Pith. sign in

REVIEW 4 major objections 5 minor 33 references

Shaping Wind-Tunnel Airflow for Unmanned Aerial Vehicles using Online Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read A seven-fan wind tunnel can be programmed in roughly six iterations to reproduce target airflow distributions, including one that halves a soaring robot's tracking error.

desk verdict A practical, well-executed experimental study on shaping wind-tunnel airflow with a coarse-model-guided online learning; the flight-test numbers and unverified convergence assumptions are soft, but the core is solid. read the letter →

arxiv 2608.03378 v1 pith:QE2TPGLJ submitted 2026-08-04 cs.RO cs.AIcs.SYeess.SY

classification cs.ROcs.AIcs.SYeess.SY
keywords onlinelearningwindtunnelcontrolairflowshapingsoaringrobotGaussianprocessregressionmodel-basediterativeunderactuatedUAVtestbed
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 a seven-propeller vertical wind tunnel can be programmed, in roughly six iterations, to reproduce a wide range of target airflow distributions from force measurements alone, without CFD simulation. The method couples a deliberately coarse physical model of each propeller's airflow with Gaussian-process regression of live sensor data, using the model's pseudo-inverse to turn the 317-point tracking error into seven motor-speed updates. The authors demonstrate convergence to uniform, Gaussian, and parabolic force fields, and show that the parabolic field, designed for passive soaring, reduces the average position tracking error of a soaring robot by 50% and attitude error by 20%. They also show that the algorithm still converges when the central motor is switched off, indicating robustness to hardware changes. A sympathetic reader would take the central claim to be that a low-fidelity model plus a small number of real-world measurements can control a turbulent flow field well enough to support repeatable aerial-robotics experiments.

What carries the argument

The central object is the update rule in Algorithm 1, $U_{k+1} = U_k + \lambda_k (\partial \Phi_{\mathrm{pr}}/\partial U|_{U_k})^+ \Delta_k$, where $\Delta_k$ is the Gaussian-process-estimated error between measured and desired force at 317 key points and $(\cdot)^+$ is the Moore-Penrose pseudo-inverse. $\Phi_{\mathrm{pr}}$ is a superposition of seven Gaussian jet models, each fitted to a small batch of on-motor thrust measurements, plus an explicit central-motor term (Eq. 4) that captures the observed edge enhancement. The coarse model supplies a descent direction that maps a high-dimensional error back to the seven motor commands; the Gaussian-process regression supplies the error measurem

What would settle it

Measure the true input-output Jacobian $\partial F/\partial U$ at several motor settings by finite differences and check whether inequality (8b) holds; if it fails for, say, a target with two sharp separated peaks, the tracking error should increase at some iteration. A simpler observation that would settle it: run Algorithm 1 on a sharp step-shaped target field and look for a monotone error decrease — any sustained increase after iteration 1 falsifies the claimed gradient alignment.

Watch

Extended reading notes

Core claim

The paper's central claim is that an online learning scheme with a highly simplified physical model converges linearly to the best achievable match of any prespecified airflow distribution over a 317-point grid, despite controlling only seven motor speeds. At each iteration, a moving force sensor collects roughly a thousand measurements, a Gaussian process reconstructs the full force field, and the motor commands are updated by $U \leftarrow U + \lambda (\partial \Phi_{\mathrm{pr}}/\partial U)^+ (F_{\mathrm{des}} - F_k)$, where $\Phi_{\mathrm{pr}}$ is a superposition of seven Gaussian jet models, with one deliberately ad hoc term for the central motor's surprising edge-amplification effect.

Load-bearing premise

The method assumes the simplified Gaussian fan model always points motor updates in the right direction, downhill on the true airflow error, but the paper never directly verifies that this alignment holds during the iterations.

Editorial extensions

If this is right

  • A multi-fan wind tunnel can be reconfigured to a new target airflow distribution in minutes, about six to seven iterations at roughly 20 seconds each, with no CFD recalibration.
  • The airflow shaping is reliable enough to serve as a testbed for aerodynamically sensitive robots: the parabolic soaring profile reduced the soaring robot's average position tracking error per step by 50% and its attitude error by 20%.
  • The algorithm remains convergent when an actuator is disabled, so hardware degradation or motor failure does not require re-deriving the model.
  • The same update rule transfers to other airflow sensing modalities, such as a fixed grid of static sensors instead of a moving probe, opening the path to real-time airflow shaping.
  • Because the method is measurement-driven, it captures turbulent and interaction effects that a purely simulated model would miss.

Reading between the lines

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

  • The load-bearing assumption is really condition (8b): the coarse model's pseudo-inverse must stay aligned with the true airflow gradient. If that alignment holds for other geometries, the same recipe could control other distributed fields with few actuators, such as temperature, pressure, or chemical concentration.
  • The ad hoc central-motor term suggests the method can absorb strong unmodeled interactions, but also that the fitted Gaussian shape and polynomials may need refitting when the mechanical layout changes; a direct transfer test across tunnel geometries would settle that.
  • A natural extension the authors do not pursue is online adaptation of the model itself: re-estimating the Gaussian width and the thrust polynomials during the iterations could relax condition (8b) and accelerate convergence on strongly nonlinear targets.
  • The 50% and 20% flight-error reductions are tied to one robot and one profile; the same protocol could be used to benchmark whether other target profiles, such as Gaussian updrafts or shear layers, yield predictable flight behavior.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents an online learning algorithm for shaping the airflow in a seven-fan vertical wind tunnel. The method combines a coarse analytical superposition model (Gaussian jets plus a central-motor correction) with Gaussian-process regression of sparse force measurements and an iterative pseudo-inverse update of the motor commands. Experiments demonstrate convergence, in roughly six to seven iterations, to three target force fields: a parabolic profile for soaring, a Gaussian profile, and a uniform profile, including with the central motor disabled. A flight experiment reports a 50% reduction in average position tracking error and a 20% reduction in average attitude tracking error under the parabolic profile. A conditional linear-convergence result is stated and proved under assumptions (8a)--(8c).

Significance. If the claims hold, the paper provides a practical, sample-efficient way to program arbitrary airflow profiles in a multi-fan testbed, which would be valuable for UAV experimentation. The experimental evidence has notable strengths: direct force measurements rather than simulation, five independent runs per target profile, robustness tests with a disabled motor, random-initial-condition tests, and 20-iteration stability checks. The convergence proof is straightforward and logically sound conditional on its assumptions. However, two load-bearing issues--the unspecified flight-performance baseline and the gap between the proven theorem and the implemented algorithm--currently prevent full validation of the paper's central claims.

major comments (4)
  1. [§V, Experiment 1] The claimed 50% reduction in average position tracking error and 20% reduction in average attitude tracking error are central to the paper's significance, but the baseline and protocol are not specified. Against what airflow profile is the comparison made? What is the flight controller, the number of flights, the duration, and the definition of 'per step'? Without this information the performance improvement cannot be evaluated or reproduced. Please report the experimental protocol, the baseline conditions, and statistics across trials.
  2. [§VI, Algorithm 1 and Prop. 1] Proposition 1 is stated for exact residual evaluations and for unconstrained updates in R^7, but Algorithm 1 uses GP-regressed residuals (Eq. 6) and the experiments clip PWM commands to [0,1] (e.g., the central motor is clipped at zero in Experiment 1). The proof in Appendix II does not account for regression error or projection/clipping. Thus the formal guarantee does not cover the implemented algorithm. Either extend the convergence analysis to the actual update, or state clearly that the theorem applies to the idealized update and that the experimental results are the evidence for the clipped/GP-based version.
  3. [§VI, Eq. (8b)] The convergence guarantee hinges on condition (8b), which requires the coarse-model pseudo-inverse to remain aligned with the true force-map Jacobian and also imposes a step-size condition. No direct verification of (8b) is given: no finite-difference estimates of ∂F/∂U are reported, and the model (Eqs. 3--5), especially the ad hoc central-motor term (Eq. 4), is fitted from a small batch without uncertainty quantification. The three demonstrated target profiles provide indirect evidence but do not establish (8b) for other profiles or hardware configurations. Please either provide empirical verification of the condition (e.g., finite-difference Jacobian checks) or explicitly reframe Prop. 1 as a motivating idealized result rather than a guarantee for the reported experiments.
  4. [§V, Generalization claim] The introduction and conclusion claim that the scheme 'can rapidly converge to a variety of prespecified airflow distributions.' The evidence covers three target fields on one hardware setup. The motor-disabled tests strengthen robustness, but they cover only two of the three profiles. Please temper the wording to match the demonstrated scope, or provide additional distributions (e.g., asymmetric or off-center profiles) to justify 'variety' more fully.
minor comments (5)
  1. [§IV-B, Eq. (4)] The central-motor term includes 0.1 + d_c^2 with no explicit units. Since d_c is a distance, specify units and ensure dimensional consistency with the polynomial force terms.
  2. [§II] Typo: 'UA Vs' should be 'UAVs' in the related-work section.
  3. [§IV-C] The statement that the length-scale of 10 cm is 'half the side length of our force measurement plate' is ambiguous. Clarify whether the plate is square and which side length is meant.
  4. [Appendix I] The polynomial fit is described as explaining the measurements 'well,' but no quantitative fit quality (e.g., R², residual standard deviation) is reported. Adding these values would support the model description.
  5. [§V, Experiments' details] The decaying learning rate λ_k = 0.8/(k+1) is used in all experiments, but the convergence theorem assumes a constant λ. A brief comment on this discrepancy would help the reader connect the theory and practice.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central claim is an experimentally measured convergence result, not a derivation from fitted model parameters.

full rationale

The algorithm's convergence to the prespecified airflow distributions is demonstrated by direct wind-tunnel measurements (Sec. V, Fig. 4), not by construction from the fitted model. The model parameters P_h/l and sigma (Appendix I) are calibrated from separate on-axis force measurements and are not functions of the target distributions F_des; the central-motor term (Eq. 4) is an ad hoc correction for observed edge coupling, not a fit to the targets. The GP regression (Eq. 6) is used only to reconstruct the field at key points from raw sensor data; its kernel length-scale is chosen by cross-validation on representative speed configurations. Proposition 1 is a conditional convergence theorem whose assumptions (8a)-(8c) are sufficient conditions; (8b) in particular requires model-gradient alignment but is not verified, which is a support gap or correctness risk, not circularity, because the theorem does not assume its conclusion and the experiments do not reduce to the model. Self-citations ([25], [28], [29], [31], [32]) are contextual related-work or application references and are not load-bearing for the central claim.

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

The paper introduces no new physical entities. Its free parameters are the fitted coefficients of the coarse model and kernel; they are disclosed but numerous, and the ad hoc central-motor offset is a clear modeling choice.

free parameters (5)
  • P_l and P_h polynomial coefficients = 3 coefficients each (quadratic), values shown in Fig. 6
    Map PWM command to force directly above each motor; fitted to measurements at 5% increments from 0 to 80% PWM.
  • sigma (Gaussian width in motor model) = Not stated numerically, 'found experimentally by fitting'
    Controls spatial extent of each motor's contribution in the coarse model; fit to 'a small batch of measurements'.
  • GP kernel length-scale l = 10 cm
    Chosen as best of three kernels via 5-fold CV on three datasets.
  • Central motor model offset 0.1 in Eq. (4) = 0.1
    Ad hoc constant making the central motor's contribution increase over edge fans as observed; not derived.
  • Learning rate schedule lambda_k = 0.8/(k+1)
    Decaying step size chosen by hand; convergence proof only requires lambda in (0,1/(2 mu c)).
assumptions (5)
  • domain assumption Superposition of individual motor flow contributions (Eq. 5)
    Assumes the total flow is the sum of seven independent Gaussian-like jets, ignoring interactions, swirl, and turbulence.
  • domain assumption Honeycomb straightener eliminates horizontal velocity components
    Justifies measuring only vertical force on a horizontal plate; stated in Section III-A.
  • domain assumption GP regression with squared exponential kernel reconstructs the force field at key points from nearby random measurements
    Used to estimate F at 317 key points without measuring them directly; kernel and length scale chosen empirically.
  • ad hoc to paper Assumptions (8a)-(8c) of Prop. 1 hold for the true flow system
    The convergence proof is conditional on L-smoothness, gradient alignment of the model pseudo-inverse, and Polyak-Lojasiewicz; none are verified experimentally.
  • domain assumption Force plate average represents the local airflow and Eq. (1) relates force to airspeed
    Used to interpret force measurements as airflow; valid only for uniform flow over the plate, which is an approximation in turbulent nonuniform flow.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Shaping Wind-Tunnel Airflow for Unmanned Aerial Vehicles using Online Learning." pith.science (2026). https://pith.science/paper/QE2TPGLJ

@misc{pith2026260803378,
  author       = {Pith},
  title        = {Pith review of: Shaping Wind-Tunnel Airflow for Unmanned Aerial Vehicles using Online Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QE2TPGLJ}},
  note         = {Machine review of arXiv:2608.03378}
}
read the original abstract

The development and testing of advanced aerial robots require experiments in controlled environments with tailored airflow profiles. This paper presents an online learning algorithm for controlling the complex airflow field in a multi-fan vertical wind tunnel. Our method combines a simplified physical model with iterative, measurement-based learning, enabling sample-efficient convergence to desired airflow distributions. We demonstrate the method's versatility by generating complex airflow, such as uniform, Gaussian, and parabolic profiles. Crucially, we show that our algorithm can produce an airflow profile specifically designed for passive soaring, greatly enhancing flight performance of a soaring robot. Variability, practical utility, and robustness of our approach are further highlighted by successful operation with a varying number of fans.

Figures

Figures reproduced from arXiv: 2608.03378 by the authors.

Figure 1
Figure 1. Our algorithm is capable of iteratively adapting the wind tunnel’s [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The figure summarizes the experimental setup used to test the algorithm. Panel (a) shows the core of the wind tunnel with the motors and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Panel (a) shows a flowchart of the algorithm. Panel (b) shows the initial airflow speeds over the wind tunnel when using the same rotational speed [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The figure shows the algorithm’s performance for different desired airflow distributions. The first three columns represent the performance when [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: This figure demonstrates the motors’ PWM commands. Plot (a) shows the commands for a parabolic target distribution. Plot (b) shows the motor [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: This figure shows the polynomial fit, which maps the PWM [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: This figure shows the PWM signals of the motors for three different tests. All three tests have the same uniform target distribution. Panel (a) [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 28 canonical work pages

  1. [1]

    Wind tunnel and hover performance test results for multicopter UAS vehicles,

    C. R. Russell, J. Jung, G. Willink, and B. Glasner, “Wind tunnel and hover performance test results for multicopter UAS vehicles,” American Helicopter Society, no. ARC-E-DAA-TN31096, 2016

  2. [2]

    Wind tunnel test of an unmanned aerial vehicle (UA V),

    C. Jindeog, L. Jangyeon, S. Bongzoo, and K. Samok, “Wind tunnel test of an unmanned aerial vehicle (UA V),”KSME International Journal, vol. 17, no. 5, pp. 776–783, 2003

  3. [3]

    Development of a wind tunnel experimental setup for testing multirotor unmanned aerial vehicles in turbulent conditions,

    J. X. J. Bannwarth, Z. J. Chen, K. A. Stol, B. A. MacDonald, and P. J. Richards, “Development of a wind tunnel experimental setup for testing multirotor unmanned aerial vehicles in turbulent conditions,” IEEE International Conference on Advanced Intelligent Mechatronics, pp. 724–729, 2018

  4. [4]

    Commercial wind-tunnels for UA Vs testing,

    WindShape company, “Commercial wind-tunnels for UA Vs testing,” (Accessed September 15, 2023). [Online]. Available: https://windshape.com/about/

  5. [5]

    Science, technology and the future of small autonomous drones,

    D. Floreano and R. J. Wood, “Science, technology and the future of small autonomous drones,”Nature, vol. 521, no. 7553, pp. 460–466, 2015

  6. [6]

    Lift production in the hovering hummingbird,

    D. R. Warrick, B. W. Tobalske, and D. R. Powers, “Lift production in the hovering hummingbird,”Proceedings of the Royal Society B: Biological Sciences, vol. 276, no. 1674, pp. 3747–3752, 2009

  7. [7]

    Acoustic and aerodynamic design and characterization of a small-scale aeroacoustic wind tunnel,

    E. Sarradj, C. Fritzsche, T. Geyer, and J. Giesler, “Acoustic and aerodynamic design and characterization of a small-scale aeroacoustic wind tunnel,”Applied Acoustics, vol. 70, no. 8, pp. 1073–1080, 2009

  8. [8]

    Design and performance of a small-scale aeroacoustic wind tunnel,

    P. Liu, Y . Xing, H. Guo, and L. Li, “Design and performance of a small-scale aeroacoustic wind tunnel,”Applied Acoustics, vol. 116, pp. 65–69, 2017

Show all 33 references
  1. [9]

    Atmospheric boundary layer sim- ulation in a new open-jet facility at LSU: CFD and experimental investigations,

    A. M. Aly and H. Gol-Zaroudi, “Atmospheric boundary layer sim- ulation in a new open-jet facility at LSU: CFD and experimental investigations,”Measurement, vol. 110, pp. 121–133, 2017

  2. [10]

    Realization of a large-scale turbulence field in a small wind tunnel,

    M. Hideharu, “Realization of a large-scale turbulence field in a small wind tunnel,”Fluid Dynamics Research, vol. 8, no. 1-4, p. 53, 1991

  3. [11]

    Atmospheric turbulence simulation tech- niques with application to flight analysis,

    S.-T. Wang and W. Frost, “Atmospheric turbulence simulation tech- niques with application to flight analysis,” Tech. Rep., 1980

  4. [12]

    P. K. Kundu, I. M. Cohen, and D. R. Dowling,Fluid mechanics, 6th ed. Academic Press, 2015

  5. [13]

    J. D. Anderson,Computational fluid dynamics: the basics with appli- cations. McGraw-Hill, 2002

  6. [14]

    Propeller-wing interaction using rapid computational methods,

    C. Agostinelli, C. B. Allen, C. Liu, G. Ferraro, and A. Rampurawala, “Propeller-wing interaction using rapid computational methods,”AIAA Applied Aerodynamics Conference, p. 2418, 2013

  7. [15]

    Computational study of propeller–wing aerodynamic interaction,

    P. Aref, M. Ghoreyshi, A. Jirasek, M. J. Satchell, and K. Bergeron, “Computational study of propeller–wing aerodynamic interaction,” Aerospace, vol. 5, no. 3, p. 79, 2018

  8. [16]

    Evaluating the longitudinal stability of an UA V using a CFD-6dof model,

    P. Sandoval, P. Cornejo, and F. Tinapp, “Evaluating the longitudinal stability of an UA V using a CFD-6dof model,”Aerospace Science and Technology, vol. 43, pp. 463–470, 2015

  9. [17]

    Physics informed deep learning (part i): Data-driven solutions of nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics informed deep learning (part i): Data-driven solutions of nonlinear partial differential equations,”arXiv preprint arXiv:1711.10561, 2017

  10. [18]

    Integrating physics-based modeling with machine learning: A survey,

    J. Willard, X. Jia, S. Xu, M. Steinbach, and V . Kumar, “Integrating physics-based modeling with machine learning: A survey,”arXiv preprint arXiv:2003.04919, 2020

  11. [19]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” Journal of Computational Physics, vol. 378, pp. 686–707, 2019

  12. [20]

    Physics- informed neural networks (PINNs) for fluid mechanics: A review,

    S. Cai, Z. Mao, Z. Wang, M. Yin, and G. E. Karniadakis, “Physics- informed neural networks (PINNs) for fluid mechanics: A review,” Acta Mechanica Sinica, vol. 37, no. 12, pp. 1727–1738, 2021

  13. [21]

    Deep learning of vortex-induced vibrations,

    M. Raissi, Z. Wang, M. S. Triantafyllou, and G. E. Karniadakis, “Deep learning of vortex-induced vibrations,”Journal of Fluid Mechanics, vol. 861, pp. 119–137, 2019

  14. [22]

    Physics- informed autoencoders for Lyapunov-stable fluid flow prediction,

    N. B. Erichson, M. Muehlebach, and M. W. Mahoney, “Physics- informed autoencoders for Lyapunov-stable fluid flow prediction,” arXiv preprint arXiv:1905.10866, 2019

  15. [23]

    State–space modeling for control based on physics-informed neural networks,

    F. Arnold and R. King, “State–space modeling for control based on physics-informed neural networks,”Engineering Applications of Artificial Intelligence, vol. 101, 104195, 2021

  16. [24]

    Pressing and rubbing: physics-informed features facilitate haptic terrain classi- fication for legged robots,

    L. Ding, P. Xu, Z. Li, R. Zhou, H. Gao, Z. Deng, and G. Liu, “Pressing and rubbing: physics-informed features facilitate haptic terrain classi- fication for legged robots,”IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 5990–5997, 2022

  17. [25]

    Stochastic online op- timization for cyber-physical and robotic systems,

    H. Ma, M. Zeilinger, and M. Muehlebach, “Stochastic online op- timization for cyber-physical and robotic systems,”arXiv preprint arXiv:2404.05318, 2024

  18. [26]

    A simple learning strategy for high-speed quadrocopter multi-flips,

    S. Lupashin, A. Sch ¨ollig, M. Sherback, and R. D’Andrea, “A simple learning strategy for high-speed quadrocopter multi-flips,”IEEE In- ternational Conference on Robotics and Automation, pp. 1642–1648, 2010

  19. [27]

    Nonholonomic yaw control of an underactuated flying robot with model-based reinforcement learning,

    N. O. Lambert, C. Schindler, D. S. Drew, and K. S. J. Pister, “Nonholonomic yaw control of an underactuated flying robot with model-based reinforcement learning,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 455–461, 2020

  20. [28]

    A learning- based iterative control framework for controlling a robot arm with pneumatic artificial muscles,

    H. Ma, D. B ¨uchler, B. Sch ¨olkopf, and M. Muehlebach, “A learning- based iterative control framework for controlling a robot arm with pneumatic artificial muscles,”Robotics: Science and Systems, pp. 1– 10, 2022

  21. [29]

    Data-efficient online learning of ball placement in robot table tennis,

    P. Tobuschat, H. Ma, D. B ¨uchler, B. Sch ¨olkopf, and M. Muehlebach, “Data-efficient online learning of ball placement in robot table tennis,” IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 567–573, 2023

  22. [30]

    A fast and reliable pick-and- place application with a spherical soft robotic arm,

    J. Zughaibi, M. Hofer, and R. D’Andrea, “A fast and reliable pick-and- place application with a spherical soft robotic arm,”IEEE International Conference on Soft Robotics, pp. 599–606, 2021

  23. [31]

    Learning-based parametrized model predictive control for trajectory tracking,

    C. Sferrazza, M. Muehlebach, and R. D’Andrea, “Learning-based parametrized model predictive control for trajectory tracking,”Optimal Control Applications and Methods, vol. 41, no. 6, pp. 2225–2249, 2020

  24. [32]

    Embodied intelligence for sustainable flight: A soaring robot with active morphological control,

    G. Elmkaiel, S. Schmitt, and M. Muehlebach, “Embodied intelligence for sustainable flight: A soaring robot with active morphological control,”npj Robotics, vol. 4, no. 28, pp. 1–12, 2026

  25. [33]

    Gaussian processes for regression,

    C. Williams and C. Rasmussen, “Gaussian processes for regression,” Advances in Neural Information Processing Systems 8, pp. 514–520, 1995

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.