Pith. sign in

REVIEW 3 major objections 6 minor 36 references

A Delay-free Control Method Based On Function Approximation And Broadcast For Robotic Surface And Multiactuator Systems

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a robotic surface's update delay can be made independent of the number of actuators by broadcasting compressed shape coefficients and letting each module compute its own control input.

desk verdict Genuine architectural idea for broadcast-based shape control, but the delay-free scaling claim is only proven for shapes with bounded complexity, not the general case. read the letter →

arxiv 2412.00492 v1 pith:6SLHU42B submitted 2024-11-30 cs.RO

Yuchen Zhao This is my paper · ORCID
classification cs.RO
keywords CellularandModularRobotsDistributedRobotSystemsShapeDisplayBroadcastControlFunctionApproximationMatchingPursuitDiscreteCosineTransformHapticsHapticInterfaces
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

Robotic surfaces control many actuators in parallel, but standard sequential control sends each actuator its own message, so the refresh delay grows linearly with the number of actuators. This paper proposes to invert that arrangement: the central computer approximates the target shape by a small set of coefficients (using a cosine transform or matching pursuit), broadcasts those coefficients to all modules at once, and each module evaluates the coefficient set together with its stored identifier to compute its own control input. The claimed consequence is a system-size-independent time delay, τ ∝ $N^{0}$, for compressible shapes. The authors support this with a 4×4 pin array, measuring constant delays as N grows while sequential control shows linear growth, and demonstrate that their method reproduces six shapes with fewer messages than the sequential baseline. The motivation is that communication, not actuation, is the usual bottleneck as robotic surfaces scale up.

What carries the argument

The central machinery is the pair (γt, f(xn, γt)): a compact coefficient message broadcast to all modules at once, together with a function form that each module evaluates locally using its stored identifier xn. The paper implements this with three function families: the discrete cosine transform fn = a0 + 2Σ at cos(kt(2xn + 1)) for extended smooth shapes; matching pursuit over a periodic Gaussian time-frequency dictionary fn = Σ at g_{st}(xn − pt) cos(2π kt xn / N + φt) for localized or mixed patterns, with gs a periodized Gaussian; and a one-term Gaussian radial basis function for object manipulation, where the coefficient message is just the amplitude, width, and two center coordinates. The load-bearing property is universal approximation: any of these families can represent the target shape to a given accuracy with a small number of coefficients when the shape is compressible, and that coefficient count—not the actuator count—determines the number of broadcast messages.

What would settle it

Drive a robotic pin array with the broadcast method using a target height field where each actuator's height is drawn independently from a uniform distribution, fix a per-actuator root-mean-squared error tolerance, and measure both the number of broadcast terms and the refresh delay as N is increased. If the term count must grow with N to hold the tolerance, then the delay will grow with N as well, which directly contradicts the paper's α = 0 scaling for that shape class.

Watch

Extended reading notes

Core claim

The central discovery is that robotic-surface control can be recast as a broadcast-plus-local-evaluation problem. Instead of shipping each actuator a setpoint, the central computer computes a compressed representation of the target shape—the amplitudes and wave vectors of a DCT, or the parameters of matching-pursuit time-frequency atoms—and broadcasts this coefficient vector γt to every module. Each module, knowing its own identifier xn, plugs the received coefficients into the common function form f(xn, γt) and computes its own reference input on-site. Since all modules receive the same message at the same time and compute in parallel, the communication delay no longer depends on the number of actuators; the paper states this as τ ∝ N^α with α = 0. Experiments on a 16-actuator pin array confirm the constant-delay scaling both for pure communication and with actuator dynamics present, show that the broadcast method needs fewer messages than sequential control to reach a given shape accuracy, and demonstrate dynamic manipulation of an 80 mm sphere using a single broadcast Gaussian coefficient set.

Load-bearing premise

The constant-delay claim holds only for target shapes that can be approximated to the required accuracy by a bounded number of broadcast coefficients that does not grow with the number of actuators; for spatially random or incompressible shapes the coefficient count must grow with N, and the delay advantage collapses.

Editorial extensions

If this is right

  • With compressible shapes, the refresh rate of a robotic surface stays fixed as the array grows, so real-time shape display and haptic feedback can be maintained at much larger scales than sequential control allows.
  • Only one broadcast frame per approximation term is needed, so the communication bandwidth requirement is set by shape complexity rather than actuator count, making single-bus hardware such as CAN sufficient for large arrays.
  • Because each actuator computes its own input, the method is naturally distributed and parallel; adding a module does not add to the message sequence, only to the set of local identifiers.
  • Dynamic tasks like object manipulation can be commanded by broadcasting time-varying coefficients (center position, width, amplitude) at a constant 60 Hz update rate independent of N, as demonstrated with the 80 mm sphere.

Reading between the lines

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

  • The method effectively trades communication bandwidth for local computation, an exchange that becomes more favorable as microcontroller cost per actuator drops; the same broadcast-and-local-decode pattern could be applied to soft actuators, swarm robots, and large pixel-dense displays.
  • The critical implicit condition is shape compressibility: the paper selects six shapes that are smooth or structured, and extrapolating to large N requires a compressibility measure relating target-shape complexity to the number of coefficients.
  • A natural extension is closed-loop operation: the broadcast method is open-loop from the shape-error perspective, but a low-rate feedback channel that broadcasts only update coefficients could correct drift without reintroducing O(N) communication, if the correction signal is itself compressible.
  • The matching-pursuit results hint at a data-dependent trade-off: DCT is better for extended patterns, MP for localized ones, so a hybrid or learned selection of basis could reduce message counts further; the paper does not explore this.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes a broadcast-based control method for robotic surface pin arrays. Instead of sending per-actuator reference signals sequentially, a central computer computes function-approximation coefficients (via discrete cosine transform, matching pursuit, or Gaussian radial basis functions) and broadcasts them to all actuation modules, which evaluate their own control inputs locally using stored identification coordinates. The paper claims a system-size-independent time delay, τ ∝ N^0, builds a 4×4 hardware prototype, and measures the time delay with and without actuator dynamics. It also characterizes shape-approximation accuracy on six shapes and demonstrates object manipulation. The authors acknowledge in the Discussion that the time-delay scaling was validated only on a small system and that the control is open-loop.

Significance. If the central claim holds for a meaningful class of target shapes, the method is a useful conceptual advance: it decouples the differential communication delay from the number of actuators and replaces per-actuator command streams with a compact coefficient broadcast. The paper includes a credible hardware implementation and clean experiments. The traveling-wave experiment confirms the predicted delay τ = T/4 independently of N, and the sequential-control prediction τ = T/4 + Tmsg(N−1) matches the data. The shape-error study provides an empirical comparison of DCT and MP against sequential control on six representative shapes, and the authors are explicit about the open-loop nature and the small validation scale. The main risk is that the abstract's unconditional 'constant time delays no matter how many actuators' is broader than the evidence, which needs a precise scope condition.

major comments (3)
  1. [§2, Eq. (2) and §4.1] The definition of τ in the Introduction ('the time between when a control message is sent to the first actuator and the time when the last actuator reaches its target position') is ambiguous, and the experiments in §4.1 actually measure the relative delay between the first and last module via cross-correlation. Under this differential-delay interpretation, the broadcast method yields τ = 0 (or the wave-propagation delay T/4) independent of N, which is true regardless of the number M of broadcast terms. However, the absolute time from the start of transmission to the completion of a shape is at least M·Tmsg plus settling time; since Eq. (1) requires M = N−1 terms for exact representation of a generic profile, the 'delay-free' property in the abstract is only guaranteed for the differential delay or for shape families with M bounded independently of N. Please state this distinction explicitly and either narrow the abstract's claim or introduce a formal compressibility assumption (e.g., bandlimited or sparse in the chosen dictionary) under which Eq. (2) is derived.
  2. [§4.2, Fig. 6] The curves labeled 'theoretical predictions' are computed by applying the same DCT or MP formulas to the same target shapes; for DCT they are exactly the mathematical truncation error of Eq. (1), and for MP they are the residuals of the fitted atoms. Agreement between these curves and the experimental points therefore verifies that the modules reproduce the computed references, but it does not independently validate the approximation accuracy. The only non-tautological model is the limited-resolution MP curve (dotted lines), which explains residual errors. Please re-label these curves as 'computed approximation error' or 'model residual' and adjust the claim of 'good agreement between experiments and theoretical predictions' accordingly.
  3. [§4.1, Fig. 5 and §4.2] The claim that the method 'requires less control messages to generate shapes with the same accuracy' is demonstrated on six 4×4 shapes, but the scaling of M with N is not addressed. For the two time-delay experiments in §4.1, the uniform pattern and the traveling wave are representable with M = 1 or 2 messages independent of N, so the experiments never exercise a regime where the number of broadcast frames grows with the system size. Because the advantage over sequential control is precisely the ratio N/M, the paper should either prove a bound on M for a defined class of shapes (e.g., bandlimited or localized patterns) or present experiments or simulations with a shape family whose complexity grows with N, such as random patterns or increasing spatial frequency.
minor comments (6)
  1. [Fig. 4 caption] The word 'boardcast' is a typo and should be 'broadcast'.
  2. [Fig. 1(b) and text near it] The figure plotting the scaling exponent α versus the number of actuators is not explained: the text says α in existing works are summarized, but it does not define the sources or how α is estimated for each work, and the star for the present work is unclear.
  3. [Eq. (1)] The DCT formula as written has a nonstandard normalization (the factor 2 and the wave vector k_t = πt/(2N)); please specify the exact DCT variant used in scipy.fftpack and the domain of x_n, so that the factor does not introduce ambiguity.
  4. [Table 1] The shape named 'random random uniform distribution' is not a well-defined target; please provide the actual 4×4 matrix or the random seed used in the experiments.
  5. [§4.3] The object-manipulation demonstration would benefit from a quantitative measure of trajectory tracking error or at least the repeatability over multiple cycles, rather than only snapshots.
  6. [§4.2, Fig. 6] The statement that the errorbar is smaller than the marker size is not accompanied by numerical standard deviations; please report the worst-case standard deviation or a table of the averaged errors.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: broadcast architecture makes delay scaling an architectural consequence, and experiments validate the implementation; shape-error 'predictions' are consistency checks, not fitted outcomes.

full rationale

The derivation chain is self-contained and non-circular. The central scaling claim τ ∝ N^0 (Eq. 2) is a direct consequence of the broadcast architecture: all modules receive the same γt message and compute fn in parallel, as stated in Sec. 2 ('all modules simultaneously receive γt and compute fn in parallel. So we have: τ ∝ N^α, α=0'). This is a design property, not a parameter fitted to data; the experiments in Sec. 4.1 validate that the CAN broadcast and local computation actually behave this way, and the traveling-wave prediction τ = T/4 follows independently from the commanded phase difference π/2 and speed v = 2π/T, not from the broadcast coefficients. The shape-error 'theoretical predictions' in Sec. 4.2 are recalculations of the same DCT/MP coefficients used to drive the hardware, so matching experiments is an implementation consistency check rather than an independent prediction; no experimental error data is used to choose the coefficients. References [34] and [36] supply external mathematical support (MP convergence and RBF universal approximation) and are not self-citations. The only caveat is a scope condition, acknowledged in Sec. 5 ('the time delay scaling is only validated on a small set of actuation modules'): Eq. (1) requires N−1 terms for an arbitrary N-actuator pattern, so the α=0 claim presumes the number of broadcast terms remains bounded as N grows. This is a generalizability limitation, not a circular step.

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

The method introduces no new physical entities or fitted constants; the approximation coefficients are the control signal, not hidden parameters. The key unstated premise is the bounded-complexity assumption on the shape class, without which the central scaling claim collapses.

assumptions (4)
  • domain assumption Target shapes have bounded complexity: they can be represented by a constant number of approximation terms as N grows.
    Introduced in Section 1 ('neighboring actuators usually have similar control inputs, which could be approximated') and Section 2; if false, the number of broadcast messages grows with N and the delay-free property fails.
  • domain assumption All actuation modules receive the broadcast message simultaneously with negligible skew.
    Assumed in Section 2 and the CAN bus implementation; any bus arbitration or repeater delay would introduce a size-dependent component.
  • domain assumption Actuator dynamics are identical across modules, so relative delays are purely kinematic.
    Used in Section 4.1 to predict tau = T/4 for the traveling wave; PID controllers run at the same rate but motor variations are not characterized.
  • standard math Matching pursuit converges (Mallat-Zhang 1993) and DCT/RBF have universal approximation properties.
    Invoked in Section 2 to justify representing arbitrary shapes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Delay-free Control Method Based On Function Approximation And Broadcast For Robotic Surface And Multiactuator Systems." pith.science (2026). https://pith.science/paper/6SLHU42B

@misc{pith2026241200492,
  author       = {Pith},
  title        = {Pith review of: A Delay-free Control Method Based On Function Approximation And Broadcast For Robotic Surface And Multiactuator Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6SLHU42B}},
  note         = {Machine review of arXiv:2412.00492}
}
read the original abstract

Robotic surface consisting of many actuators can change shape to perform tasks, such as facilitating human-machine interactions and transporting objects. Increasing the number of actuators can enhance the robot's capacity, but controlling them requires communication bandwidth to increase equally in order to avoid time delays. We propose a novel control method that has constant time delays no matter how many actuators are in the robot. Having a distributed nature, the method first approximates target shapes, then broadcasts the approximation coefficients to the actuators, and relies on themselves to compute the inputs. We build a robotic pin array and measure the time delay as a function of the number of actuators to confirm the system size-independent scaling behavior. The shape-changing ability is achieved based on function approximation algorithms, i.e. discrete cosine transform or matching pursuit. We perform experiments to approximate target shapes and make quantitative comparison with those obtained from standard sequential control method. A good agreement between the experiments and theoretical predictions is achieved, and our method is more efficient in the sense that it requires less control messages to generate shapes with the same accuracy. Our method is also capable of dynamic tasks such as object manipulation.

Figures

Figures reproduced from arXiv: 2412.00492 by the authors.

Figure 1
Figure 1. (a) The 4×4 pin array; (b) time delay scaling exponent α is plotted vs. the number of actuators. Our work is highlighted in a yellow star. This paper proposes a new control method for robotic surface that has system size-independent time delay. The central computer broadcasts features of the target shape to individual actuation modules and has them calculate their inputs on-site. The motivations are: (1) neighboring… view at source ↗
Figure 2
Figure 2. An illustration of our control method. At time [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. (a) A picture of a single linear actuation module. The rectangular cover is removed to expose [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Experimental time delay scaling without actuator dynamics. (a) The control inputs of two [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Experimental time delay scaling with actuator dynamics. (a) The motor shaft angular position [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Characterization of shape-changing ability. (a) The experimental setup for shape measurement; [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Illustration of object manipulation capability. Figure (a) to (h) are snapshots of a blue sphere [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages

  1. [1]

    Continuum robot surfaces: Smart saddles and seats,

    I. D. Walker, “Continuum robot surfaces: Smart saddles and seats,” in Mechatronics and Robotics Engineering for Advanced and Intelligent Manufacturing. Springer, Cham, 2017, pp. 97–105

    work page 2017

  2. [2]

    Physical telepresence: shape capture and display for embodied, computer-mediated remote collaboration,

    D. Leithinger, S. Follmer, A. Olwal, and H. Ishii, “Physical telepresence: shape capture and display for embodied, computer-mediated remote collaboration,” in Proceedings of the 27th annual ACM symposium on User interface software and technology, ser. UIST ’14. New York, NY, USA: Association for Computing Machinery, 2014, pp. 461–470

    work page 2014

  3. [3]

    A Novel Refreshable Braille Display Based on the Layered Electromagnetic Driving Mechanism of Braille Dots,

    H. Chen, W. Tao, C. Liu, Q. Shen, Y. Wu, L. Ruan, and W. Yang, “A Novel Refreshable Braille Display Based on the Layered Electromagnetic Driving Mechanism of Braille Dots,” IEEE Transactions on Haptics, vol. 16, no. 1, pp. 96–105, 2023

    work page 2023

  4. [4]

    Materiable: Rendering Dynamic Material Properties in Response to Direct Physical Touch with Shape Changing Interfaces,

    K. Nakagaki, L. Vink, J. Counts, D. Windham, D. Leithinger, S. Follmer, and H. Ishii, “Materiable: Rendering Dynamic Material Properties in Response to Direct Physical Touch with Shape Changing Interfaces,” in Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, ser. CHI ’16. New York, NY, USA: Association for Computing Machinery,...

    work page 2016

  5. [5]

    Visuo-Haptic Illusions for Improving the Perceived Performance of Shape Displays,

    P. Abtahi and S. Follmer, “Visuo-Haptic Illusions for Improving the Perceived Performance of Shape Displays,” in Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, ser. CHI ’18. New York, NY, USA: Association for Computing Machinery, 2018, pp. 1–13

    work page 2018

  6. [6]

    inFORCE: Bi-directional ‘Force’ Shape Display for Haptic Interaction,

    K. Nakagaki, D. Fitzgerald, Z. J. Ma, L. Vink, D. Levine, and H. Ishii, “inFORCE: Bi-directional ‘Force’ Shape Display for Haptic Interaction,” in Proceedings of the Thirteenth International Con- ference on Tangible, Embedded, and Embodied Interaction, ser. TEI ’19. New York, NY, USA: Association for Computing Machinery, 2019, pp. 615–623

    work page 2019

  7. [7]

    Control strategies for small- scaled conveyor modules enabling highly flexible material flow systems,

    C. Uriarte, A. Asphandiar, H. Thamer, A. Benggolo, and M. Freitag, “Control strategies for small- scaled conveyor modules enabling highly flexible material flow systems,” Procedia CIRP, vol. 79, pp. 433–438, 2019

    work page 2019

  8. [8]

    Trajectory Planning and Tracking of Multiple Objects on a Soft Robotic Table Using a Hierarchical Search on Time-Varying Potential Fields,

    Z. Chen, Z. Deng, J. S. Dhupia, M. Stommel, and W. Xu, “Trajectory Planning and Tracking of Multiple Objects on a Soft Robotic Table Using a Hierarchical Search on Time-Varying Potential Fields,” IEEE Transactions on Robotics, vol. 40, pp. 351–363, 2024

    work page 2024

Show all 36 references
  1. [9]

    Design and Characterization of a Novel Robotic Surface for Application to Compressed Physical Environments *,

    Y. Wang, C. Frazelle, R. Sirohi, L. Li, I. D. Walker, and K. E. Green, “Design and Characterization of a Novel Robotic Surface for Application to Compressed Physical Environments *,” in2019 International Conference on Robotics and Automation (ICRA), 2019, pp. 102–108, iSSN: 2577-087X

    2019

  2. [10]

    Ori-Pixel, a Multi-DoFs Origami Pixel for Modular Reconfig- urable Surfaces,

    M. Salerno, J. Paik, and S. Mintchev, “Ori-Pixel, a Multi-DoFs Origami Pixel for Modular Reconfig- urable Surfaces,” IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 6988–6995, 2020, conference Name: IEEE Robotics and Automation Letters

    2020

  3. [11]

    Soft Robotic Mannequin: Design and Algorithm for Deformation Control,

    Y. Tian, G. Fang, J. S. Petrulis, A. Weightman, and C. C. L. Wang, “Soft Robotic Mannequin: Design and Algorithm for Deformation Control,” IEEE/ASME Transactions on Mechatronics, pp. 1–10, 2022, conference Name: IEEE/ASME Transactions on Mechatronics

    2022

  4. [12]

    Adaptive mould d100,

    A. A. Moulds, “Adaptive mould d100,” 2024. [Online]. Available: https://adapamoulds.com/portfolio- item/adaptive-mould-d100/ 12

    2024

  5. [13]

    Floor system providing omnidirec- tional movement of a person walking in a virtual reality environment,

    L. S. Smoot, G. D. Niemeyer, D. L. Christensen, and R. Bristow, “Floor system providing omnidirec- tional movement of a person walking in a virtual reality environment,” US Patent US10 416 754B2, Sept., 2019

    2019

  6. [14]

    A multifunctional soft robotic shape display with high-speed actuation, sensing, and control,

    B. K. Johnson, M. Naris, V. Sundaram, A. Volchko, K. Ly, S. K. Mitchell, E. Acome, N. Kellaris, C. Keplinger, N. Correll, J. S. Humbert, and M. E. Rentschler, “A multifunctional soft robotic shape display with high-speed actuation, sensing, and control,” Nature Communications,...

    2023

  7. [15]

    A Compact Modular Soft Surface With Reconfigurable Shape and Stiffness,

    M. A. Robertson, M. Murakami, W. Felt, and J. Paik, “A Compact Modular Soft Surface With Reconfigurable Shape and Stiffness,” IEEE/ASME Transactions on Mechatronics, vol. 24, no. 1, pp. 16–24, 2019, conference Name: IEEE/ASME Transactions on Mechatronics

    2019

  8. [16]

    Robotic surfaces with reversible, spatiotemporal control for shape morphing and object manipulation,

    K. Liu, F. Hacker, and C. Daraio, “Robotic surfaces with reversible, spatiotemporal control for shape morphing and object manipulation,” Science Robotics, vol. 6, no. 53, p. eabf5116, 2021, eprint: https://www.science.org/doi/pdf/10.1126/scirobotics.abf5116

    2021 doi

  9. [17]

    Dimension reduction in a feedback loop using the SVD: Results on controllability and stability,

    R. C. Winck and W. J. Book, “Dimension reduction in a feedback loop using the SVD: Results on controllability and stability,” Automatica, vol. 49, no. 10, pp. 3084–3089, 2013

    2013

  10. [18]

    inFORM: dynamic physical affordances and constraints through shape and object actuation,

    S. Follmer, D. Leithinger, A. Olwal, A. Hogge, and H. Ishii, “inFORM: dynamic physical affordances and constraints through shape and object actuation,” in Proceedings of the 26th annual ACM sympo- sium on User interface software and technology, ser. UIST ’13. New York, NY, USA...

    2013

  11. [19]

    ArrayBot: Reinforcement Learning for Generalizable Distributed Manipulation through Touch,

    Z. Xue, H. Zhang, J. Cheng, Z. He, Y. Ju, C. Lin, G. Zhang, and H. Xu, “ArrayBot: Reinforcement Learning for Generalizable Distributed Manipulation through Touch,” June 2023, arXiv:2306.16857 [cs]

    2023 arXiv

  12. [20]

    shapeShift: 2D Spatial Ma- nipulation and Self-Actuation of Tabletop Shape Displays for Tangible and Haptic Interaction,

    A. F. Siu, E. J. Gonzalez, S. Yuan, J. B. Ginsberg, and S. Follmer, “shapeShift: 2D Spatial Ma- nipulation and Self-Actuation of Tabletop Shape Displays for Tangible and Haptic Interaction,” in Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, ser. ...

    2018

  13. [21]

    Closed-loop shape control of a Haptic Jamming de- formable surface,

    A. A. Stanley, K. Hata, and A. M. Okamura, “Closed-loop shape control of a Haptic Jamming de- formable surface,” in 2016 IEEE International Conference on Robotics and Automation (ICRA), May 2016, pp. 2718–2724

    2016

  14. [22]

    Relief: a scalable actuated shape display,

    D. Leithinger and H. Ishii, “Relief: a scalable actuated shape display,” in Proceedings of the fourth international conference on Tangible, embedded, and embodied interaction, ser. TEI ’10. New York, NY, USA: Association for Computing Machinery, 2010, pp. 221–222

  15. [23]

    J. Chen, W. Cranton, and M. Fihn, Handbook of visual display technology. Springer Publishing Company, Incorporated, 2011

    2011

  16. [24]

    Practical Structure Design and Control for Digital Clay

    H. Zhu and W. J. Book, “Practical Structure Design and Control for Digital Clay.” American Society of Mechanical Engineers Digital Collection, 2004, pp. 1051–1058

    2004

  17. [25]

    Construction and control of massive hydraulic miniature-actuator-sensor array,

    ——, “Construction and control of massive hydraulic miniature-actuator-sensor array,” in 2006 IEEE Conference on Computer Aided Control System Design, 2006 IEEE International Conference on Con- trol Applications, 2006 IEEE International Symposium on Intelligent Control, Oct. 20...

    2006

  18. [26]

    A control loop structure based on semi-nonnegative matrix factorization for input-coupled systems,

    R. C. Winck, J. Kim, W. J. Book, and H. Park, “A control loop structure based on semi-nonnegative matrix factorization for input-coupled systems,” in 2012 American Control Conference (ACC). IEEE, 2012, pp. 3484–3489

    2012

  19. [27]

    A Control Loop Structure Based on Singular Value Decomposition for Input-Coupled Systems

    R. C. Winck and W. J. Book, “A Control Loop Structure Based on Singular Value Decomposition for Input-Coupled Systems.” American Society of Mechanical Engineers Digital Collection, May 2012, pp. 329–336

    2012

  20. [28]

    Command Generation Techniques for a Pin Array using the SVD and the SNMF,

    R. C. Winck, J. Kim, W. J. Book, and H. Park, “Command Generation Techniques for a Pin Array using the SVD and the SNMF,” IF AC Proceedings Volumes, vol. 45, no. 22, pp. 411–416, 2012

    2012

  21. [29]

    The SVD System for First-Order Linear Systems,

    R. C. Winck and W. J. Book, “The SVD System for First-Order Linear Systems,” IEEE Transactions on Control Systems Technology, vol. 23, no. 3, pp. 1213–1220, May 2015

    2015

  22. [30]

    Passivity and practical considerations for the SNMF System,

    ——, “Passivity and practical considerations for the SNMF System,” in 2016 IEEE International Conference on Robotics and Automation (ICRA). Stockholm, Sweden: IEEE, May 2016, pp. 2669– 2674

    2016

  23. [31]

    Multiplicative valve to control many cylinders,

    K. M. Ferguson, D. Tong, and R. C. Winck, “Multiplicative valve to control many cylinders,” in 2020 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM). Boston, MA, USA: IEEE, July 2020, pp. 673–678

    2020

  24. [32]

    Scalable Fluidic Matrix Circuits for Controlling Large Arrays of Individually Addressable Actuators,

    S. Jadhav, P. E. Glick, M. Ishida, C. Chan, I. Adibnazari, J. P. Schulze, N. Gravish, and M. T. Tolley, “Scalable Fluidic Matrix Circuits for Controlling Large Arrays of Individually Addressable Actuators,” Advanced Intelligent Systems, vol. 5, no. 8, p. 2300011, 2023

    2023

  25. [33]

    Passively addressed robotic morphing surface (parms) based on machine learning,

    J. Wang, M. Sotzing, M. Lee, and A. Chortos, “Passively addressed robotic morphing surface (parms) based on machine learning,” Science Advances, vol. 9, no. 29, p. eadg8019, 2023

    2023

  26. [34]

    Matching pursuits with time-frequency dictionaries,

    S. Mallat and Zhifeng Zhang, “Matching pursuits with time-frequency dictionaries,” IEEE Transac- tions on Signal Processing, vol. 41, no. 12, pp. 3397–3415, Dec. 1993

    1993

  27. [35]

    Mallat, A Wavelet Tour of Signal Processing: The Sparse Way,Third Edition, 3rd ed

    S. Mallat, A Wavelet Tour of Signal Processing: The Sparse Way,Third Edition, 3rd ed. Elsevier Inc., Academic Press, Dec. 2008

    2008

  28. [36]

    Universal approximation using radial-basis-function networks,

    J. Park and I. W. Sandberg, “Universal approximation using radial-basis-function networks,” Neural computation, vol. 3, no. 2, pp. 246–257, 1991. 14

    1991

Pith tools

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