RMSEpose 0.05 vs 0.08 are significant. Additionally, no ablation removes the analytical layer or replaces it with an equivalent-capacity unconstrained module, so the contribution of the structured layer is not isolated. Please provide repeated-seed statistics, dataset sizes, and an ablation.
  4. [Sec. 4.3.3 / Sec. 7.1] The cascade operating-point selection is described only textually. The threshold τ, the choice criterion on validation data, and the resulting precision/recall of the stability filter are not reported. Since Stage 2 is trained only on ground-truth stable samples but evaluated on predicted-stable samples, the train/eval distribution mismatch should be quantified (e.g., by reporting E[Δx] on true-positive vs false-positive stable samples). This is important for interpreting RMSEpose.
minor comments (6)
  1. [Sec. 5.2, Eq. (29)] The definition of B_i uses cross-products but does not specify the branch joint axes and lengths (a_i, b_i) in coordinates; please include for reproducibility.
  2. [Sec. 6.1] The phrase 'consistent with' is vague; specify the physics engine, contact model, solver settings, and friction parameters, and release code and data for reproducibility.
  3. [References] References [24], [31], [34], and [41] contain inserted editorial notes ('Key used here as shorthand...') that should be removed before publication.
  4. [Fig. 9 / Sec. 7.1] Axis labels and units are missing in Fig. 9; also, the signed-work proxy should be described as translational only, not as a full passivity test.
  5. [Notation / Sec. 7] The terms 'Cascade AMINN' and 'AMINN' are used interchangeably; define the baseline 'Cascade MLP' explicitly in Sec. 7.1.
  6. [Sec. 4.2.4, Eq. (12)] The inverse of (J_q + εI) is used, but J_q may be rectangular; specify pseudo-inverse or regularization choices.

Circularity Check

1 steps flagged

The reported passivity-violation rate is forced by construction: the model outputs Δx = K_x^{-1}w_e with K_x positive definite, so signed work w_e·Δx is always positive.

specific steps
  1. self definitional [Sec. 4.2.4 Eq. (19); Sec. 7.1; Table 1]
    "ΔˆxHT = K −1 x w e. (19) ... we compute a signed-work value from the translational force and predicted translational displacement for each sample. Negative signed work is counted as a passivity violation. ... Table 1 ... Passivity↓ 0.00 0.65"

    The AMINN displacement output is defined as K_x^{-1} w_e, and K_x is constructed to be positive definite: each spring block is constrained K_{s,b} ≻ 0 (Eq. 10), tendon terms are PSD, the inner regularized term is SPD, and the outer expression adds ε_{Kx} I. Hence w_e^T Δx = w_e^T K_x^{-1} w_e > 0 for every nonzero wrench. The signed-work passivity proxy can never be negative, so the 0.00 passivity-violation rate in Table 1 is an algebraic consequence of the chosen output layer, not an empirical or learned result. Comparing this with an unconstrained MLP baseline is therefore a definitional comparison for this metric, not evidence that the network learned physical consistency.

full rationale

I found one genuine circular step: the headline 'energy-based physical consistency' result, specifically the 0.00 passivity-violation rate, is guaranteed by the model's own definition. Since Eq. (19) sets predicted displacement to K_x^{-1} w_e and K_x is positive semidefinite-plus-regularization by construction, the translational signed-work proxy is always positive. This is not an empirical finding and does not support the claim that AMINN 'achieves' better physical consistency in a way that could have gone either way. The other empirical claims (stability F1, displacement RMSE) are ordinary supervised-learning results on simulator-generated data and are not circular by the paper's own equations, although their external validity is limited because the simulator is described only as 'consistent with' the same parallel-mechanism/tendon-compliance model family and no hardware validation is reported. I did not count the simulator-consistency issue as an additional circular step because the paper does not show that the simulator's ground truth is exactly the same as Eq. (7); that is a validation-independence concern rather than an in-text reduction. There are no load-bearing self-citations, imported uniqueness theorems, or ansatz-smuggling chains. Overall, partial circularity: one headline metric reduces by construction, while the core learning pipeline has independent content.

Axiom & Free-Parameter Ledger

8 free parameters · 6 axioms · 1 invented entities

The ledger shows what the paper adds versus what it assumes. Every mechanical quantity inside the analytical layer—spring blocks, tendon gains and directions, gate parameters—is learned from simulation data, not measured or derived; the simulator that generates all ground truth is itself uncalibrated against hardware. The structural priors (parallel-mechanism representation, spring-plus-rank-one-tendon decomposition, pass-1-informed gating) are domain assumptions the paper does not defend with independent evidence. No new physical entities are introduced; the tendon-engagement gate is a latent construct with no independent measurement.

free parameters (8)
  • Branch spring blocks K_s,b (SPD) = learned, values not reported
    The mechanics head outputs per-branch SPD spring stiffness (Eq. 10); these are the core learned quantities of the analytical layer.
  • Tendon gains k_b = learned
    Per-branch scalar tendon stiffness magnitudes forming rank-one tendon terms (Eq. 15).
  • Tendon direction vectors d_b = learned
    Per-branch tendon directions used in the rank-one tendon stiffness (Eq. 15).
  • Gate parameters alpha_b, beta_b, rho_b = learned
    Parameterize sigmoid tendon engagement from pass-1 deflection (Eqs. 13-14).
  • Loss weights lambda_tr, lambda_rot, lambda_gate, lambda_ten = not reported
    Balance pose fitting versus structural regularization (Eqs. 22-24); values affect the trained behavior and are not given.
  • Simulator parameters (springs, tendon routing, friction, contacts) = not disclosed
    The simulator (Sec. 6.1) generates all ground-truth labels; its compliance parameters determine the dataset and are not reported.
  • Synergy subspace projection A, offset o, scaling delta = hand-defined
    Sampling subspace (Sec. 6.2) restricts grasp diversity to three axes and shapes the training distribution.
  • Regularization eps_M, eps_Kx, eps_Jq and cascade threshold tau = not reported / validation-selected
    Numerical-stability terms (Eqs. 11, 12, 18) and the operating point chosen on validation (Sec. 4.3.3).
axioms (6)
  • domain assumption Quasi-static, rigid-body, Coulomb-frictional contact regime; high-speed transients out of scope.
    Stated in Sec. 3; the entire stiffness mapping Δw ≈ Kx Δx and the displacement targets presuppose this regime.
  • domain assumption The hand-tool system is a parallel mechanism with palm as base and tool as moving platform.
    Sec. 5.2 (Eqs. 25-30); underpins the Jacobians Jx, Jq used in the analytical layer and given as model inputs.
  • ad hoc to paper Joint compliance decomposes into branch-wise SPD springs plus rank-one tendon terms with sigmoid engagement gating.
    Eqs. 10-17; the core structural prior of AMINN. If real tendon-driven compliance has inter-branch coupling or rate-dependent terms not captured by this decomposition, predictions are misspecified.
  • ad hoc to paper Pass-1 spring-only deflection predicts tendon engagement.
    Eqs. 13-14; the two-pass inference assumes tendon activity is a deterministic function of the spring-only response, with no tendon-tension measurement supervising it.
  • domain assumption The simulator is a valid oracle for ground-truth stability and in-hand displacement.
    Sec. 6; all labels come from a simulator built 'consistent with' the analytical model. No hardware validation is reported, so the oracle's fidelity is unverified.
  • standard math Cartesian stiffness Kx is locally constant over the displacement increment.
    Eq. 4; standard linearization used throughout grasp-stiffness literature.
invented entities (1)
  • Tendon-engagement gate g_b no independent evidence
    purpose: Latent in [0,1] modulating per-branch tendon stiffness based on pass-1 deflection alignment; promoted to confident 0/1 by L_gate.
    A postulated internal quantity never measured or validated against tendon-tension sensing; its correctness is only indirectly assessed via RMSE on displacement.

pith-pipeline@v1.3.0-alltime-deepseek · 12888 in / 24438 out tokens · 205007 ms · 2026-08-01T17:41:26.484721+00:00 · methodology

0 comments
read the original abstract

In robotic manipulation studies, grasping is often treated as a binary success or failure problem, usually defined by whether the object simply stays in the hand. For forceful tool use, however, this view is insufficient because grasp compliance becomes a critical factor governing how the hand and tool behave under load. Compliance arises from coupled kinematics, grasp configuration, passive mechanics, and contact conditions, producing nonlinear behavior in which deformation and interaction forces influence each other. Understanding this relationship is essential for predictive models of how a grasped tool and a compliant hand jointly respond to external loading. In underactuated hands, these effects are amplified: such designs offer low cost and adaptive grasping, but make compliance behavior more difficult to model and predict. Our goal is therefore to develop a predictive model for grasped tool behavior during forceful interactions. To address this challenge, we introduce an analytical model informed neural network (AMINN), a hybrid predictive model that combines an analytical mechanics layer with data driven learning to estimate grasp stability and in hand tool displacement under external loading. The model is evaluated on a three finger underactuated robotic hand and shows strong predictive capability with mechanically meaningful outputs across diverse loading conditions. Compared with a black box multilayer perceptron baseline, AMINN also achieves better energy based physical consistency. Beyond prediction accuracy alone, this framework advances physically interpretable learning for robotic manipulation and supports more reliable, safer, and more trustworthy autonomous tool use in safety critical settings during forceful interaction.

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

46 extracted references · 2 canonical work pages

  1. [1]

    Mason, M. T. and Salisbury, J. K., 1985,Robot Hands and the Mechanics of Manipulation, MIT Press, Cambridge, MA

    1985

  2. [2]

    AnalysisofMultifingeredHands,

    Kerr,J.andRoth,B.,1986,“AnalysisofMultifingeredHands,”TheInternational Journal of Robotics Research,4(4), pp. 3–17

    1986

  3. [3]

    On the Modeling of Grasps with a Multi- Fingered Hand,

    Brodsky, V. and Shoham, M., 1995, “On the Modeling of Grasps with a Multi- Fingered Hand,”Computational Kinematics, J.-P. Merlet and B. Ravani, eds., Springer Netherlands, Dordrecht

    1995

  4. [4]

    Constructing force-closure grasps,

    Nguyen, V.-D., 1986, “Constructing force-closure grasps,”Proceedings. 1986 IEEE International Conference on Robotics and Automation, Vol. 3, pp. 1368– 1373, doi: 10.1109/ROBOT.1986.1087483

    arXiv 1986

  5. [5]

    Planning optimal grasps,

    Ferrari, C. and Canny, J., 1992, “Planning optimal grasps,”Proceedings 1992 IEEE International Conference on Robotics and Automation, pp. 2290–2295 vol.3, doi: 10.1109/ROBOT.1992.219918

    arXiv 1992

  6. [6]

    Grasp Quality Measures: Review and Per- formance,

    Roa, M. A. and Suarez, R., 2014, “Grasp Quality Measures: Review and Per- formance,” Autonomous Robots,38, pp. 65–88

    2014

  7. [7]

    Deep Learning for Detecting Robotic Grasps,

    Lenz, I., Lee, H., and Saxena, A., 2015, “Deep Learning for Detecting Robotic Grasps,” The International Journal of Robotics Research,34(4-5), pp. 705–724

    2015

  8. [8]

    Supersizing self-supervision: Learn- ing to grasp from 50K tries and 700 robot hours,

    Pinto, L. and Gupta, A., 2016, “Supersizing self-supervision: Learn- ing to grasp from 50K tries and 700 robot hours,”2016 IEEE Inter- national Conference on Robotics and Automation (ICRA), pp. 3406–3413, doi: 10.1109/ICRA.2016.7487517

    arXiv 2016

  9. [9]

    Learning hand-eye coordination for robotic grasping with deep learning and large-scale data collection,

    Levine, S., Pastor, P., Krizhevsky, A., Ibarz, J., andQuillen, D., 2018, “Learning hand-eye coordination for robotic grasping with deep learning and large-scale data collection,” The International journal of robotics research,37(4-5), pp. 421–436

    2018

  10. [10]

    Large-scale Supervised Learning of the Grasp Robustness of Surface Patch Pairs,

    Seita, D., Pokorny, F. T., Mahler, J., Kragic, D., Franklin, M. J., Canny, J. F., and Goldberg, K. Y., 2016, “Large-scale Supervised Learning of the Grasp Robustness of Surface Patch Pairs,”2016 IEEE International Conference on Simulation, Modeling, and Programming for Autonomous Robots (SIMPAR), pp. 216–223, doi: 10.1109/SIMPAR.2016.7862399

    arXiv 2016

  11. [11]

    Dex-Net 2.0: Deep Learning to Plan Robust Grasps With Synthetic Point Clouds and Analytic Grasp Metrics,

    Mahler, J., Liang, J., Niyaz, S., Laskey, M., Doan, R., Liu, X., Aparicio Ojea, J., and Goldberg, K., 2017, “Dex-Net 2.0: Deep Learning to Plan Robust Grasps With Synthetic Point Clouds and Analytic Grasp Metrics,”Robotics: Science and Systems, doi: 10.15607/RSS.2017.XIII.058

    work page doi:10.15607/rss.2017.xiii.058 2017

  12. [12]

    QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipula- tion,

    Kalashnikov, D., Irpan, A., Pastor, P., Ibarz, J., Herzog, A., Jang, E., Quillen, D., Holly, E., Kalakrishnan, M., Vanhoucke, V., and Levine, S., 2018, “QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipula- tion,” CoRR,abs/1806.10293

    Pith/arXiv arXiv 2018

  13. [13]

    Tactile Grasp Stability Classification Based on Graph Convolutional Networks,

    Mi, T., Que, D., Fang, S., Zhou, Z., Ye, C., Liu, C., Yi, Z., and Wu, X., 2021, “Tactile Grasp Stability Classification Based on Graph Convolutional Networks,”2021 IEEE International Conference on Real-time Computing and Robotics (RCAR), pp. 875–880, doi: 10.1109/RCAR52367.2021.9517085

    arXiv 2021

  14. [14]

    A comparison of stiffness analysis methods for robotic systems,

    Goncalves, R. S., Carbone, G., Carvalho, J. C. M., and Ceccarelli, M., 2016, “A comparison of stiffness analysis methods for robotic systems,” International journal of mechanics and control,17(2), pp. 35–58

    2016

  15. [15]

    Stiffness mapping for parallel manipulators,

    Gosselin, C., 1990, “Stiffness mapping for parallel manipulators,” IEEE Trans- actions on Robotics and Automation,6(3), pp. 377–382

    1990

  16. [16]

    Jacobian, Manipulability, ConditionNumber, andAccuracy of Parallel Robots,

    Merlet, J.-P., 2006, “Jacobian, Manipulability, ConditionNumber, andAccuracy of Parallel Robots,” Journal of Mechanical Design,128(1), pp. 199–206

    2006

  17. [17]

    Design Considerations of New Six Degrees-of-Freedom Parallel Robots,

    Sima’an, N., Glozman, D., and Shoham, M., 1998, “Design Considerations of New Six Degrees-of-Freedom Parallel Robots,”Proceedings of the 1998 IEEE International Conference on Robotics and Automation, pp. 1327–1333

    1998

  18. [18]

    Robot Analysis: The Mechanics of Serial and Parallel Manipulators,

    Tsai, L. W., 1999, “Robot Analysis: The Mechanics of Serial and Parallel Manipulators,” https://api.semanticscholar.org/CorpusID:59698769

    1999

  19. [19]

    Fundamentals of ma- nipulator stiffness modeling using matrix structural analysis,

    Klimchik, A., Pashkevich, A., and Chablat, D., 2019, “Fundamentals of ma- nipulator stiffness modeling using matrix structural analysis,” Mechanism and Machine Theory,133, pp. 365–394

    2019

  20. [20]

    Rigid- ity analysis of T3R1 parallel robot with uncoupled kinematics,

    Bouzgarrou, B. C., Fauroux, J.-C., Gogu, G., and Heerah, Y., 2004, “Rigid- ity analysis of T3R1 parallel robot with uncoupled kinematics,” https://api. semanticscholar.org/CorpusID:211575117

    2004

  21. [21]

    A method for analyzing parallel mechanism stiffness including elastic deformations in the structure,

    Yoon, W.-K., Suehiro, T., Tsumaki, Y., and Uchiyama, M., 2002, “A method for analyzing parallel mechanism stiffness including elastic deformations in the structure,”IEEE/RSJ International Conference on Intelligent Robots and Sys- tems, Vol. 3, pp. 2875–2880 vol.3, doi: 10.1109/IRDS.2002.1041707

    Pith/arXiv arXiv 2002

  22. [22]

    Computing and controlling compliance of a robotic hand,

    Cutkosky, M. and Kao, I., 1989, “Computing and controlling compliance of a robotic hand,” IEEE Transactions on Robotics and Automation,5(2), pp. 151– 165

    1989

  23. [23]

    Conservative Congruence Transformation for Joint and Cartesian Stiffness Matrices of Robotic Hands and Fingers,

    Chen, S.-F. and Kao, I., 2000, “Conservative Congruence Transformation for Joint and Cartesian Stiffness Matrices of Robotic Hands and Fingers,” The International Journal of Robotics Research,19(9), pp. 835–847

    2000

  24. [24]

    Enhanced Stiffness Mod- eling of Manipulators with Passive Joints,

    Pashkevich, A., Klimchik, A., and Chablat, D., 2011, “Enhanced Stiffness Mod- eling of Manipulators with Passive Joints,” Mechanism and Machine Theory, 46(5), pp. 662–679, Key used here as shorthand for passive-joint projection methods; see also the 2012 T-RO ‘Computational Aspects‘ follow-up

    2011

  25. [25]

    Redundancy in Parallel Mechanisms: A Review,

    Schreiber, L.-T. and Gosselin, C., 2018, “Redundancy in Parallel Mechanisms: A Review,” Applied Mechanics Reviews,70(1), p. 010802

    2018

  26. [26]

    Redundant Actuation of Parallel Manipulators,

    Müller, A., 2008, “Redundant Actuation of Parallel Manipulators,”Parallel Manipulators, Towards New Applications, H. Wu, ed., IntechOpen

    2008

  27. [27]

    Stiffness Synthesis of a Variable Geom- etry Six-Degrees-of-Freedom Double Planar Parallel Robot,

    Simaan, N. and Shoham, M., 2003, “Stiffness Synthesis of a Variable Geom- etry Six-Degrees-of-Freedom Double Planar Parallel Robot,” The International Journal of Robotics Research,22(9), pp. 757–775

    2003

  28. [28]

    Analysis of stiffness controllability of a redundant cable-driven parallel robot based on its configuration,

    Cui, Z. and Tang, X., 2021, “Analysis of stiffness controllability of a redundant cable-driven parallel robot based on its configuration,” Mechatronics,75, p. 102519

    2021

  29. [29]

    Static and dynamic stiffness analyses of cable-driven parallel robots with non-negligible cable mass and elasticity,

    Yuan, H., Courteille, E., and Deblaise, D., 2015, “Static and dynamic stiffness analyses of cable-driven parallel robots with non-negligible cable mass and elasticity,” Mechanism and Machine Theory,85, pp. 64–81

    2015

  30. [30]

    Directional Stiffness Modulation of Par- allel Robots With Kinematic Redundancy and Variable Stiffness Joints,

    Orekhov, A. L. and Simaan, N., 2019, “Directional Stiffness Modulation of Par- allel Robots With Kinematic Redundancy and Variable Stiffness Joints,” Journal of Mechanisms and Robotics,11(5), p. 051003

    2019

  31. [31]

    Simulation and Design of Underactu- ated Mechanical Hands,

    Laliberté, T. and Gosselin, C. M., 1998, “Simulation and Design of Underactu- ated Mechanical Hands,” Mechanism and Machine Theory,33(1-2), pp. 39–57

    1998

  32. [32]

    Joint Coupling Design of Underactuated Grip- pers,

    Dollar, A. and Howe, R., 2006, “Joint Coupling Design of Underactuated Grip- pers,” . 8

    2006

  33. [33]

    ‘Underactuation in robotic graspinghands’,

    Laliberté, T., Birglen, L., and Gosselin, C., 2002, “‘Underactuation in robotic graspinghands’,”JapaneseJournalofMachineIntelligenceandRoboticControl, 4, pp. 77–87

    2002

  34. [34]

    Birglen,L.,Laliberté,T.,andGosselin,C.,2008,UnderactuatedRoboticHands, Springer Tracts in Advanced Robotics, Springer, Berlin, Heidelberg, Key retains 2007 to match your citation label

    2008

  35. [35]

    Robotic Grasping and Manipulation of Unstructured Ob- jects: A Review of Underactuated Hand Design,

    Dollar, A. M., 2011, “Robotic Grasping and Manipulation of Unstructured Ob- jects: A Review of Underactuated Hand Design,” The International Journal of Robotics Research,30(9), pp. 1147–1169

    2011

  36. [36]

    A Design and Analysis Tool for Under- actuated Compliant Hands,

    Ciocarlie, M. T. and Allen, P. K., 2009, “A Design and Analysis Tool for Under- actuated Compliant Hands,”2009 IEEE/RSJ International Conference on Intel- ligent Robots and Systems, pp. 5234–5239, doi: 10.1109/IROS.2009.5354089

    arXiv 2009

  37. [37]

    The Ability of Underactuated Hands to GraspandHoldObjects,

    Kragten, G. A. and Herder, J. L., 2010, “The Ability of Underactuated Hands to GraspandHoldObjects,” MechanismandMachineTheory,45(3), pp.408–425

    2010

  38. [38]

    Stable, Open-Loop Precision Ma- nipulation with Underactuated Hands,

    Odhner, L. U. and Dollar, A. M., 2015, “Stable, Open-Loop Precision Ma- nipulation with Underactuated Hands,” The International Journal of Robotics Research,34(11), pp. 1347–1360

    2015

  39. [39]

    Evaluation of grasp stiffness in un- deractuated compliant hands,

    Malvezzi, M. and Prattichizzo, D., 2013, “Evaluation of grasp stiffness in un- deractuated compliant hands,”2013 IEEE International Conference on Robotics and Automation, pp. 2074–2079, doi: 10.1109/ICRA.2013.6630855

    arXiv 2013

  40. [40]

    Underactuation Design for Tendon-Driven Hands via Optimization of Mechanically Realizable ManifoldsinPostureandTorqueSpaces,

    Chen, T., Wang, L., Haas-Heger, M., and Ciocarlie, M., 2020, “Underactuation Design for Tendon-Driven Hands via Optimization of Mechanically Realizable ManifoldsinPostureandTorqueSpaces,”IEEETransactionsonRobotics,36(3), pp. 708–723

    2020

  41. [41]

    A Highly-Underactuated Robotic Hand With Force and Joint Angle Sensors,

    Wang, L., DelPreto, J., Bhattacharyya, S., Weisz, J., and Allen, P. K., 2011, “A Highly-Underactuated Robotic Hand With Force and Joint Angle Sensors,” 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 1380–1385, doi: 10.1109/IROS.2011.6048748, Used here as the underactuated sensing reference; the user-provided IEEE document id...

    arXiv 2011

  42. [42]

    Learning-based object’s stiffness and shape estimation with confidence level in multi-fingered hand grasping,

    Kutsuzawa, K., Matsumoto, M., Owaki, D., and Hayashibe, M., 2024, “Learning-based object’s stiffness and shape estimation with confidence level in multi-fingered hand grasping,” Frontiers in Neurorobotics,Volume 18 - 2024

    2024

  43. [43]

    A Data- Driven Framework for Learning Dexterous Manipulation of Unknown Objects,

    Morgan, A. S., Hang, K., Bircher, W. G., and Dollar, A. M., 2019, “A Data- Driven Framework for Learning Dexterous Manipulation of Unknown Objects,” 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 8273–8280, doi: 10.1109/IROS40897.2019.8968458

    arXiv 2019

  44. [44]

    PINN-Ray: A Physics-Informed Neural Net- work to Model Soft Robotic Fin Ray Fingers,

    Wang, X., Dabrowski, J. J., Pinskier, J., Liow, L., Viswanathan, V., Scalzo, R., and Howard, D., 2024, “PINN-Ray: A Physics-Informed Neural Net- work to Model Soft Robotic Fin Ray Fingers,”2024 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS), pp. 247–254, doi: 10.1109/IROS58592.2024.10802217

    arXiv 2024

  45. [45]

    Contact-Invariant Optimiza- tion for Hand Manipulation,

    Mordatch, I., Popović, Z., and Todorov, E., 2012, “Contact-Invariant Optimiza- tion for Hand Manipulation,”Eurographics/ACM SIGGRAPH Symposium on Computer Animation, pp. 137–144, doi: 10.2312/SCA/SCA12/137-144

    work page doi:10.2312/sca/sca12/137-144 2012

  46. [46]

    A Continuous Tele- operation Subspace With Empirical and Algorithmic Mapping Algorithms for Nonanthropomorphic Hands,

    Meeker, C., Haas-Heger, M., and Ciocarlie, M., 2022, “A Continuous Tele- operation Subspace With Empirical and Algorithmic Mapping Algorithms for Nonanthropomorphic Hands,” IEEE Transactions on Automation Science and Engineering,19(1), pp. 373–386. 9

    2022