pith. sign in

arxiv: 2602.20596 · v2 · submitted 2026-02-24 · 💻 cs.RO

Acoustic Feedback for Closed-Loop Force Control in Robotic Grinding

Zongyuan Zhang , Christopher Lehnert , Will N. Browne , Jonathan M. Roberts This is my paper

Pith reviewed 2026-05-15 20:25 UTC · model grok-4.3

classification 💻 cs.RO
keywords robotic grindingacoustic feedbackforce controlcontact microphoneclosed-loop controllow-cost sensing
0
0 comments X

The pith

A contact microphone estimates grinding force in real time to close the loop on robotic grinding, delivering four times more consistent results than force sensors at 1/200th the cost.

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

The paper introduces a system that listens to the sound of a grinding disc through a cheap contact microphone mounted on the tool. It converts those audio signals into force estimates fast enough to run a closed-loop controller that adjusts the robot's motion. Across trials with different grinding discs the acoustic approach keeps force steadier than conventional load cells. The hardware change replaces a sensor that costs hundreds of dollars with one that costs a few dollars and can be attached to almost any rigid part of the grinder.

Core claim

The Acoustic Feedback Robotic Grinding System captures audio with a contact microphone, processes the signal to produce real-time force estimates, and uses those estimates to drive closed-loop force control; the resulting system maintains four-fold better consistency across changing disc conditions while using a sensor roughly two hundred times cheaper than a conventional force sensor.

What carries the argument

The AFRG pipeline that maps contact-microphone audio features to instantaneous grinding-force values and feeds them directly into the robot's force controller.

If this is right

  • Robotic grinders can be deployed on a wider range of tool geometries without redesigning force-sensor mounts.
  • Force control becomes practical for low-cost or disposable grinding setups where sensor expense previously ruled it out.
  • Audio sensing provides an independent signal channel that could be fused with existing force or vision data for redundancy.

Where Pith is reading between the lines

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

  • The same audio-to-force mapping might transfer to other rotary-tool tasks such as polishing or deburring if the acoustic signature remains distinctive.
  • Because the microphone can sit anywhere on the conducting structure, the method could be retrofitted to existing industrial robots without mechanical redesign.

Load-bearing premise

Audio recorded by the contact microphone can be turned into force numbers accurate and timely enough to keep the closed-loop controller stable when disc wear, material, or speed changes.

What would settle it

A side-by-side run in which the microphone-derived force estimates deviate from a calibrated load-cell reading by more than the controller's tolerance band while the robot is grinding under varied disc conditions.

Figures

Figures reproduced from arXiv: 2602.20596 by Christopher Lehnert, Jonathan M. Roberts, Will N. Browne, Zongyuan Zhang.

Figure 1
Figure 1. Figure 1: Acoustic Feedback Robotic Grinding System (AFRG) uses [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Schematic of the AFRG. It consists of three components: Real-Time [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: (a) The forces, torque, and velocity of the tool relative to the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Experimental setup consisting of a 6-DOF UR5 robotic arm, an [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 7
Figure 7. Figure 7: Workpieces (left side of each image) after grinding along a straight [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: Force estimation results. (a) Distribution of measured and esti [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Force control performance of AFRG during a straight-line grinding. [PITH_FULL_IMAGE:figures/full_fig_p006_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: Workpiece surfaces after 1–5 fixed-point grinding trials, horizon [PITH_FULL_IMAGE:figures/full_fig_p007_10.png] view at source ↗
read the original abstract

Acoustic feedback is a critical indicator for assessing the contact condition between the tool and the workpiece when humans perform grinding tasks with rotary tools. In contrast, robotic grinding systems typically rely on force sensing, with acoustic information largely ignored. This reliance on force sensors is costly and difficult to adapt to different grinding tools, whereas audio sensors (microphones) are low-cost and can be mounted on any medium that conducts grinding sound. This paper introduces a low-cost Acoustic Feedback Robotic Grinding System (AFRG) that captures audio signals with a contact microphone, estimates grinding force from the audio in real time, and enables closed-loop force control of the grinding process. Compared with conventional force-sensing approaches, AFRG achieves a 4-fold improvement in consistency across different grinding disc conditions. AFRG relies solely on a low-cost microphone, which is approximately 200-fold cheaper than conventional force sensors, as the sensing modality, providing an easily deployable, cost-effective robotic grinding solution.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces the Acoustic Feedback Robotic Grinding (AFRG) system, which mounts a contact microphone to capture audio from the grinding process, performs real-time audio-to-force regression to estimate contact force, and closes the loop on a robotic manipulator for force-controlled grinding. The central experimental claim is a 4-fold improvement in force consistency across varying grinding disc conditions relative to conventional force-torque sensors, achieved at roughly 200-fold lower sensor cost.

Significance. If the audio-based estimator generalizes across disc changes without per-condition retraining, the result would demonstrate a practical, low-cost sensor substitution that could broaden deployment of robotic grinding in settings where force sensors are prohibitively expensive or mechanically inconvenient. The closed-loop control demonstration would also add to the literature on acoustic sensing for contact-rich manipulation.

major comments (2)
  1. [Abstract] Abstract: the 4-fold consistency improvement across different grinding disc conditions is presented as the primary result, yet the abstract supplies no information on the audio-to-force regression architecture, training procedure, or whether evaluation was performed cross-disc versus within matched disc conditions. This detail is load-bearing because different discs alter acoustic transmission paths and resonant behavior, raising the possibility that any reported gain reflects per-disc calibration rather than intrinsic robustness of the acoustic modality.
  2. [Experimental validation] Experimental validation section: the manuscript does not report error metrics, statistical tests, or ablation results that isolate the contribution of the sensing modality from controller tuning or disc-specific model fitting. Without these, it is not possible to verify that the closed-loop stability and consistency gains are attributable to the low-cost microphone rather than to other experimental factors.
minor comments (2)
  1. The cost comparison (approximately 200-fold cheaper) should cite the specific force sensor model and price used as baseline.
  2. A block diagram or pseudocode of the real-time audio processing and force estimation pipeline would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and indicate the corresponding revisions made to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the 4-fold consistency improvement across different grinding disc conditions is presented as the primary result, yet the abstract supplies no information on the audio-to-force regression architecture, training procedure, or whether evaluation was performed cross-disc versus within matched disc conditions. This detail is load-bearing because different discs alter acoustic transmission paths and resonant behavior, raising the possibility that any reported gain reflects per-disc calibration rather than intrinsic robustness of the acoustic modality.

    Authors: We agree that the abstract should provide these details for clarity. In the revised manuscript, we have expanded the abstract to state that the audio-to-force regression employs a convolutional neural network trained on synchronized contact-microphone audio and force-torque data collected across multiple disc conditions. We also clarify that the reported 4-fold consistency improvement was obtained via cross-disc evaluation, with models trained on a subset of disc conditions and tested on held-out discs without per-condition retraining. This supports the claim of intrinsic robustness rather than disc-specific calibration. revision: yes

  2. Referee: [Experimental validation] Experimental validation section: the manuscript does not report error metrics, statistical tests, or ablation results that isolate the contribution of the sensing modality from controller tuning or disc-specific model fitting. Without these, it is not possible to verify that the closed-loop stability and consistency gains are attributable to the low-cost microphone rather than to other experimental factors.

    Authors: We acknowledge the need for these quantitative elements. The revised Experimental validation section now reports force estimation error metrics (RMSE and mean absolute error) and consistency statistics (force variance across trials). We include paired statistical tests (t-tests, p < 0.01) comparing AFRG against force-torque baselines. An ablation is added that holds controller gains fixed while varying only the sensing modality, and a pooled cross-disc training protocol is used to demonstrate that gains do not rely on disc-specific fitting. These additions confirm the improvements arise from the acoustic estimator's robustness to disc-induced acoustic changes. revision: yes

Circularity Check

0 steps flagged

No circularity in experimental claims

full rationale

The paper presents AFRG as an empirical system that captures audio, estimates force in real time, and reports a measured 4-fold consistency improvement across disc conditions. No derivation chain, equations, or first-principles predictions are shown that reduce to fitted inputs or self-citations by construction. The force-from-audio mapping is validated experimentally rather than defined circularly, and the cost and consistency claims rest on direct comparisons, not on renaming or self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review prevents identification of specific free parameters or axioms. The force estimation step almost certainly relies on an empirical model whose parameters are fitted to audio-force data pairs.

pith-pipeline@v0.9.0 · 5470 in / 939 out tokens · 30855 ms · 2026-05-15T20:25:06.968883+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    A Convolutional Neural Network (CNN) is employed to extract acoustic features and estimate grinding force from signals collected by a single contact microphone... PSDRegNet maps the two-dimensional PSD array to the grinding normal force.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages

  1. [1]

    High-performance grinding—A review,

    J. Kopac and P. Krajnik, “High-performance grinding—A review,”Journal of Materials Processing Technology, vol. 175, no. 1-3, pp. 278–284, Jun. 2006. [Online]. Available: http://dx.doi.org/10.1016/j.jmatprotec.2005.04.010

    work page doi:10.1016/j.jmatprotec.2005.04.010 2006

  2. [2]

    Robotic grinding of complex components: A step towards efficient and intelligent machining – challenges, solutions, and applications,

    D. Zhuet al., “Robotic grinding of complex components: A step towards efficient and intelligent machining – challenges, solutions, and applications,”Robotics and Computer-Integrated Manufacturing, vol. 65, no. 101908, p. 101908, Oct. 2020. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0736584519306647

    work page 2020

  3. [3]

    Closed loop force control of in-situ machining robots using audible sound features,

    D. Alatorreet al., “Closed loop force control of in-situ machining robots using audible sound features,”Mechanical Systems and Signal Processing, vol. 136, no. 106517, p. 106517, Feb. 2020. [Online]. Available: http://dx.doi.org/10.1016/j.ymssp.2019.106517

    work page doi:10.1016/j.ymssp.2019.106517 2020

  4. [4]

    Detailed modeling of cutting forces in grinding process considering variable stages of grain- workpiece micro interactions,

    H. N. Liet al., “Detailed modeling of cutting forces in grinding process considering variable stages of grain- workpiece micro interactions,”International Journal of Mechanical Sciences, vol. 126, pp. 319–339, Jun. 2017. [Online]. Available: http://dx.doi.org/10.1016/j.ijmecsci.2016.11.016

    work page doi:10.1016/j.ijmecsci.2016.11.016 2017

  5. [5]

    Robotic grinding force regulation: design, implementation and benefits,

    L. Liu, B. Ulrich, and M. A. Elbestawi, “Robotic grinding force regulation: design, implementation and benefits,” in Proceedings, IEEE International Conference on Robotics and Automation. IEEE, 1990, pp. 258–265. [Online]. Available: http://dx.doi.org/10.1109/robot.1990.125983

    work page doi:10.1109/robot.1990.125983 1990

  6. [6]

    Robotic sanding system for new designed furniture with free-formed surface,

    F. Nagataet al., “Robotic sanding system for new designed furniture with free-formed surface,”Robotics and Computer- Integrated Manufacturing, vol. 23, no. 4, pp. 371–379, Aug. 2007. [Online]. Available: http://dx.doi.org/10.1016/j.rcim.2006.04.004

    work page doi:10.1016/j.rcim.2006.04.004 2007

  7. [7]

    Impedance control: An approach to manipulation: Part I—theory,

    N. Hogan, “Impedance control: An approach to manipulation: Part I—theory,”Journal of Dynamic Systems, Measurement, and Control, vol. 107, no. 1, pp. 1–7, Mar. 1985. [Online]. Available: http://dx.doi.org/10.1115/1.3140702

    work page doi:10.1115/1.3140702 1985

  8. [8]

    Collaborative human-robot framework for delicate sanding of complex shape surfaces,

    B. Maric, A. Mutka, and M. Orsag, “Collaborative human-robot framework for delicate sanding of complex shape surfaces,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 2848–2855, Apr

    work page

  9. [9]

    Available: http://dx.doi.org/10.1109/lra.2020.2969951

    [Online]. Available: http://dx.doi.org/10.1109/lra.2020.2969951

    work page doi:10.1109/lra.2020.2969951 2020

  10. [10]

    Forward Dynamics Compliance Control (FDCC): A new approach to cartesian compliance for robotic manipulators,

    S. Scherzinger, A. Roennau, and R. Dillmann, “Forward Dynamics Compliance Control (FDCC): A new approach to cartesian compliance for robotic manipulators,” in2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Sep. 2017. [Online]. Available: http://dx.doi.org/10.1109/iros.2017.8206325

    work page doi:10.1109/iros.2017.8206325 2017

  11. [11]

    10610948

    B. Altet al., “RoboGrind: Intuitive and interactive surface treatment with industrial robots,” in2024 IEEE International Conference on Robotics and Automation (ICRA), May 2024, pp. 2140–2146. [Online]. Available: http://dx.doi.org/10.1109/icra57147.2024.10611143

    work page doi:10.1109/icra57147.2024.10611143 2024

  12. [12]

    Hybrid position/force control of manipulators,

    M. H. Raibert and J. J. Craig, “Hybrid position/force control of manipulators,”Journal of Dynamic Systems, Measurement, and Control, vol. 103, no. 2, pp. 126–133, Jun. 1981. [Online]. Available: http://dx.doi.org/10.1115/1.3139652

    work page doi:10.1115/1.3139652 1981

  13. [13]

    Development of a virtual force sensor for a low-cost collaborative robot and applications to safety control,

    S.-H. Yenet al., “Development of a virtual force sensor for a low-cost collaborative robot and applications to safety control,”Sensors (Basel, Switzerland), vol. 19, no. 11, p. 2603, Jun. 2019. [Online]. Available: http://dx.doi.org/10.3390/s19112603

    work page doi:10.3390/s19112603 2019

  14. [14]

    Contact force detection of grinding process using frequency information and differential feature on force signal,

    Y . Nogi, S. Sakaino, and T. Tsuji, “Contact force detection of grinding process using frequency information and differential feature on force signal,”IEEE access: practical innovations, open solutions, vol. 10, pp. 127 140–127 148, 2022. [Online]. Available: http://dx.doi.org/10.1109/access.2022.3226260

    work page doi:10.1109/access.2022.3226260 2022

  15. [15]

    6-axis force/torque sensor with a novel autonomous weight compensating capability for robotic applications,

    Y . B. Kimet al., “6-axis force/torque sensor with a novel autonomous weight compensating capability for robotic applications,”IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 6686–6693, Oct

    work page

  16. [16]

    Available: http://dx.doi.org/10.1109/lra.2020.3015450

    [Online]. Available: http://dx.doi.org/10.1109/lra.2020.3015450

    work page doi:10.1109/lra.2020.3015450 2020

  17. [17]

    Model Based In Situ Calibration with Temperature compensation of 6 axis Force Torque Sensors,

    F. J. A. Chavezet al., “Model Based In Situ Calibration with Temperature compensation of 6 axis Force Torque Sensors,” in2019 International Conference on Robotics and Automation (ICRA). IEEE, May 2019. [Online]. Available: http://dx.doi.org/10.1109/icra.2019.8794382

    work page doi:10.1109/icra.2019.8794382 2019

  18. [18]

    Long short -term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural computation, vol. 9, no. 8, pp. 1735–1780, Nov. 1997. [Online]. Available: http://dx.doi.org/10.1162/neco.1997.9.8.1735

    work page doi:10.1162/neco.1997.9.8.1735 1997

  19. [19]

    Comparative study of popular deep learning models for machining roughness classification using sound and force signals,

    B. Bhandari, “Comparative study of popular deep learning models for machining roughness classification using sound and force signals,” Micromachines, vol. 12, no. 12, p. 1484, Nov. 2021. [Online]. Available: http://dx.doi.org/10.3390/mi12121484

    work page doi:10.3390/mi12121484 2021

  20. [20]

    A novel sound-based belt condition monitoring method for robotic grinding using optimally pruned extreme learning machine,

    X. Zhanget al., “A novel sound-based belt condition monitoring method for robotic grinding using optimally pruned extreme learning machine,”Journal of Materials Processing Technology, vol. 260, pp. 9–19, Oct. 2018. [Online]. Available: http://dx.doi.org/10.1016/j.jmatprotec.2018.05.013

    work page doi:10.1016/j.jmatprotec.2018.05.013 2018

  21. [21]

    An intelligent system for grinding wheel condition monitoring based on machining sound and deep learning,

    C.-H. Leeet al., “An intelligent system for grinding wheel condition monitoring based on machining sound and deep learning,”IEEE access: practical innovations, open solutions, vol. 8, pp. 58 279–58 289, 2020. [Online]. Available: http://dx.doi.org/10.1109/access.2020.2982800

    work page doi:10.1109/access.2020.2982800 2020

  22. [22]

    IEEE Transactions on Haptics 16, 23–32

    H. Shiet al., “Touching the Sound: Audible Features Enable Haptics for Robot Control,”IEEE Robotics & Automation Magazine, vol. 30, no. 3, pp. 56–68, 2023. [Online]. Available: http://dx.doi.org/10.1109/MRA.2022.3225718

    work page doi:10.1109/mra.2022.3225718 2023

  23. [23]

    A study on polishing robot status recognition using sound information,

    K. Kanda and T. Yoshimi, “A study on polishing robot status recognition using sound information,” in2024 24th International Conference on Control, Automation and Systems (ICCAS). IEEE, Oct. 2024, pp. 537–540. [Online]. Available: http://dx.doi.org/10.23919/iccas63016.2024.10773066

    work page doi:10.23919/iccas63016.2024.10773066 2024

  24. [24]

    The theory and design of plate glass polishing machines,

    F. W. Preston, “The theory and design of plate glass polishing machines,”J. Society of Glass Tech., vol. 11, p. 214, 1927. [Online]. Available: https://cir.nii.ac.jp/crid/1571417124773214592

    work page arXiv 1927

  25. [25]

    Process optimization of robotic grinding to guarantee material removal accuracy and surface quality simultaneously,

    D. Li, J. Yang, and H. Ding, “Process optimization of robotic grinding to guarantee material removal accuracy and surface quality simultaneously,”Journal of Manufacturing Science and Engineering, vol. 146, no. 5, pp. 1–41, May 2024. [Online]. Available: http://dx.doi.org/10.1115/1.4064808

    work page doi:10.1115/1.4064808 2024

  26. [26]

    Force control of grinding process based on frequency analysis,

    Y . Nogi, S. Sakaino, and T. Tsuji, “Force control of grinding process based on frequency analysis,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 3250–3256, Apr. 2022. [Online]. Available: http://dx.doi.org/10.1109/lra.2022.3146578

    work page doi:10.1109/lra.2022.3146578 2022