pith. machine review for the scientific record. sign in

arxiv: 2604.19618 · v1 · submitted 2026-04-21 · 💻 cs.RO

Recognition: unknown

Autonomous UAV Pipeline Near-proximity Inspection via Disturbance-Aware Predictive Visual Servoing

Wen Li , Hui Wang , Jinya Su , Cunjia Liu , Wen-Hua Chen , Shihua Li
Authors on Pith no claims yet

Pith reviewed 2026-05-10 01:42 UTC · model grok-4.3

classification 💻 cs.RO
keywords UAV pipeline inspectionvisual servoingmodel predictive controldisturbance estimationquadrotorimage-based controlextended-state Kalman filter
0
0 comments X

The pith

Coupling quadrotor dynamics with image feature kinematics in a predictive controller with disturbance estimation enables reliable autonomous near-proximity pipeline inspection.

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

The paper develops a framework for a quadrotor to inspect pipelines from close range using visual information alone. It builds a single prediction model that links the drone's physical motion to how features move in the camera image, then adds a filter to estimate and cancel out lumped disturbances such as wind. A simple velocity rule keeps forward speed steady while the drone climbs or descends unknown slopes. Real experiments show the system cuts image errors by more than half on straight pipes and succeeds on windy and bent sections where a standard visual-servoing controller loses track.

Core claim

The ESKF-PRE-VMPC framework augments image-based visual servoing model predictive control with an extended-state Kalman filter that predicts image features and estimates disturbances; the estimated disturbances are fed directly into the prediction model, allowing the controller to maintain visual alignment under low-rate images and external forces.

What carries the argument

The unified predictive model that couples quadrotor dynamics with image feature kinematics, extended by extended-state Kalman filtering with image prediction (ESKF-PRE) to supply disturbance estimates to the VMPC optimizer.

If this is right

  • Direct image-space prediction inside the control loop removes the need for separate image-to-world transformations at each step.
  • Disturbance estimates allow the optimizer to reject wind and other forces without explicit wind sensors.
  • Terrain-adaptive vertical velocity is generated on the fly from image measurements alone, eliminating the need for prior slope maps.
  • The same structure completes both straight-pipe and bend-pipe runs while a baseline visual-servoing method fails.

Where Pith is reading between the lines

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

  • The approach could transfer to inspection of other long linear assets such as power lines or rail tracks by changing only the image-feature definition.
  • Combining the filter with a simple onboard map of previously seen pipe sections might reduce reliance on continuous visual contact.
  • The reported open-source hardware changes suggest the method can be reproduced on low-cost platforms for field trials.

Load-bearing premise

The single model that links drone motion to image-feature motion stays accurate enough for prediction even when visual measurements arrive slowly and unmodeled forces are present.

What would settle it

A flight test on a pipeline segment with wind gusts or curvature beyond the reported trials in which the UAV loses visual lock on the pipe or deviates beyond the claimed RMSE bounds.

Figures

Figures reproduced from arXiv: 2604.19618 by Cunjia Liu, Hui Wang, Jinya Su, Shihua Li, Wen-Hua Chen, Wen Li.

Figure 1
Figure 1. Figure 1: Coordinate frames in quadrotor inspection. In this study, bold lowercase letters denote vectors, bold uppercase letters denote matrices, and regular letters denote scalars. For frame-dependent variables, a presuperscript indi￾cates that the quantity is expressed in the corresponding local frame, whereas a postsuperscript indicates that the quantity is expressed in the world frame. For example, Bω denotes t… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of image features and modelling, where blue dashed lines indicate the same virtual plane, and l denotes the corresponding projection on the normalized image plane. Accordingly, the desired image feature for pipeline centering and vertical alignment is given by χd = [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Control framework of the proposed ESKF-PRE-VMPC for UAV autonomous pipeline inspection. The framework integrates image feature extraction and prediction, state and lumped-disturbance estimation, and model predictive control. A downward-facing camera captures pipeline images, from which the pipeline edges and the midline are extracted. To address the low sampling rate of visual measurements, an image featur… view at source ↗
Figure 4
Figure 4. Figure 4: ESKF estimator with image feature prediction. Remark 4: Although the image features are first predicted in an open-loop manner, these predictions are treated as pseudo￾measurements in the ESKF. The final image features used by the MPC are obtained from the ESKF estimates, thereby forming a closed-loop image-feature estimation scheme. Specifically, the measurement-noise covariance Rk in the ESKF is designed… view at source ↗
Figure 5
Figure 5. Figure 5: Pipeline model is composed of eight segments with each 8 meters long. Top: an exemplary real-world pipeline; Bottom: schematic of the 3D pipeline in this work, showing the curvature and slope of each segment. In simulation, the measurements are obtained from the onboard camera and the Gazebo ground truth. To assess the performance of the estimator, zero-mean Gaussian white noise N (0, σ2I) is added to the … view at source ↗
Figure 8
Figure 8. Figure 8: Modified Crazyflie platform with a downward camera with a resolution of 756 × 560. Left: side view; Right: bottom view [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Hardware and communication architecture of the modified Crazyflie platform for real-world experiments [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparative results under wind-free and wind cases. of the original Crazyflie platform, which is barely sufficient to carry components beyond motion capture markers, the propulsion system is modified to increase its payload capa￾bility to at least 6 g. This enhancement enables the onboard integration of a lightweight camera and image-transmission module (as shown in [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 12
Figure 12. Figure 12: Comparative results of ESKF-VMPC and ESKF-PRE-VMPC. servoing model predictive control. By coupling quadrotor dynamics with image feature kinematics, a unified predictive model was established for control-oriented visual servoing. To handle low-rate visual updates, measurement degradation, and environmental uncertainties, an extended-state Kalman filtering scheme with image feature prediction was integrate… view at source ↗
read the original abstract

Reliable pipeline inspection is critical to safe energy transportation, but is constrained by long distances, complex terrain, and risks to human inspectors. Unmanned aerial vehicles provide a flexible sensing platform, yet reliable autonomous inspection remains challenging. This paper presents an autonomous quadrotor near-proximity pipeline inspection framework for three-dimensional scenarios based on image-based visual servoing model predictive control (VMPC). A unified predictive model couples quadrotor dynamics with image feature kinematics, enabling direct image-space prediction within the control loop. To address low-rate visual updates, measurement noise, and environmental uncertainties, an extended-state Kalman filtering scheme with image feature prediction (ESKF-PRE) is developed, and the estimated lumped disturbances are incorporated into the VMPC prediction model, yielding the ESKF-PRE-VMPC framework. A terrain-adaptive velocity design is introduced to maintain the desired cruising speed while generating vertical velocity references over unknown terrain slopes without prior terrain information. The framework is validated in high-fidelity Gazebo simulations and real-world experiments. In real-world tests, the proposed method reduces RMSE by 52.63% and 75.04% in pipeline orientation and lateral deviation in the image, respectively, for straight-pipeline inspection without wind, and successfully completes both wind-disturbance and bend-pipeline tasks where baseline method fails. An open-source nano quadrotor is modified for indoor experimentation.

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 / 1 minor

Summary. The manuscript presents an autonomous quadrotor framework for near-proximity pipeline inspection using image-based visual servoing model predictive control (VMPC). It introduces a unified predictive model coupling quadrotor dynamics with image feature kinematics, an extended-state Kalman filter with image feature prediction (ESKF-PRE) to handle low-rate updates, noise, and lumped disturbances, and a terrain-adaptive velocity design for unknown slopes. The ESKF-PRE-VMPC approach is validated in Gazebo simulations and real-world experiments on a modified open-source nano quadrotor, claiming 52.63% and 75.04% RMSE reductions in pipeline orientation and lateral deviation (straight pipeline, no wind) plus successful completion of wind-disturbance and bend-pipeline tasks where the baseline fails.

Significance. If the performance claims hold with stronger validation, the work advances UAV infrastructure inspection by integrating predictive image-space control with disturbance estimation, addressing practical challenges like low-rate vision and environmental uncertainty. The open-source platform modification is a clear strength for reproducibility.

major comments (2)
  1. [Real-world experiments] Real-world experiments (as summarized in the abstract): the central performance claims rest on end-to-end RMSE reductions and task-completion rates, yet no separate quantitative validation of the unified quadrotor+image-kinematics predictive model is provided (e.g., horizon-wise feature prediction residuals versus ground truth, or ablation isolating the coupling accuracy under low-rate visual updates). This leaves the weakest assumption unexamined and undermines confidence in why ESKF-PRE-VMPC succeeds where the baseline fails.
  2. [Abstract and results reporting] Abstract and results reporting: the 52.63% and 75.04% RMSE figures are given without error bars, number of trials, statistical tests, or implementation details for the baseline method and any post-hoc tuning of filter/controller parameters. This makes the magnitude of improvement difficult to interpret or reproduce.
minor comments (1)
  1. The terrain-adaptive velocity design is described at a high level; adding a short pseudocode or explicit reference to how vertical velocity references are generated from image features would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment point by point below and will incorporate revisions to improve the manuscript's clarity, rigor, and reproducibility.

read point-by-point responses

  1. Referee: [Real-world experiments] Real-world experiments (as summarized in the abstract): the central performance claims rest on end-to-end RMSE reductions and task-completion rates, yet no separate quantitative validation of the unified quadrotor+image-kinematics predictive model is provided (e.g., horizon-wise feature prediction residuals versus ground truth, or ablation isolating the coupling accuracy under low-rate visual updates). This leaves the weakest assumption unexamined and undermines confidence in why ESKF-PRE-VMPC succeeds where the baseline fails.

    Authors: We acknowledge that the manuscript's validation is primarily end-to-end and that isolating the predictive model's accuracy would strengthen the presentation. The unified coupling is central to handling low-rate updates and disturbances, as evidenced by the framework's ability to complete tasks where the baseline fails. In the revised version, we will add quantitative validation of the predictive model, including horizon-wise feature prediction residuals versus ground truth from the real-world data and an ablation isolating the coupling's contribution under low-rate visual updates. These additions will be placed in the results section to directly address the concern. revision: yes

  2. Referee: [Abstract and results reporting] Abstract and results reporting: the 52.63% and 75.04% RMSE figures are given without error bars, number of trials, statistical tests, or implementation details for the baseline method and any post-hoc tuning of filter/controller parameters. This makes the magnitude of improvement difficult to interpret or reproduce.

    Authors: We agree that the current reporting of the RMSE improvements lacks sufficient statistical detail and implementation transparency, which limits interpretability. In the revised manuscript, we will update both the abstract and results section to include error bars on the reported percentages, the number of trials performed, statistical significance tests (e.g., paired t-tests), and expanded details on the baseline implementation along with the tuning process for ESKF-PRE and VMPC parameters. This will enhance reproducibility without altering the core claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation relies on standard quadrotor rigid-body dynamics coupled to image-feature kinematics, an extended-state Kalman filter with prediction (ESKF-PRE), and model-predictive control (VMPC). None of these components are defined in terms of the final performance metrics; the reported RMSE reductions and task-completion results are experimental outcomes, not quantities obtained by re-arranging the same fitted parameters or self-citations. No uniqueness theorems, ansatzes, or load-bearing self-citations are invoked to close the loop. The framework therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Review performed on abstract only; the paper relies on standard quadrotor rigid-body dynamics and pinhole-camera feature kinematics (standard_math), plus the assumption that lumped disturbances can be treated as slowly varying states in the extended Kalman filter (domain_assumption). No free parameters or invented entities are explicitly named in the abstract.

axioms (2)
  • standard math Quadrotor translational and rotational dynamics can be accurately represented by a 6-DOF rigid-body model coupled to image-feature velocity kinematics.
    Invoked when the unified predictive model is introduced in the abstract.
  • domain assumption Lumped disturbances (wind, model mismatch) vary slowly enough to be estimated as an extended state from image measurements.
    Central to the ESKF-PRE design described in the abstract.

pith-pipeline@v0.9.0 · 5561 in / 1489 out tokens · 51596 ms · 2026-05-10T01:42:13.643009+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

33 extracted references · 1 canonical work pages · 1 internal anchor

  1. [1]

    Inspection and maintenance of oil & gas pipelines: a review of policies,

    H. Iqbal, S. Tesfamariam, H. Haider, and R. Sadiq, “Inspection and maintenance of oil & gas pipelines: a review of policies,”Structure and Infrastructure Engineering, vol. 13, no. 6, pp. 794–815, 2017

    2017

  2. [2]

    Ai meets uavs: A survey on ai empowered uav perception systems for precision agriculture,

    J. Su, X. Zhu, S. Li, and W.-H. Chen, “Ai meets uavs: A survey on ai empowered uav perception systems for precision agriculture,” Neurocomputing, vol. 518, pp. 242–270, 2023

    2023

  3. [3]

    A uav-based aircraft surface defect inspection system via external constraints and deep learning,

    Y . Liu, J. Dong, Y . Li, X. Gong, and J. Wang, “A uav-based aircraft surface defect inspection system via external constraints and deep learning,”IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1–15, 2022

    2022

  4. [4]

    A review of gnss-independent uav navigation techniques,

    N. Gyagenda, J. V . Hatilima, H. Roth, and V . Zhmud, “A review of gnss-independent uav navigation techniques,”Robotics and Autonomous Systems, vol. 152, p. 104069, 2022

    2022

  5. [5]

    Visual servo control. i. basic ap- proaches,

    F. Chaumette and S. Hutchinson, “Visual servo control. i. basic ap- proaches,”IEEE Robotics & Automation Magazine, vol. 13, no. 4, pp. 82–90, 2006. 11

    2006

  6. [6]

    What matters in constructing a visual servoing scheme: A review of key issues and solutions,

    F. Xu, H. Wang, and H. Gao, “What matters in constructing a visual servoing scheme: A review of key issues and solutions,”IEEE/ASME Transactions on Mechatronics, 2025

    2025

  7. [7]

    Robust nonlinear model predictive control based visual servoing of quadrotor uavs,

    K. Zhang, Y . Shi, and H. Sheng, “Robust nonlinear model predictive control based visual servoing of quadrotor uavs,”IEEE/ASME Transac- tions on Mechatronics, vol. 26, no. 2, pp. 700–708, 2021

    2021

  8. [8]

    RAPTOR: A Foundation Policy for Quadrotor Control

    J. Eschmann, D. Albani, and G. Loianno, “Raptor: A foundation policy for quadrotor control,”arXiv preprint arXiv:2509.11481, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  9. [9]

    Robust online predictive visual servoing for autonomous landing of a rotor uav,

    L. Yang, X. Wang, Z. Liu, and L. Shen, “Robust online predictive visual servoing for autonomous landing of a rotor uav,”IEEE Transactions on Intelligent Vehicles, vol. 10, no. 4, pp. 2818–2835, 2025

    2025

  10. [10]

    Da-mppi: Disturbance-aware model predictive path integral via active disturbance estimation and compensa- tion,

    H. Zhang, J. Su, J. Yang, and S. Li, “Da-mppi: Disturbance-aware model predictive path integral via active disturbance estimation and compensa- tion,” in2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025, pp. 3752–3759

    2025

  11. [11]

    Robust model predictive control using tubes,

    W. Langson, I. Chryssochoos, S. Rakovi ´c, and D. Q. Mayne, “Robust model predictive control using tubes,”Automatica, vol. 40, no. 1, pp. 125–133, 2004

    2004

  12. [12]

    Tracking control of small-scale helicopters using explicit nonlinear mpc augmented with disturbance observers,

    C. Liu, W.-H. Chen, and J. Andrews, “Tracking control of small-scale helicopters using explicit nonlinear mpc augmented with disturbance observers,”Control Engineering Practice, vol. 20, no. 3, pp. 258–268, 2012

    2012

  13. [13]

    Disturbance-observer-based control and related methods–an overview,

    W.-H. Chen, J. Yang, L. Guo, and S. Li, “Disturbance-observer-based control and related methods–an overview,”IEEE Transactions on Indus- trial Electronics, vol. 63, no. 2, pp. 1083–1095, 2016

    2016

  14. [14]

    Disturbance rejection mpc for tracking of wheeled mobile robot,

    Z. Sun, Y . Xia, L. Dai, K. Liu, and D. Ma, “Disturbance rejection mpc for tracking of wheeled mobile robot,”IEEE/ASME Transactions on Mechatronics, vol. 22, no. 6, pp. 2576–2587, 2017

    2017

  15. [15]

    Design of a prediction-accuracy-enhanced continuous-time mpc for disturbed sys- tems via a disturbance observer,

    J. Yang, W. X. Zheng, S. Li, B. Wu, and M. Cheng, “Design of a prediction-accuracy-enhanced continuous-time mpc for disturbed sys- tems via a disturbance observer,”IEEE Transactions on Industrial Electronics, vol. 62, no. 9, pp. 5807–5816, 2015

    2015

  16. [16]

    Model predictive control using efficient gaussian processes for unknown disturbance inputs,

    J. Graßhoff, G. M ¨annel, H. S. Abbas, and P. Rostalski, “Model predictive control using efficient gaussian processes for unknown disturbance inputs,” in2019 IEEE 58th Conference on Decision and Control (CDC). IEEE, 2019, pp. 2708–2713

    2019

  17. [17]

    Computationally efficient infinite-horizon indefinite model predictive control with distur- bance preview information,

    S. Zhan, W.-H. Chen, T. Steffen, and J. V . Ringwood, “Computationally efficient infinite-horizon indefinite model predictive control with distur- bance preview information,”Automatica, vol. 146, p. 110667, 2022

    2022

  18. [18]

    PRED- MPPI: Disturbance-preview and efficient MPPI for robust quadrotor tracking with hardware validation,

    H. Zhang, J. Ge, J. Su, Y . Pan, J. Yang, W.-H. Chen, and S. Li, “PRED- MPPI: Disturbance-preview and efficient MPPI for robust quadrotor tracking with hardware validation,” in2026 IEEE International Con- ference on Robotics and Automation (ICRA)

  19. [19]

    Nonlinear model predictive dynamic positioning of a remotely operated vehicle with wave disturbance preview,

    K. L. Walker, L.-B. Jordan, and F. Giorgio-Serchi, “Nonlinear model predictive dynamic positioning of a remotely operated vehicle with wave disturbance preview,”The International Journal of Robotics Research, vol. 44, no. 6, pp. 932–951, 2025

    2025

  20. [20]

    Predicting the future state of disturbed lti systems: A solution based on high-order observers,

    A. Castillo and P. Garcia, “Predicting the future state of disturbed lti systems: A solution based on high-order observers,”Automatica, vol. 124, p. 109365, 2021

    2021

  21. [21]

    Autonomous power line inspection with drones via perception-aware mpc,

    J. Xing, G. Cioffi, J. Hidalgo-Carri ´o, and D. Scaramuzza, “Autonomous power line inspection with drones via perception-aware mpc,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 1086–1093

    2023

  22. [22]

    Visual servoing nmpc applied to uavs for photovoltaic array inspection,

    E. Velasco-S ´anchez, L. F. Recalde, B. S. Guevara, J. Varela-Ald´as, F. A. Candelas, S. T. Puente, and D. C. Gandolfo, “Visual servoing nmpc applied to uavs for photovoltaic array inspection,”IEEE Robotics and Automation Letters, 2024

    2024

  23. [23]

    Fast model pre- dictive image-based visual servoing for quadrotors,

    P. Roque, E. Bin, P. Miraldo, and D. V . Dimarogonas, “Fast model pre- dictive image-based visual servoing for quadrotors,” in2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 7566–7572

    2020

  24. [24]

    Disturbance observer based adaptive model predictive control for uncalibrated visual servoing in constrained environments,

    Z. Qiu, S. Hu, and X. Liang, “Disturbance observer based adaptive model predictive control for uncalibrated visual servoing in constrained environments,”ISA transactions, vol. 106, pp. 40–50, 2020

    2020

  25. [25]

    Dr-mpc: Disturbance- resilient model predictive visual servoing control for quadrotor uav pipeline inspect,

    W. Li, J. Su, C. Liu, W.-H. Chen, and S. Li, “Dr-mpc: Disturbance- resilient model predictive visual servoing control for quadrotor uav pipeline inspect,” in2025 IEEE/RSJ International Conference on In- telligent Robots and Systems (IROS), 2025, pp. 20 738–20 745

    2025

  26. [26]

    Crazyflie 2.0 quadrotor as a platform for research and education in robotics and control engineering,

    W. Giernacki, M. Skwierczy ´nski, W. Witwicki, P. Wro ´nski, and P. Kozierski, “Crazyflie 2.0 quadrotor as a platform for research and education in robotics and control engineering,” in2017 22nd Interna- tional Conference on Methods and Models in Automation and Robotics (MMAR), 2017, pp. 37–42

    2017

  27. [27]

    Multirotor aerial vehicles: Model- ing, estimation, and control of quadrotor,

    R. Mahony, V . Kumar, and P. Corke, “Multirotor aerial vehicles: Model- ing, estimation, and control of quadrotor,”IEEE Robotics & Automation Magazine, vol. 19, no. 3, pp. 20–32, 2012

    2012

  28. [28]

    R. I. Hartley and A. Zisserman,Multiple View Geometry in Computer Vision, 2nd ed. Cambridge University Press, ISBN: 0521540518, 2004

    2004

  29. [29]

    On extended state based kalman filter design for a class of nonlinear time-varying uncertain systems,

    W. Bai, W. Xue, Y . Huang, and H. Fang, “On extended state based kalman filter design for a class of nonlinear time-varying uncertain systems,”Science China Information Sciences, vol. 61, pp. 1–16, 2018

    2018

  30. [30]

    Dynamic visual servoing with image moments for a quadrotor using a virtual spring approach,

    R. Ozawa and F. Chaumette, “Dynamic visual servoing with image moments for a quadrotor using a virtual spring approach,” in2011 IEEE International Conference on Robotics and Automation, 2011, pp. 5670– 5676

    2011

  31. [31]

    Acado toolkit–an open-source framework for automatic control and dynamic optimization,

    B. Houska, H. J. Ferreau, and M. Diehl, “Acado toolkit–an open-source framework for automatic control and dynamic optimization,”Optimal Control Applications and Methods, vol. 32, no. 3, pp. 298–312, 2011

    2011

  32. [32]

    qpOASES: A parametric active-set algorithm for quadratic program- ming,

    H. J. Ferreau, C. Kirches, A. Potschka, H. G. Bock, and M. Diehl, “qpOASES: A parametric active-set algorithm for quadratic program- ming,”Mathematical Programming Computation, vol. 6, pp. 327–363, 2014

    2014

  33. [33]

    An overview of null space projections for redundant, torque-controlled robots,

    A. Dietrich, C. Ott, and A. Albu-Sch ¨affer, “An overview of null space projections for redundant, torque-controlled robots,”The International Journal of Robotics Research, vol. 34, no. 11, pp. 1385–1400, 2015

    2015