Pith. sign in

REVIEW 3 major objections 6 minor 39 references

Validating Terrain Models in Digital Twins for Trustworthy sUAS Operations

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A three-dimensional validation framework for terrain-aware digital shadows traces the largest real-world drone geolocation errors to altitude mismatch between the vehicle and the terrain model, not to the terrain data itself.

desk verdict A solid framework paper undermined by using the DEM as its own ground truth for the key altitude-error claims. read the letter →

arxiv 2508.16104 v1 pith:OH45RLSA submitted 2025-08-22 cs.SE cs.RO

Arturo Miguel Russell Bernal , Maureen Petterson , Pedro Antonio Alarcon Granadeno , Michael Murphy , James Mason , Jane Cleland-Huang This is my paper
classification cs.SEcs.RO
keywords terrain-awaredigitalshadowsUASvalidationtwingeolocationaccuracyaltitudemismatchsimulation-to-real-worldtestingterrainmodel
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 tries to establish that terrain models for small drone operations need dedicated validation, and that a three-dimensional testing framework can provide it. The framework organizes tests by level (unit through acceptance), fidelity (software-in-the-loop, hardware-in-the-loop, live flight), and functional/environmental complexity, with every use-case step mapped to one of eight named operational challenges. Demonstrated on a Terrain-Aware Digital Shadow built from USGS elevation, hydrography, land cover, and satellite imagery, the approach produced field measurements of about 1.5 m average and 4.2 m maximum horizontal geolocation error, with elevation error below 1 m. The paper's key observed failure mode is altitude mismatch: when a drone's altitude belief disagrees with the terrain model's elevation, geolocation errors exceed 10 m even though the terrain model returns the correct intersection point for the ray it is given. The reason to care is that this reframes where trust should be placed: an accurate terrain model does not guarantee trustworthy geolocation unless the drone's own altitude and attitude estimates are validated against it.

What carries the argument

The load-bearing mechanism is the Terrain-Aware Digital Shadow (TDS): a unidirectional digital representation built by fusing USGS DEM, hydrography, land-cover, and transportation datasets with satellite imagery segmented by a computer-vision model, and organized in STRTree-indexed cells that can be queried by latitude/longitude. The validation mechanism is the three-dimensional framework, which tests the TDS across test level (unit, integration, system, acceptance), operational fidelity (software-in-the-loop, hardware-in-the-loop, real-world), and functional/environmental complexity, with each test step mapped to challenges C1–C8. The framework's diagnostic power comes from the mapping: whe

What would settle it

Set up a surveyed ground-control point at the FARM-FLAT and FARM-GULLY sites, compare the TDS elevation and the sUAS barometric/GPS altitude against the surveyed value during a hover, then re-run the geolocation test. If the DEM matches the survey and the drone altitude does not, the paper's altitude-mismatch explanation is confirmed; if the DEM deviates as much as the drone, the dominant error is terrain-model bias, not vehicle altitude error.

Watch

Extended reading notes

Core claim

The paper establishes that a terrain model for sUAS operations can be validated as an engineered system by running it through three orthogonal dimensions—test granularity, realism fidelity, and functional/environmental complexity—with each use-case step explicitly mapped to one or more of eight operational challenges (C1–C8). Applied to a Terrain-Aware Digital Shadow that fuses USGS DEMs, hydrography, land cover, and satellite-image segmentation, the framework produced repeatable numbers: average horizontal geolocation error of about 1.5 m, maximum about 4.2 m, and elevation error below 1 m across flat and gully sites. The decisive observation is an altitude-mismatch failure mode: when the s

Load-bearing premise

The paper assumes the USGS-derived elevation stored in the terrain model is the true altitude at the test sites; if that elevation is wrong, what it calls drone altitude error could actually be terrain-model error.

Editorial extensions

If this is right

  • If the framework is adopted, validation plans for terrain-aware autonomy can be specified as a matrix of test level, fidelity, and complexity, with each step tied to named operational challenges.
  • The altitude-mismatch failure mode implies flight systems should cross-check sUAS altitude readings against terrain-model elevation before trusting geolocation outputs.
  • Simulation-only validation would have missed the real-world GPS, gimbal, and altitude interactions; the hardware-in-the-loop stage is necessary to catch integration issues.
  • The measured 1.5 m average and 4.2 m maximum horizontal error establishes a field baseline for geolocation accuracy that future terrain-model improvements can be measured against.
  • The challenge-to-test-step mapping supports cyclic testing: field failures feed back into new unit tests in software-in-the-loop or hardware-in-the-loop environments.

Reading between the lines

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

  • The paper does not separate DEM bias from drone sensor error because it treats TDS elevation as truth; a natural extension is to survey the test sites with RTK ground control and re-run the same tests to decompose the altitude mismatch.
  • Because each test step is mapped to a named challenge, the same structure could generate runtime confidence intervals for geolocation by propagating current altitude error through the ray-casting math, giving the human operator a live trust signal rather than a post-test explanation.
  • The three-dimensional organization appears transferable to other environmental digital twin components, such as weather and airspace models, by redefining the challenge set while keeping the test-level, fidelity, and complexity dimensions.
  • The acceptance-test dimension is explicitly not yet executed; the framework's own logic predicts that human-in-the-loop studies of uncertainty visualization will be the next binding constraint for operational trust.
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

3 major / 6 minor

Summary. The paper proposes a three-dimensional validation framework for terrain models used in small uncrewed aircraft system (sUAS) digital twins. Dimension D1 covers test levels (unit, integration, system, acceptance), D2 covers operational fidelity (software-in-the-loop, hardware-in-the-loop, real-world), and D3 covers functional and environmental complexity. The authors construct a Terrain-Aware Digital Shadow (TDS) by fusing USGS elevation, hydrography, land-cover, and transportation data with satellite imagery and a semantic segmentation model. They identify eight operational challenges (C1–C8), then apply the framework to a geolocation use case in which a sUAS detects a person and computes the person's latitude, longitude, and altitude by ray-casting into the TDS. Field tests at two farm locations and one ground-based site reveal altitude mismatches between the sUAS and the TDS, geolocation errors of roughly 1.5 m average and 4.2 m maximum, and errors over 10 m under poor GPS conditions, which the paper attributes primarily to sUAS altitude error. The paper's central claim is that this framework enables systematic detection and analysis of real-world failures that simulation alone would miss.

Significance. The proposed framework is a useful organizational contribution: it ties test levels, fidelity, and complexity into a single workflow, explicitly maps test steps to named challenges (C1–C8), and demonstrates the value of HIL testing before field deployment. The paper also provides a concrete reproducibility artifact in the unit-test example of Table I, and it is unusually honest about its limitations, explicitly stating that acceptance tests were not conducted and that only two drone models and hilly (not mountainous) terrain were used. If the quantitative claims were properly supported, the paper would offer a credible template for terrain-model validation in sUAS operations. However, the empirical evidence as presented is not yet strong enough to support the paper's central validation claims, because the reported errors are not measured against independent ground truth and the altitude-error attribution is confounded with DEM error.

major comments (3)
  1. [Section V-D, Figs. 10–11] The quantitative claims—'average latitude–longitude error across all experiments was approximately 1.5m, with a maximum of around 4.2m' and 'elevation error remained below 1m'—are not backed by a defined error metric, number of trials, variance, confidence intervals, or a description of how ground-truth coordinates were obtained. More importantly, the 'elevation error' is not a terrain-validation metric: the returned geolocation altitude is read from the same TDS that is used for ray casting, so an error below 1m only shows internal consistency of the projection algorithm, not agreement with real terrain. The paper should either provide independent ground-truth measurements (e.g., surveyed markers or RTK GNSS) for both position and elevation, or clearly rephrase these numbers as algorithmic self-consistency checks rather than validation results.
  2. [Section V-D, Figs. 9–11] The central causal explanation—that sUAS altitude error causes the observed geolocation offsets—is confounded because the TDS elevation (274–275 m amsl at FARM-FLAT) is treated as the reference truth. The observed 5–8 m discrepancies between sUAS altitude readings and the TDS elevation could equally be explained by vertical error in the underlying USGS DEM, by GPS altitude error, or by a combination of both. No independent surveyed elevation at the actual takeoff, hover, or stare points is reported. To support the attribution to sUAS error, the authors need an independent elevation reference (e.g., a total station, RTK base station, or surveyed benchmark) at each test location, or at minimum a sensitivity analysis that varies the DEM error within its published RMSE bounds.
  3. [Sections IV and VI] The paper's title and contribution claim to validate terrain models, but the empirical demonstration is an exploratory case study of one geolocation use case, with no acceptance tests (explicitly stated in Section V-C) and only two drone platforms in non-mountainous terrain. The framework itself may be sound and the lessons learned valuable, but the evidence presented does not yet validate the terrain model's accuracy against the physical world. The authors should either add validation results with independent ground truth and broader coverage, or reframe the contribution as a validation framework with a preliminary case study and clearly separate framework-level claims from measured accuracy claims.
minor comments (6)
  1. [Section II-B] Typo: 'retreived' should be 'retrieved'.
  2. [Section III-B, C6] 'field-of-view incurabilities' appears to be a typo; presumably 'inaccuracies' or 'limitations'.
  3. [Section V-C] Typo in 'HIL to Real-Wold shift' (should be 'Real-World').
  4. [Section V-D] Typo: 'above mean seal level' should be 'above mean sea level'.
  5. [Figure 8] The figure shows GPS uncertainty of roughly 4 m and 2.5 m on the ground and 1 m and 0.3 m in the air, but no explanation of how these uncertainties are computed. Adding a sentence on the error metric would help.
  6. [Section IV-B] The three example scenarios are introduced but only the first is elaborated in the use case; the other two are never revisited. Consider removing them or connecting them to future work.

Circularity Check

1 steps flagged · score 4.0 of 10

Self-referential altitude reference: the TDS under validation is used as the ground-truth elevation, so the sUAS-altitude-error explanation and the <1m elevation-error claim are defined relative to the model itself.

  1. self definitional [Section V-D (Results and Discussion), Figs. 9-11]
    "where elevation readings were expected to be 274-275m above mean sea level (amsl) according to the TDS. Both tests showed persistent altitude deviations from the TDS and from each other... In contrast, elevation error remained below 1m, even in FARM-GULLY tests where elevation ranged from 268m to 274m above mean sea level."

    The expected elevation at the test site is defined by the TDS ('according to the TDS'), so the sUAS 'altitude deviation' is measured relative to the model under test. The geolocation 'elevation error' likewise compares a TDS ray-cast intersection altitude with a target LLA altitude that is also a TDS value. No surveyed ground-truth elevation is reported. Consequently the central field attribution (sUAS altitude error causes geolocation offsets, Figs. 10-11) and the 'elevation error <1m' result are consistency checks against the TDS, not independent validation of the terrain model; the TDS is assumed correct by construction in the error definition.

full rationale

The paper's main contribution is a three-dimensional validation framework; no model parameters are fitted, no prediction is generated from a fitted input, and the framework's organization is not derived from the measurements. Thus the core claim is not circular. The field demonstration, however, contains a self-referential altitude reference: the TDS, constructed from USGS DEMs, is the object under validation, yet its own elevation is used as the expected value when measuring sUAS altitude deviations and geolocation elevation error. This confounds sUAS altitude error with terrain-model error and weakens the specific causal explanation in Figs. 10-11. It does not collapse the entire derivation because the lat-lon geolocation offsets and the framework's structure have independent content. Self-citations to the authors' prior DroneResponse and NOMAD work are descriptive rather than load-bearing. Overall score 4: partial self-reference in the key quantitative analysis, but no fitted-input-as-prediction or self-citation chain.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The paper does not fit any parameter to make its claim work, but it rests on the accuracy of external data sources (USGS DEM), on a CV segmentation model with acknowledged false positives, and on an unstated ground-truth measurement protocol. The centroid-based nearest-neighbor query is an acknowledged approximation. No new physical entities are introduced; the TDS is an implemented system, not a postulated entity.

free parameters (1)
  • grid_cell_size = not stated
    The terrain model resolution is 'adjustable through the grid cell size' (Section II-B); results depend on this setting but the value used in field tests is not reported. It is a hand-chosen parameter, not fitted to data.
assumptions (6)
  • domain assumption USGS DEM elevation data is accurate enough to serve as reference truth for terrain elevation at the test sites.
    Fig 9 and Fig 10 treat TDS elevation (274-275m amsl) as the expected value when measuring sUAS altitude error; if the DEM is biased, the altitude mismatch is partly a model error. This assumption is load-bearing for the main observed failure mode.
  • standard math The pinhole camera projection and ray-casting model correctly map image pixels to rays through the terrain.
    Section III-C and use case step 5 rely on extracting a ray from camera to ground and intersecting with TDS; standard geometric optics, not independently verified in this paper.
  • domain assumption The DeepLabv3 model fine-tuned on LandCover.AI provides sufficiently accurate semantic segmentation for buildings, woodlands, water, and roads.
    Section II-B merges CV segmentations into the TDS; the paper notes water false positives due to shading, so this assumption is partially acknowledged.
  • domain assumption The STRTree centroid nearest-neighbor query returns the correct elevation for a given lat/lon.
    Section II-B: the paper itself notes 'current approach is based on centroid-based queries' and future enhancements may use radius-based searches; at coarse resolution this can misassign elevation.
  • domain assumption Ground-truth coordinates of the detected person (the 'actual coordinates' in the success criterion) are accurate.
    The use case success criterion compares computed LLA to 'actual coordinates', but the paper never describes how those were measured; this assumption is unstated and load-bearing for all reported geolocation errors.
  • standard math WGS84 coordinate reference and Haversine distance are appropriate for computing lat-lon error.
    Used in Section V-D to quantify geolocation error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Validating Terrain Models in Digital Twins for Trustworthy sUAS Operations." pith.science (2026). https://pith.science/paper/OH45RLSA

@misc{pith2026250816104,
  author       = {Pith},
  title        = {Pith review of: Validating Terrain Models in Digital Twins for Trustworthy sUAS Operations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OH45RLSA}},
  note         = {Machine review of arXiv:2508.16104}
}
read the original abstract

With the increasing deployment of small Unmanned Aircraft Systems (sUAS) in unfamiliar and complex environments, Environmental Digital Twins (EDT) that comprise weather, airspace, and terrain data are critical for safe flight planning and for maintaining appropriate altitudes during search and surveillance operations. With the expansion of sUAS capabilities through edge and cloud computing, accurate EDT are also vital for advanced sUAS capabilities, like geolocation. However, real-world sUAS deployment introduces significant sources of uncertainty, necessitating a robust validation process for EDT components. This paper focuses on the validation of terrain models, one of the key components of an EDT, for real-world sUAS tasks. These models are constructed by fusing U.S. Geological Survey (USGS) datasets and satellite imagery, incorporating high-resolution environmental data to support mission tasks. Validating both the terrain models and their operational use by sUAS under real-world conditions presents significant challenges, including limited data granularity, terrain discontinuities, GPS and sensor inaccuracies, visual detection uncertainties, as well as onboard resources and timing constraints. We propose a 3-Dimensions validation process grounded in software engineering principles, following a workflow across granularity of tests, simulation to real world, and the analysis of simple to edge conditions. We demonstrate our approach using a multi-sUAS platform equipped with a Terrain-Aware Digital Shadow.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 33 canonical work pages

  1. [1]

    Annual nlcd (national land cover database) — the next generation of land cover mapping,

    U.S. Geological Survey, Earth Resources Observation and Science (EROS) Center, “Annual nlcd (national land cover database) — the next generation of land cover mapping,” U.S. Geological Survey, Fact Sheet 2025–3001, 2025, first release (Collection 1.0) covers land cover and change from 1985 to 2023 at 30m resolution across CONUS. Accessed June 25, 2025. [O...

    work page 2025

  2. [2]

    Nhdplus high resolution (nhdplus hr),

    U.S. Geological Survey, “Nhdplus high resolution (nhdplus hr),” 2023, accessed June 25, 2025. [Online]. Available: https://www.usgs.gov/nat ional-hydrography/nhdplus-high-resolution

    work page 2023

  3. [3]

    Land-coverage aware path-planning for multi-uav swarms in search and rescue scenarios,

    P. A. Alarcon Granadeno and J. Cleland-Huang, “Land-coverage aware path-planning for multi-uav swarms in search and rescue scenarios,”

    work page

  4. [4]

    On the notion of digital twins: A modeling perspective,

    B. Tekinerdogan, “On the notion of digital twins: A modeling perspective,” Syst., vol. 11, no. 1, p. 15, 2023. [Online]. Available: https://doi.org/10.3390/systems11010015

    work page doi:10.3390/systems11010015 2023

  5. [5]

    The 3d elevation program and energy for the nation,

    C. Thatcher, V . Lukas, and J. Stoker, “The 3d elevation program and energy for the nation,” U.S. Geological Survey, Fact Sheet 2019–3051,

    work page 2019

  6. [6]

    The national map - transportation datasets,

    U.S. Geological Survey, “The national map - transportation datasets,” U.S. Geological Survey, National Geospatial Program, 2023, vector data for roads, trails, and transportation features. Data available from 2014 to present. Accessed June 25, 2025. [Online]. Available: https://data.usg s.gov/datacatalog/data/USGS:ad3d631d-f51f-4b6a-91a3-e617d6a58b4e

    work page 2023

  7. [7]

    Hyriver: Hydroclimate data retriever,

    T. Chegini, H. Li, B. Lehner, and C. H. David, “Hyriver: Hydroclimate data retriever,” Journal of Open Source Software , vol. 6, no. 66, p. 3175, 2021. [Online]. Available: https://doi.org/10.21105/joss.03175

    doi:10.21105/joss.03175 2021

  8. [8]

    STR: A simple and efficient algorithm for R-tree packing,

    S. T. Leutenegger, J. M. Edgington, and M. A. Lopez, “STR: A simple and efficient algorithm for R-tree packing,” in Proceedings of the 13th International Conference on Data Engineering (ICDE) . IEEE, Feb. 1997, pp. 497–506. [Online]. Available: https://doi.org/10.1109/ICDE.1 997.581888

    work page doi:10.1109/icde.1 1997

Show all 39 references
  1. [9]

    GEOS Computational Geometry Library,

    GEOS contributors, “GEOS Computational Geometry Library,” https: //libgeos.org/, 2024, open Source Geospatial Foundation

    2024

  2. [10]

    JTS Topology Suite,

    Martin Davis, “JTS Topology Suite,” https://www.tsusiatsoftware.net/jt s/main.html, 2018, jTS Topology Suite

    2018

  3. [11]

    JTS Topology Suite,

    LocationTech, “JTS Topology Suite,” https://locationtech.github.io/jts/, 2017, jTS Topology Suite

    2017

  4. [12]

    Mapbox api documentation,

    Mapbox, “Mapbox api documentation,” https://docs.mapbox.com/api/ov erview/, 2025, accessed: 2025-02-15

    2025

  5. [13]

    Rethinking atrous convolution for semantic image segmentation,

    L.-C. Chen, G. Papandreou, F. Schroff, and H. Adam, “Rethinking atrous convolution for semantic image segmentation,” 2017

    2017

  6. [14]

    Landcover.ai: Dataset for automatic mapping of build- ings, woodlands, water and roads from aerial imagery,

    A. Boguszewski, D. Batorski, N. Ziemba-Jankowska, T. Dziedzic, and A. Zambrzycka, “Landcover.ai: Dataset for automatic mapping of build- ings, woodlands, water and roads from aerial imagery,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (...

    2021

  7. [15]

    NOMAD: A natural, occluded, multi-scale aerial dataset, for emergency response scenarios,

    A. M. Russell Bernal, W. J. Scheirer, and J. Cleland-Huang, “NOMAD: A natural, occluded, multi-scale aerial dataset, for emergency response scenarios,” in IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2024, Waikoloa, HI, USA, January 3-8, 2024 . IEEE, 202...

    2024

  8. [16]

    The next generation of human-drone partnerships: Co-designing an emer- gency response system,

    A. Agrawal, S. J. Abraham, B. Burger, C. Christine, L. Fraser, J. M. Hoeksema, S. Hwang, E. Travnik, S. Kumar, W. Scheirer et al. , “The next generation of human-drone partnerships: Co-designing an emer- gency response system,” in Proceedings of the 2020 CHI Conference on Huma...

    2020

  9. [17]

    Nav- igating the black box: Operational lenses for AI-enabled drone governance,

    D. Hernandez, K. Harris, T. Hernandez, and R. Morales, “Nav- igating the black box: Operational lenses for AI-enabled drone governance,” MIT Science Policy Review , vol. 6, aug 5 2025, https://sciencepolicyreview.pubpub.org/pub/yc0xe805

    2025

  10. [18]

    Controlling software projects,

    P. Rook, “Controlling software projects,” Software engineering journal , vol. 1, no. 1, pp. 7–16, 1986

    1986

  11. [19]

    Human–machine teaming with small unmanned aerial systems in a mape-k environment,

    J. Cleland-Huang, T. Chambers, S. Zudaire, M. T. Chowdhury, A. Agrawal, and M. Vierhauser, “Human–machine teaming with small unmanned aerial systems in a mape-k environment,” ACM Transactions on Autonomous and Adaptive Systems , vol. 19, no. 1, pp. 1–35, 2024

    2024

  12. [20]

    Self-adaptation of loosely coupled systems across a system of small uncrewed aerial sys- tems,

    T. Chambers, J. Cleland-Huang, and M. Vierhauser, “Self-adaptation of loosely coupled systems across a system of small uncrewed aerial sys- tems,” in Proceedings of the 12th ACM/IEEE International Workshop on Software Engineering for Systems-of-Systems and Software Ecosystems ...

    2024

  13. [21]

    Cognitive guardrails for open-world decision making in autonomous drone swarms,

    J. Cleland-Huang, P. A. A. Granadeno, A. M. Russell Bernal, D. Hernan- dez, M. Murphy, M. Petterson, and W. Scheirer, “Cognitive guardrails for open-world decision making in autonomous drone swarms,” arXiv preprint arXiv:2505.23576, 2025

    2025 arXiv

  14. [22]

    An environmentally complex requirement for safe separation distance between uavs,

    P. Alarcon Granadeno, A. M. Russell Bernal, M. N. A. Islam, and J. Cleland-Huang, “An environmentally complex requirement for safe separation distance between uavs,” in 32nd IEEE International Requirements Engineering Conference, RE 2024 - Workshops, Reykjavik, Iceland, June 2...

    2024

  15. [23]

    Open Robotics, “Gazebo,” https://gazebosim.org/home, 2025, gazebo

    2025

  16. [24]

    Psych- occlusion: Using visual psychophysics for aerial detection of occluded persons during search and rescue,

    A. M. Russell Bernal, J. Cleland-Huang, and W. J. Scheirer, “Psych- occlusion: Using visual psychophysics for aerial detection of occluded persons during search and rescue,” in 2025 IEEE/CVF Winter Confer- ence on Applications of Computer Vision (WACV) . IEEE, 2025, pp. 3383–3395

    2025

  17. [25]

    Hierarchically organized computer vision in support of multi-faceted search for missing persons,

    Arturo Miguel Russell Bernal and Jane Cleland-Huang, “Hierarchically organized computer vision in support of multi-faceted search for missing persons,” in 2023 IEEE 17th International Conference on Automatic Face and Gesture Recognition (FG) . IEEE, 2023, pp. 1–7

    2023

  18. [26]

    Region proposal approach for human detection on aerial imagery,

    ˇZ. Maru ˇsi´c, D. Bo ˇzi´c-ˇStuli´c, S. Gotovac, and T. Maru ˇsi´c, “Region proposal approach for human detection on aerial imagery,” in 2018 3rd International Conference on Smart and Sustainable Technologies (SpliTech). IEEE, 2018, pp. 1–6

    2018

  19. [27]

    Deep learning approach in aerial imagery for supporting land search and rescue missions,

    D. Bo ˇzi´c-ˇStuli´c, ˇZ. Maru ˇsi´c, and S. Gotovac, “Deep learning approach in aerial imagery for supporting land search and rescue missions,” International Journal of Computer Vision , vol. 127, no. 9, pp. 1256– 1278, 2019

    2019

  20. [28]

    Automatic person detection in search and rescue operations using deep cnn detectors,

    S. Sambolek and M. Ivasic-Kos, “Automatic person detection in search and rescue operations using deep cnn detectors,” Ieee Access, vol. 9, pp. 37 905–37 922, 2021

    2021

  21. [29]

    Ultralytics yolo11,

    G. Jocher and J. Qiu, “Ultralytics yolo11,” 2024. [Online]. Available: https://github.com/ultralytics/ultralytics

    2024

  22. [30]

    Van Brummelen, Heavenly mathematics: The forgotten art of spher- ical trigonometry

    G. Van Brummelen, Heavenly mathematics: The forgotten art of spher- ical trigonometry . Princeton University Press, 2017

    2017

  23. [31]

    Autonomous driving validation and verification using digital twins

    H. Pikner, M. Malayjerdi, M. Bellone, B. C. Baykara, and R. Sell, “Autonomous driving validation and verification using digital twins.” VEHITS, pp. 204–211, 2024

    2024

  24. [32]

    Verification and validation for a digital twin for augmenting current sora practices with air-to-air collision hazards prediction from small uncooperative flying objects,

    S. Matalonga, J. Black, and J. Riordan, “Verification and validation for a digital twin for augmenting current sora practices with air-to-air collision hazards prediction from small uncooperative flying objects,” Journal of Intelligent & Robotic Systems , vol. 110, no. 3, p. 109, 2024

    2024

  25. [33]

    Design of virtual- real cooperative flight testbed for multiple unmanned aerial vehicles,

    Y . Wang, G. Xu, Y . Hu, T. Long, J. Sun, and J. Li, “Design of virtual- real cooperative flight testbed for multiple unmanned aerial vehicles,” in Advances in Guidance, Navigation and Control. ICGNC 2022 , ser. Lecture Notes in Electrical Engineering, L. Yan, H. Duan, and Y ....

    2022

  26. [34]

    A step-by-step guide to creating a robust autonomous drone testing pipeline,

    Y . Jiang, Y . Deng, S. Schroder, L. Liang, S. Gambhir, A. James, A. Seth, J. Pirrie, Y . Zhang, and X. Zheng, “A step-by-step guide to creating a robust autonomous drone testing pipeline,” 2025. [Online]. Available: https://arxiv.org/abs/2506.11400

    2025 arXiv

  27. [35]

    Testing and validation framework for autonomous aerial vehicles,

    M. Akbas, “Testing and validation framework for autonomous aerial vehicles,” Journal of Aviation/Aerospace Education & Research , 01 2021

    2021

  28. [36]

    Generating edge cases for testing autonomous vehicles using real-world data,

    D. Karunakaran, J. S. Berrio Perez, and S. Worrall, “Generating edge cases for testing autonomous vehicles using real-world data,” Sensors, vol. 24, no. 1, 2024. [Online]. Available: https://www.mdpi.com/142 4-8220/24/1/108

    2024

  29. [37]

    An open-source uav digital twin framework: A case study on remote sensing in the andean mountains,

    E. Valencia, F. Toapanta, G. O ˜na et al., “An open-source uav digital twin framework: A case study on remote sensing in the andean mountains,” Journal of Intelligent & Robotic Systems , vol. 111, p. 71, 2025

    2025

  30. [2020]

    Available: https://doi.org/10.3133/fs20193051

    [Online]. Available: https://doi.org/10.3133/fs20193051

    doi

  31. [2025]

    Available: https://arxiv.org/abs/2505.08060

    [Online]. Available: https://arxiv.org/abs/2505.08060

    arXiv

Pith tools

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