Pith. sign in

REVIEW 3 major objections 4 minor 19 references

Aerial Layouting: Design and Control of a Compliant and Actuated End-Effector for Precise In-flight Marking on Ceilings

T0 review · 3 major / 4 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 compliant, actuated end-effector mounted on a drone can mark lines on ceilings with millimetre accuracy, without requiring an accurate model of the drone or sophisticated control.

desk verdict A genuinely new compliant end-effector for aerial ceiling marking with a strong ablation study, but the 'millimetre accuracy' headline is measured through a self-referential camera loop and one reported spring constant is off by orders of magnitude. read the letter →

arxiv 2608.10987 v1 pith:3TUXPRGE submitted 2026-08-11 cs.RO

Christian Lanegger , Marco Ruggia , Marco Tognon , Lionel Ott , Roland Siegwart This is my paper · ORCID
classification cs.RO
keywords aerialmanipulationend-effectordesignGough-Stewartplatformcompliantmechanismmillimetreaccuracyceilingmarkingomnidirectionalvehicleablationstudy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that millimetre-precision marking on ceilings can be achieved not by making the drone itself more precise, but by adding a smart mechanical end-effector that handles the accuracy at the tool tip. The end-effector is a passive Gough-Stewart platform (two platforms connected by six spring-dampers) topped with three actuated omni-wheels and an onboard camera that tracks the moving upper platform. In flight tests, the full system followed circular and 'hello' trajectories with mean absolute errors of roughly 1 to 3 mm and a worst-case bound under 4 mm, while the same drone flying free had errors above 20 mm. The authors conclude that this makes precise aerial layouting practical without the need for an accurate system model or a complex control architecture.

What carries the argument

The load-bearing component is a compliant Gough-Stewart platform, a six-degree-of-freedom parallel mechanism whose lower platform is fixed to the drone and whose upper platform is connected by six spring-dampers. These springs provide passive vibration isolation and allow the drone to push the platform against the ceiling, and the platform geometry is optimised by maximising the smallest eigenvalue of the Hessian of the total energy field, so that the upper platform is pulled back to the centre from any displacement. On the upper platform, three actuated omni-wheels create multiple contact points and closed-loop positioning: a cascaded controller (a proportional outer loop with feed-forward velocity and a PI inner loop) drives the wheel servos to follow the desired trajectory. An upward-facing camera on the lower platform estimates the relative pose of the upper platform by tracking a checkerboard fiducial pattern, feeding the controller and providing the measurement used to report accuracy.

What would settle it

Directly measure the lines left by the marker on the ceiling (e.g., with a calibrated scanner or precision ruler) and compare them to the commanded trajectories; if the actual line deviation systematically exceeds the reported mean absolute errors of 1–3 mm, then the camera-plus-drone-ground-truth error chain is masking the true marking error.

Watch

Extended reading notes

Core claim

The central discovery is that four design features — compliance, multiple contact points, actuation, and self-containment — together let a simple aerial vehicle mark surfaces with millimetre accuracy. The paper demonstrates this with an ablation study: removing actuation feedback, compliance, or the extra contact points degrades end-effector error from about 1.0 mm to 14–40 mm, while tracking a circle with the full system keeps the mean absolute error around 1 mm. The geometry of the Gough-Stewart platform is chosen by an optimisation that maximises the smallest eigenvalue of the energy field's Hessian, ensuring the upper platform always returns to its centred position after a disturbance. In a velocity sweep on complex trajectories, this end-effector achieves tracking errors comparable to a state-of-the-art delta-arm system that uses nonlinear model predictive control, and it does so without any retuning of the flight controller.

Load-bearing premise

The claim of millimetre marking accuracy assumes that the onboard camera's estimate of the upper-platform pose, measured as about 0.8 mm accurate in a static test, stays that accurate during active flight and contact, because the marked lines themselves are never directly measured.

Editorial extensions

If this is right

  • The end-effector improves tracking accuracy by an order of magnitude over the bare drone: full-system mean absolute error is about 1.0 mm in the design-validation circle, versus 22.6 mm in free flight.
  • Repeatedly marking the same 250 mm circle shows the end-effector error rarely exceeds 2 mm, even where the drone's own tracking error peaks at 3.5 cm, meaning the tool actively compensates for vehicle drift.
  • Across a velocity sweep up to 27.5 cm/s on a 'hello' trajectory, end-effector mean absolute error stays below 3 mm (worst-case below 4 mm), matching the 2.1–2.8 mm reported for a nonlinear-model-predictive-control delta-arm system.
  • The ablation study shows that closed-loop actuation is the dominant error-reducing feature, with compliance adding a further improvement; without feedback the error remains large.
  • Because the end-effector is self-contained and requires no controller retuning of the vehicle, the same tool can be transferred to different aerial platforms.

Reading between the lines

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

  • If the onboard camera maintains its 0.8 mm pose accuracy under real-world lighting and vibration, the same end-effector should be mountable on a conventional quadrotor, not just an omnidirectional one, since the design does not depend on the vehicle's model.
  • The energy-field optimisation used to stabilise the Gough-Stewart geometry could be reused as a general design tool for passive compliant manipulators, choosing leg placements that guarantee a return-to-centre workspace.
  • A direct measurement of the ink lines left on the ceiling would settle whether the reported accuracy reflects true marking quality or the fidelity of the camera-ground-truth estimation chain; the paper only infers tool error from drone pose measurements.
  • The approach may extend to inclined and curved ceilings (which the paper names as future work) by letting the compliant stage absorb out-of-plane motion while the omni-wheels follow the surface.
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 / 4 minor

Summary. The paper presents a novel aerial layouting end-effector based on a passive Gough-Stewart platform with three actuated omni-wheels, an upward-facing camera tracking a ChArUco board, and a local feedback controller. The geometry is chosen through an optimization of the Hessian of a spring-energy field. Experiments include an ablation study, a ten-trial repeatability test, trajectory-size sweeps, and a velocity sweep compared with the NMPC delta-arm system of [18]. The central claim is that the system marks lines on ceilings with millimetre accuracy without requiring an accurate model or sophisticated control of the aerial vehicle.

Significance. If the central claim is supported, this is a valuable systems contribution to aerial manipulation: a mechanically compliant, self-contained end-effector that reduces end-effector error by an order of magnitude relative to the bare aerial vehicle, with a clear ablation showing the role of actuation and feedback. The comparison to a state-of-the-art NMPC delta-arm system is informative, and the repeatability tests provide useful data. However, the advertised 'millimetre marking' claim is currently inferred from transformed pose estimates rather than from the actual marked lines, and the proposed MWCE error bound is not a true worst-case bound. These issues are load-bearing for the central claim and require additional validation before acceptance.

major comments (3)
  1. [IV-B and Section IV-A] The claimed millimetre marking accuracy is not directly measured. Section IV-B states that the end-effector is occluded from the Vicon system during marking and that the end-effector error is obtained by transforming the aerial-vehicle Vicon ground truth through the camera-based estimate of the upper-platform pose. Because the same camera tracking system provides feedback to the end-effector controller and is also used to compute the reported error, any bias, latency, or dynamic degradation of the camera estimate is invisible in the MAE. The static camera-accuracy characterisation in Section IV-A was performed while the upper platform was manually moved, not under flight vibration, motor-induced shaking, or contact. To support the abstract and Section IV-E claims, the authors should measure the actual marked lines, provide an independent ground-truth measurement of the upper platform during flight/contact, or at least characterise the camera error under realistic operating conditions and show that the claimed accuracy is robust to that error.
  2. [Section IV-B] The Mean Worst Case Error (MWCE) is defined as the sum of the end-effector MAE and the camera tracking error, and is described as 'an upper bound on the worst case performance.' A sum of two mean errors is not a worst-case bound and cannot bound the tail of the true error distribution. For example, in Table II the full system reports P90 = 1.7 mm and MWCE = 1.8 mm, but this does not protect against a larger tail in the camera error distribution. The authors should report the error distribution, a genuine worst-case or quantile bound, or direct measurements of marking error instead of calling MAE + camera MAE an upper bound.
  3. [Section III-D and Table I] There is a numerical inconsistency in the force-torque sensor discussion. Table I lists k_spring = 0.2 N/m, but Section III-D states that 'for a spring constant of k_spring = 0.2 N/m, the sensor would require to sense a force of 0.07 N to measure a translation of 1 mm.' For k = 0.2 N/m, a 1 mm displacement requires 0.0002 N, not 0.07 N; for k = 0.2 N/mm it would require 0.2 N. The same inconsistency affects the statement that a force of 15 N compresses the end-effector to its nominal height. This must be corrected with consistent units and values, since the argument against a force-torque sensor rests on this calculation.
minor comments (4)
  1. [Table II] The header alignment and row entries for Table II are difficult to parse; in particular the entry '2.47.7' appears to be two numbers ('2.4' and '7.7') without a separator, and the distinction between 'End-Effector xy-error' and 'OMA V xyz-error' columns should be clarified.
  2. [Section III-D] The sentence 'Currently, no camera feedback is currently used to correct for any error in Z-direction' contains a duplicated 'currently' and should be rephrased.
  3. [Section III-B] The stability optimisation uses the arbitrary scaling that 1 mm of translation is equivalent to 1 degree of rotation in the Hessian-based metric. Since this scaling can change the optimised geometry, a sensitivity analysis or a physically motivated justification would strengthen the design contribution.
  4. [Section IV-E] The comparison with [18] would be more convincing if the authors reported the same error metric and trajectory normalisation used by [18], and if they explicitly stated whether the comparison is based on measured marking output or on pose tracking error.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the precision claim is an experimental result with an indirect measurement, not a derivation from fitted inputs.

full rationale

The paper's derivation chain is: derive four design objectives, optimize the Gough-Stewart geometry by maximizing the smallest Hessian eigenvalue of an energy field (Eq. 4), implement camera-based tracking of the upper platform, and evaluate via ablation, repeatability, and a velocity sweep benchmarked against [18]. No model parameter is fitted to the marking-error outcome; the stability metric is used to select geometry, and the ablation studies independently toggle contact, compliance, actuation, and feedback. The only author-overlapping self-citation is [1], used for the vehicle's impedance controller; it is a pre-existing component, not a theorem invoked to justify the end-effector's accuracy. The notable weakness is in Section IV-B, where the end-effector is occluded from Vicon and the reported MAE is computed from Vicon aerial-vehicle poses transformed through the same camera estimate used by the upper-platform controller. The paper explicitly acknowledges this: 'The end-effector Mean Absolute Error (MAE) presented in the following experiments does thus not include the tracking error of the camera system.' It then defines MWCE as MAE plus the statically measured 0.8 mm camera error and calls it an upper bound, although a sum of means is not a worst-case bound and the camera error was not characterized under flight vibration or contact. This is a measurement-validity limitation that weakens the 'millimetre accuracy' claim, but it is not circular: the camera was independently calibrated against Vicon in Section IV-A, and the reported MAE is not produced by fitting or defining the target result into the measurement. No circular step is present.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a set of design parameters (geometry, spring force, wheel size) and modeling assumptions (planar contact, constant height, energy-based stability). None of these are fitted to the marking-error outcome, so the circularity burden is low. The most fragile assumptions are the accuracy of the camera tracking chain and the constancy of the drone's Z position during contact.

free parameters (4)
  • Pose scaling in stability metric = 1 mm position = 1 degree attitude
    Hand-chosen weighting in the six-dimensional energy Hessian (Section III-B). Different scalings yield different optimal geometries, and this ratio is not justified from physical requirements.
  • Nominal compression force = 15 N
    Chosen as a trade-off between available thrust and ceiling grip (Section III-A). It determines spring stiffness and contact behavior, and it is central to the multiple-contact-points design.
  • Spring stiffness k_spring = 0.2 N/m (Table I)
    The printed value is inconsistent with the 15 N force needed to reach nominal height and with the 0.07 N force quoted for 1 mm deflection. This inconsistency affects the quantitative basis of the energy analysis and the force-torque sensor feasibility argument.
  • Omni-wheel diameter = 100 mm
    Design choice for clearance and positional resolution (Section III-C). It affects actuation authority and the achievable tracking accuracy.
assumptions (5)
  • domain assumption The smallest eigenvalue of the energy Hessian at nominal height is a valid proxy for the upper platform's tendency to return to its centered position.
    Section III-B assumes this stability metric predicts physical recovery behavior; it is validated only by visual inspection of energy fields, not by dynamic stability tests or direct measurement of recovery.
  • ad hoc to paper One millimetre of translation is equivalent to one degree of rotation in the stability metric.
    Explicitly stated in Section III-B as a chosen scaling. It is arbitrary and directly influences which geometry is selected as optimal.
  • domain assumption The upper platform can be modeled as a three-wheeled omnidirectional robot moving on a planar ceiling without slip.
    Section III-C uses this kinematic model for the controller. It assumes ideal omni-wheel behavior, which may not hold under varying contact forces and ceiling friction.
  • domain assumption The aerial vehicle can maintain a constant Z reference to keep the end-effector compressed at its nominal height.
    Section III-E states no camera feedback is used to correct Z errors; the stored reference is kept constant. This assumes the drone's height control is accurate enough not to affect contact force or wheel traction.
  • domain assumption The camera tracking error and the end-effector MAE are independent and can be added to form the Mean Worst Case Error.
    Section IV-B defines MWCE as the sum of MAE and camera error. The actual correlation between camera errors and vehicle motion during marking is not measured.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Aerial Layouting: Design and Control of a Compliant and Actuated End-Effector for Precise In-flight Marking on Ceilings." pith.science (2026). https://pith.science/paper/3TUXPRGE

@misc{pith2026260810987,
  author       = {Pith},
  title        = {Pith review of: Aerial Layouting: Design and Control of a Compliant and Actuated End-Effector for Precise In-flight Marking on Ceilings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3TUXPRGE}},
  note         = {Machine review of arXiv:2608.10987}
}
read the original abstract

Aerial robots have demonstrated impressive feats of precise control, such as dynamic flight through openings or highly complex choreographies. Despite the accuracy needed for these tasks, there are problems that require levels of precision that are challenging to achieve today. One such problem is aerial interaction. Advances in aerial robot design and control have made such contact-based tasks possible and opened up research into challenging real-world tasks, including contact-based inspection. However, while centimetre accuracy is sufficient and achievable for inspection tasks, the positioning accuracy needed for other problems, such as layouting on construction sites or general push-and-slide tasks, is millimetres. To achieve such a high precision, we propose a new aerial system composed of an aerial vehicle equipped with a novel "smart" end-effector leveraging a stability-optimized Gough-Stewart mechanism. We present its design process and features incorporating the principles of compliance, multiple contact points, actuation, and self-containment. In experiments, we verify that the design choices made for our novel end-effector are necessary to obtain the desired positioning precision. Furthermore, we demonstrate that we can reliably mark lines on ceilings with millimetre accuracy without the need for precise modeling or sophisticated control of the aerial robot.

Figures

Figures reproduced from arXiv: 2608.10987 by the authors.

Figure 1
Figure 1. Layouting end-effector based on a passive Gough [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Definition of geometry variables used for the opti [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Energy field for the optimized Gough-Stewart structure [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Mean absolute error of the upper platform in the xy [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Average tracking error of the end-effector (top) and OMAV (bottom) for ten marking sequences of a circle with a [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Absolute tracking error of end-effector for different [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Violin plots (top) of the absolute tracking error of the end-effector and OMAV with corresponding visual error (bottom) [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 10 canonical work pages

  1. [18]

    Aerial Manipulation Using Hybrid Force and Position NMPC Applied to Aerial Writing

    Dimos Tzoumanikas, Felix Graule, Qingyue Yan, Dhruv Shah, Marija Popovic, and Stefan Leutenegger. Aerial Manipulation Using Hybrid Force and Position NMPC Applied to Aerial Writing. InProceedings of Robotics: Science and Systems, Corvalis, Oregon, USA, July 2020. doi: 10.15607/RSS.2020.XVI.046

    work page doi:10.15607/rss.2020.xvi.046 2020

  2. [1]

    Karen Bodie, Maximilian Brunner, Michael Pantic, Ste- fan Walser, Patrick Pf ¨andler, Ueli Angst, Roland Sieg- wart, and Juan I. Nieto. An Omnidirectional Aerial Ma- nipulation Platform for Contact-Based Inspection.CoRR, abs/1905.03502, 2019

    arXiv 1905

  3. [2]

    Dy- namic End Effector Tracking With an Omnidirectional Parallel Aerial Manipulator.IEEE Robotics and Automa- tion Letters, 6(4):8165–8172, 2021

    Karen Bodie, Marco Tognon, and Roland Siegwart. Dy- namic End Effector Tracking With an Omnidirectional Parallel Aerial Manipulator.IEEE Robotics and Automa- tion Letters, 6(4):8165–8172, 2021. doi: 10.1109/LRA. 2021.3101864

    arXiv 2021

  4. [3]

    An integrated delta manipulator for aerial repair: A new aerial robotic system.IEEE Robotics & Automation Magazine, 26(1):54–66, 2019

    Pisak Chermprayong, Ketao Zhang, Feng Xiao, and Mirko Kovac. An integrated delta manipulator for aerial repair: A new aerial robotic system.IEEE Robotics & Automation Magazine, 26(1):54–66, 2019

    work page 2019

  5. [4]

    Uni- fied temporal and spatial calibration for multi-sensor systems

    Paul Furgale, Joern Rehder, and Roland Siegwart. Uni- fied temporal and spatial calibration for multi-sensor systems. In2013 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 1280–1286, 2013. doi: 10.1109/IROS.2013.6696514

    arXiv 2013

  6. [5]

    Getting Ready for Robotics in Property Development and Building, 2018

    Boston Consulting Group. Getting Ready for Robotics in Property Development and Building, 2018. Last accessed on 2022-01-24

    work page 2018

  7. [6]

    A 6-DOF passive vibra- tion isolator based on Stewart structure with X-shaped legs.Nonlinear Dynamics, 91(1):157–185, 2018

    Feizhou Hu and Xingjian Jing. A 6-DOF passive vibra- tion isolator based on Stewart structure with X-shaped legs.Nonlinear Dynamics, 91(1):157–185, 2018

    work page 2018

  8. [7]

    Wall contact by octo-rotor UA V with one DoF manipulator for bridge inspection

    Takahiro Ikeda, Shogo Yasui, Motoharu Fujihara, Kenichi Ohara, Satoshi Ashizawa, Akihiko Ichikawa, Akihisa Okino, Takeo Oomichi, and Toshio Fukuda. Wall contact by octo-rotor UA V with one DoF manipulator for bridge inspection. In2017 IEEE/RSJ International Con- ference on Intelligent Robots and Systems (IROS), pages 5122–5127, 2017. doi: 10.1109/IROS.201...

    arXiv 2017

Show all 19 references
  1. [8]

    A. E. Jimenez-Cano, J. Braga, G. Heredia, and A. Ollero. Aerial manipulator for structure inspection by contact from the underside.IEEE International Conference on Intelligent Robots and Systems, 2015-December:1879– 1884, 12 2015. ISSN 21530866. doi: 10.1109/IROS. 2015.7353623

    2015

  2. [9]

    Direct Force Feedback Control and Online Multi-Task Optimization for Aerial Manipulators.IEEE Robotics and Automation Letters, 5(2):331–338, 4 2020

    Gabriele Nava, Quentin Sabl ´e, Marco Tognon, Daniele Pucci, and Antonio Franchi. Direct Force Feedback Control and Online Multi-Task Optimization for Aerial Manipulators.IEEE Robotics and Automation Letters, 5(2):331–338, 4 2020. ISSN 23773766. doi: 10.1109/ LRA.2019.2958473

    2020

  3. [10]

    Past, Present, and Future of Aerial Robotic Manipulators.IEEE Transac- tions on Robotics, pages 1–20, 2021

    Anibal Ollero, Marco Tognon, Alejandro Suarez, Dongjun Lee, and Antonio Franchi. Past, Present, and Future of Aerial Robotic Manipulators.IEEE Transac- tions on Robotics, pages 1–20, 2021. doi: 10.1109/TRO. 2021.3084395

    2021

  4. [11]

    Open Source Computer Vision Library, 2015

    OpenCV . Open Source Computer Vision Library, 2015

    2015

  5. [12]

    Sanchez-Cuevas, Antonio Gonzalez-Morgado, Nicolas Cortes, Diego B

    Pedro J. Sanchez-Cuevas, Antonio Gonzalez-Morgado, Nicolas Cortes, Diego B. Gayango, Antonio E. Jimenez- Cano, An ´ıbal Ollero, and Guillermo Heredia. Fully- Actuated Aerial Manipulator for Infrastructure Contact Inspection: Design, Modeling, Localization, and Con- trol.Sensor...

    2020 doi

  6. [13]

    A platform with six degrees of freedom

    Doug Stewart. A platform with six degrees of freedom. Proceedings of the institution of mechanical engineers, 180(1):371–386, 1965

    1965

  7. [14]

    A modified Stewart platform manipulator with improved dexterity.IEEE Transactions on Robotics and Automation, 9(2):166–173, 1993

    Robert S Stoughton and Tatsuo Arai. A modified Stewart platform manipulator with improved dexterity.IEEE Transactions on Robotics and Automation, 9(2):166–173, 1993

    1993

  8. [15]

    Vega, Manuel Fernandez, Guillermo Heredia, and Anibal Ollero

    Alejandro Suarez, Victor M. Vega, Manuel Fernandez, Guillermo Heredia, and Anibal Ollero. Benchmarks for Aerial Manipulation.IEEE Robotics and Automation Letters, 5(2):2650–2657, 2020. doi: 10.1109/LRA.2020. 2972870

    2020 doi

  9. [16]

    Tello-Chavez, Enrico Gas- parin, Quentin Sabl ´e, Davide Bicego, Anthony Mallet, Marc Lany, Gilles Santi, Bernard Revaz, Juan Cort´es, and Antonio Franchi

    Marco Tognon, Hermes A. Tello-Chavez, Enrico Gas- parin, Quentin Sabl ´e, Davide Bicego, Anthony Mallet, Marc Lany, Gilles Santi, Bernard Revaz, Juan Cort´es, and Antonio Franchi. A Truly-Redundant Aerial Manipulator System With Application to Push-and-Slide Inspection in Indu...

    2019

  10. [17]

    Novel Aerial Manipulator for Accurate and Robust Industrial NDT Contact Inspection: A New Tool for the Oil and Gas Inspection Industry.Sensors 2019, V ol

    Miguel ´Angel Trujillo, Jos ´e Ramiro Mart ´ınez-De Dios, Carlos Mart´ın, Antidio Viguria, and An´ıbal Ollero. Novel Aerial Manipulator for Accurate and Robust Industrial NDT Contact Inspection: A New Tool for the Oil and Gas Inspection Industry.Sensors 2019, V ol. 19, Page 13...

    2019

  11. [19]

    Gareth Pierce, and Juan Nieto

    Robert Watson, Mina Kamel, Dayi Zhang, Gordon Do- bie, Charles MacLeod, S. Gareth Pierce, and Juan Nieto. Dry Coupled Ultrasonic Non-Destructive Evaluation Us- ing an Over-Actuated Unmanned Aerial Vehicle.IEEE Transactions on Automation Science and Engineering, pages 1–16, 202...

    2021

Pith tools

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