Pith. sign in

REVIEW 4 major objections 5 minor 58 references

Demonstrating ViSafe: Vision-enabled Safety for High-speed Detect and Avoid

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Control barrier functions give vision-only drones a provable safety layer, and ViSafe demonstrated it in flight at closure rates up to 144 km/h.

desk verdict Real engineering milestone—first vision-only DAA at 144 km/h closure—but the 'provably safe' claim does not survive contact with the paper's own data and admitted perception gaps. read the letter →

arxiv 2505.03694 v2 pith:VDBEABRU submitted 2025-05-06 cs.RO cs.AI

Parv Kapoor , Ian Higgins , Nikhil Keetha , Jay Patrikar , Brady Moon , Zelin Ye , Yao He , Ivan Cisneros
show 4 more authors
Yaoyu Hu Changliu Liu Eunsuk Kang Sebastian Scherer
This is my paper · ORCID
classification cs.ROcs.AI
keywords detectandavoidcontrolbarrierfunctionsvision-onlycollisionavoidancemulti-cameratrackingdigitaltwinhardware-in-the-loopsmalluncrewedaircraftnon-cooperativeintruder
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

ViSafe is a full-stack, vision-only Detect and Avoid system for small uncrewed aircraft that cannot carry radar, transponders, or ADS-B. The paper's central claim is that control barrier functions (CBFs) can encode the separation requirement and, when fed by a multi-camera deep-learning tracker, can give provably safe runtime guarantees for self-separation in high-speed encounters. The authors support the claim with a hardware-in-the-loop digital-twin campaign of roughly 12,000 simulated trajectories across varied weather, lighting, and encounter geometry, plus real-world flights with closure rates up to 144 km/h. They are also explicit that the formal guarantee assumes always-accurate global perception, which real vision does not supply, so the deployed system's demonstrated benefit is empirical risk reduction rather than an unconditional proof.

What carries the argument

The load-bearing object is the control barrier function $h(x) = c + d_{\mathrm{thresh}}^n - d^n - k\dot d$, with the safe set defined as $\{x : h(x) \le 0\}$. The CBF condition $\dot h(x) \le -\lambda h(x)$ is added to a quadratic program whose objective is to stay as close as possible to the nominal control input, subject to the vehicle's actuation limits. This turns the safety specification into a runtime filter that only modifies the command when the barrier is about to be violated. The rest of the stack, including the multi-camera tracker, coordinate-frame conversion, decoupled angle and range Kalman filters, and the detection-range reaction profile, exists to feed this QP with the best available state estimates and to decide when a track can be trusted.

What would settle it

Run the digital-twin below-horizon scenarios with the intruder starting outside the reliable detection range; the paper's own table shows thousands of separation violations in that regime, so a claim of unconditional runtime safety would fail exactly there.

Watch

Extended reading notes

Core claim

The paper's central discovery, stated in its own terms, is that a perceptual input-focused control barrier function can turn imperfect monocular vision into a usable safety layer for high-speed airborne collision avoidance. The CBF takes the form $h(x) = c + d_{\mathrm{thresh}}^n - d^n - k\dot d$, and the condition $\dot h(x) \le -\lambda h(x)$ is inserted as a constraint in a quadratic program that minimally adjusts the nominal controller while respecting actuation limits. Around this core, ViSafe integrates a multi-camera extension of a deep-learning aircraft tracker, two decoupled Kalman filters for angles and range because range is noisier, and a reaction-time profile derived from an aviation standard to know at what range a track is reliable. In real-world tests, including head-on encounters at a 144 km/h closure rate, the fastest scenarios showed zero loss of separation, and the digital-twin results were conservative relative to field results.

Load-bearing premise

The formal safety proof holds only when the vision system gives complete and accurate state information with no false negatives, and the paper states this is not true in the real world.

Editorial extensions

If this is right

  • If the system is right, small uncrewed aircraft can perform Detect and Avoid without cooperative surveillance equipment, handling non-cooperative intruders such as balloons and rogue drones on lightweight hardware.
  • The digital twin is conservative relative to field results, so hardware-in-the-loop simulation can serve as a predeployment safety evaluation before costly flight tests.
  • At closure rates up to 144 km/h, the fast real-world scenarios reported no loss of separation, which is the regime where reaction time matters most.
  • The formal forward-invariance guarantee is conditional on continuous accurate perception, so in deployment the paper's evidence is a measured reduction in collision risk rather than an unconditional proof.
  • The below-horizon weakness means ViSafe's guarantees are stronger when the intruder is at or above the horizon; below-horizon encounters in simulation retained a high risk ratio and thousands of separation violations.

Reading between the lines

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

  • This suggests a concrete design rule: choose $d_{\mathrm{thresh}}$ from the detector's measured reliable-detection range rather than from geometry alone, so the CBF is only asked to act when the perception pipeline can actually support it.
  • The same architecture could transfer to other latency-limited platforms: gate a barrier function on the distance at which the perception system is statistically reliable, which matters for ground robots, marine vessels, and any sensor with a detection-latency budget.
  • The paper's own results imply the next bottleneck is perceptual rather than dynamical; improving below-horizon tracking or making the CBF robust to false negatives could recover a formal guarantee that the current system only shows empirically.
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

4 major / 5 minor

Summary. The paper presents ViSafe, a vision-only detect-and-avoid (DAA) system for small uncrewed aerial vehicles. It extends the AirTrack detection and tracking model to multiple cameras, fuses tracks with two Kalman filters, and uses a control-barrier-function-based quadratic program as a supervisory safety controller. The system is evaluated in a hardware-in-the-loop digital twin built on Isaac Sim and in real-world flight tests with closure rates up to 144 km/h. The central claim is that ViSafe can 'provide provably safe runtime guarantees for self-separation in high-speed aerial operations.'

Significance. The engineering and experimental effort is substantial: the authors built a custom SWaP-C payload, integrated an edge-AI detection pipeline with a CBF controller, constructed a realistic digital twin, and conducted extensive real-world flight testing. If the safety claims were properly qualified, the paper would be a valuable systems benchmark for vision-only DAA. However, the headline claim of provably safe runtime guarantees is not established by the proof and is contradicted by the paper's own data in below-horizon scenarios. The work therefore does not currently support its central contribution.

major comments (4)
  1. [Abstract and Sec. IV.C (Eqs. (10)-(13))] The 'provably safe runtime guarantee' claim assumes exact, continuous full-state information about the intruder at every time step. The CBF condition (10) and the QP (13) require h and its Lie derivatives to be evaluated from the state x, but the vision pipeline provides discrete detections that can be missing or stale. Section VI.A.4 explicitly acknowledges that the formal guarantee requires 'always accurate global perception, i.e., no false negatives, which is not the case in the real world.' A false negative or track loss is an interval in which the CBF constraint is inactive or evaluated on incorrect or stale state, so forward invariance of the safe set is not established and Eq. (1) is not enforced. The abstract's claim therefore is not supported by the analysis.
  2. [Tables II and III] The empirical results contradict the claim that ViSafe 'consistently ensures self-separation.' In the digital twin, the below-horizon aggregation over E1-E3 yields P(NMAC)=0.91 and 5466 violations out of 6000 trajectories (Table II), and in the real-world below-horizon runs ViSafe has 4 violations out of 6 runs (Table III). These are large fractions of the tested scenarios, not rare edge cases. Section VI.B identifies below-horizon intruders as a known failure mode, but the abstract and conclusion do not qualify the safety claim accordingly.
  3. [Sec. IV.C (Eq. (12))] The paper does not prove forward invariance of the actual safe set d≥d_thresh. The constructed function h=c+d^n_thresh-d^n-k·ḋ can be negative while d<d_thresh when ḋ is sufficiently positive, so the QP constraint ḣ≤-λh does not by itself imply Eq. (1). The statement that h≤0∩d≥d_thresh is forward invariant under this control law is cited to [32] without a proof in this paper. Since the formal guarantee is a central claim, this missing argument is load-bearing.
  4. [Sec. V.C and Sec. VI.A.2] Part of the 'guarantee' is engineered rather than derived: the CBF hyperparameters (k=0.2, c=0.01, n=0.3, λ=0.2) are tuned for the specific ownship to maintain QP feasibility, and the reliable detection range d_max is computed from AirTrack's own empirical statistics (Sec. V.C). These are reasonable engineering choices, but they mean the safety certification is not a parameter-free or system-independent guarantee as the abstract implies.
minor comments (5)
  1. [Sec. III] The sentence 'This assumption also implies that the township agent possesses no collision avoidance algorithm' appears to contain a typo: 'township' should be 'ownship.'
  2. [Sec. V.G.2] The heading 'ViSafe is robust to weather and lightning conditions' should be 'weather and lighting conditions,' since the experiments vary lighting, not lightning.
  3. [Table III] The reported values are internally inconsistent for some scenarios: for E3, the ViSafe row has 1 violation out of 4 runs but P(NMAC)=0.5 and risk ratio 0.25, which imply 2 and 1 violations, respectively. Please clarify the metric definitions and report counts consistently.
  4. [Sec. V.C] The reaction-time calculation states d_max=163.2×l but does not specify the units of the 163.2 factor or the focal length used; please make the computation explicit.
  5. [Fig. 3 and Eqs. (15)-(17)] The angle α=π+θ is used in the encounter geometry and in the expressions for ḋ and d̈, but it is not defined in the text; define it where the encounter geometry is introduced.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the CBF safety argument is a standard conditional theorem, and the paper's overclaiming is a scope issue rather than a circular derivation.

full rationale

The paper's formal safety argument is a standard CBF/QP construction: Eq. (10) is the textbook forward-invariance condition from Ames et al. [7], and Eq. (13) enforces it online. The paper does not define its safety predicate in terms of the result it claims, nor does it fit a parameter and then relabel that fit as a prediction. The hyperparameters k, c, n, and λ are tuned for QP feasibility (Sec. VI.A.2), which is a synthesis choice, not a circular derivation. The detection-range profile (Sec. V.C) inherits empirical statistics from the authors' prior AirTrack work [17]; that is a self-citation, but it is external, falsifiable empirical evidence about a perception module, not an assumption that entails the paper's avoidance-safety conclusion. The dependence on [32] for the invariant-set argument is likewise a peer-reviewed external theorem, and the paper explicitly concedes in Sec. VI.A.4 that the unconditional 'always accurate global perception' assumption fails in the real world; this is a limitation of the guarantee, not a circular substitution. The abstract's 'provably safe runtime guarantees' wording overstates the conditional proof and the below-horizon violation data (Tables II/III), but overclaiming a conditional result is a correctness/scope problem, which this pass does not score as circularity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central safety guarantee rests on tuned CBF hyperparameters and on assumptions (constant-velocity intruder, planar motion, reliable detection range) that the paper itself acknowledges are violated in real deployment.

free parameters (6)
  • CBF hyperparameter k = 0.2
    Tuned to ensure QP feasibility for the Aurelia X6 ownship (Sec IV.C).
  • CBF hyperparameter c = 0.01
    Offset in barrier function h(x), tuned for the safety margin (Sec IV.C).
  • CBF hyperparameter n = 0.3
    Exponent in h(x), tuned to shape the safe set and avoid singularities (Sec IV.C).
  • CBF hyperparameter λ = 0.2
    Decay rate in the condition ˙h ≤ -λh, tuned for the specific ownship (Sec IV.C).
  • Separation threshold d_thresh = 0.05 nm (shown in Fig. 1)
    Defines the safe set in Eq. (11); value appears only in a figure, not stated in text.
  • Kalman filter noise covariances = not specified
    The paper says decoupled filters required significant joint tuning; actual covariances are not listed (Sec IV.B).
assumptions (4)
  • domain assumption Intruder maintains a constant velocity vector and does not react to the ownship (Eq. 4).
    Used to derive ¨d in Eq. (16); acknowledged as a limitation in Sec VI.B.
  • domain assumption Ownship dynamics are planar unicycle with bounded inputs (Eq. 2).
    The avoidance problem is modeled in the horizontal plane; vertical motion is ignored.
  • standard math The CBF condition ˙h ≤ -λh guarantees forward invariance of the safe set (Eq. 10).
    Standard CBF theory from Ames et al. [7]; the specific intersection invariant set h≤0∩d≥d_thresh is cited to [32] without proof.
  • domain assumption Vision detections are reliable within the empirically derived range profile (163.2×l from ASTM F3442).
    The system only activates avoidance when a track exists; the paper concedes in Sec VI.A.4 that false negatives occur in practice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Demonstrating ViSafe: Vision-enabled Safety for High-speed Detect and Avoid." pith.science (2026). https://pith.science/paper/VDBEABRU

@misc{pith2026250503694,
  author       = {Pith},
  title        = {Pith review of: Demonstrating ViSafe: Vision-enabled Safety for High-speed Detect and Avoid},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VDBEABRU}},
  note         = {Machine review of arXiv:2505.03694}
}
read the original abstract

Assured safe-separation is essential for achieving seamless high-density operation of airborne vehicles in a shared airspace. To equip resource-constrained aerial systems with this safety-critical capability, we present ViSafe, a high-speed vision-only airborne collision avoidance system. ViSafe offers a full-stack solution to the Detect and Avoid (DAA) problem by tightly integrating a learning-based edge-AI framework with a custom multi-camera hardware prototype designed under SWaP-C constraints. By leveraging perceptual input-focused control barrier functions (CBF) to design, encode, and enforce safety thresholds, ViSafe can provide provably safe runtime guarantees for self-separation in high-speed aerial operations. We evaluate ViSafe's performance through an extensive test campaign involving both simulated digital twins and real-world flight scenarios. By independently varying agent types, closure rates, interaction geometries, and environmental conditions (e.g., weather and lighting), we demonstrate that ViSafe consistently ensures self-separation across diverse scenarios. In first-of-its-kind real-world high-speed collision avoidance tests with closure rates reaching 144 km/h, ViSafe sets a new benchmark for vision-only autonomous collision avoidance, establishing a new standard for safety in high-speed aerial navigation.

Figures

Figures reproduced from arXiv: 2505.03694 by the authors.

Figure 1
Figure 1. We demonstrate ViSafe, a high-speed vision-only airborne collision avoidance system. Top: Rendering of a real-world flight test log where the ViSafe system detects an incoming intruder with a 144 km/h closure rate and performs an avoidance maneuver to ensure safe separation. The annotations showcase the detections and multi-camera tracks from the vision-based aircraft detection and tracking system, while the traject… view at source ↗
Figure 2
Figure 2. Overview of our ViSafe framework for real-world testing and hardware-in-the-loop simulation: Firstly, the onboard sensors or digital twin simulation stream the multi-cam videos to the AirTrack visual detection module, which detects the intruder across multiple views. Then, these detections, along with the ownship state information, are sent to the multi-view fusion and coordinate frame conversion module, which then … view at source ↗
Figure 3
Figure 3. Encounter geometry and information required for vision-enabled collision avoidance. The velocity of the ownship vown and heading with respect to North χown are obtained from the ownship odometry. The intruder’s range d, azimuth θ, velocity vint , and heading with respect to North χint are obtained using the visual detection module. while the range is predicted by the monocular deep learning model (AirTrack). It is p… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Diversity of the airborne collision testing scenarios: (a) The various encounter geometries used for real-world & simulation flight testing. (b) The diverse weather and lighting conditions that were used to evaluate ViSafe’s robustness in simulation. In the simulation,…
Figure 5
Figure 5. Figure 5: Average horizontal rate of closure comparisons across different encounter geometries in real-world testing: Higher values indicate that agents are moving apart, showcasing diverging & safe trajectories. Under different testing scenarios, it can be seen that ViSafe cons…
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 48 canonical work pages

  1. [32]

    Control in a safe set: Addressing safety in human-robot interactions

    Changliu Liu and Masayoshi Tomizuka. Control in a safe set: Addressing safety in human-robot interactions. In IEEE/ACM International Conference on Human-Robot Interaction, 2014. 6

    work page 2014

  2. [1]

    URL https://registry

    Airborne object tracking dataset. URL https://registry. opendata.aws/airborne-object-tracking. 3, 4

    work page

  3. [2]

    URL https://github

    Dronekit-python: Developer sdk for python to control drones using the mavlink protocol. URL https://github. com/dronekit/dronekit-python. 8

    work page

  4. [3]

    https://www.esri.com/en-us/ arcgis/products/arcgis-cityengine/overview

    Procedural City Generator — 3D City Maker — ArcGIS CityEngine — esri.com. https://www.esri.com/en-us/ arcgis/products/arcgis-cityengine/overview. [Accessed 11-12-2024]. 9

    work page 2024

  5. [4]

    https://developer

    Isaac Sim — developer.nvidia.com. https://developer. nvidia.com/isaac/sim. [Accessed 11-12-2024]. 9

    work page 2024

  6. [5]

    URL https://developer.nvidia

    Nvidia isaac sim, 2022. URL https://developer.nvidia. com/isaac-sim. 2

    work page 2022

  7. [6]

    Discrete control barrier functions for safety-critical control of discrete systems with application to bipedal robot navigation

    Ayush Agrawal and Koushil Sreenath. Discrete control barrier functions for safety-critical control of discrete systems with application to bipedal robot navigation. In Robotics: Science and Systems , 2017. URL https: //api.semanticscholar.org/CorpusID:1780280. 2

    work page 2017

  8. [7]

    Ames, Samuel Coogan, Magnus Egerst- edt, Gennaro Notomista, Koushil Sreenath, and Paulo Tabuada

    Aaron D. Ames, Samuel Coogan, Magnus Egerst- edt, Gennaro Notomista, Koushil Sreenath, and Paulo Tabuada. Control barrier functions: Theory and appli- cations. In 2019 18th European Control Conference (ECC), pages 3420–3431, 2019. doi: 10.23919/ECC. 2019.8796030. 2, 6

    arXiv 2019

Show all 58 references
  1. [8]

    Optimized airborne collision avoid- ance in mixed equipage environments

    DM Asmar, MJ Kochenderfer, DM Asmar, and MJ Kochenderfer. Optimized airborne collision avoid- ance in mixed equipage environments. Lincoln Labora- tory, MIT Report , 2013. 2

    2013

  2. [9]

    Simple online and realtime tracking

    Alex Bewley, Zongyuan Ge, Lionel Ott, Fabio Ramos, and Ben Upcroft. Simple online and realtime tracking. In 2016 IEEE international conference on image processing (ICIP), pages 3464–3468. IEEE, 2016. 3

    2016

  3. [10]

    Robust control barrier functions under high relative degree and input constraints for satellite trajectories, 2021

    Joseph Breeden and Dimitra Panagou. Robust control barrier functions under high relative degree and input constraints for satellite trajectories, 2021. URL https: //arxiv.org/abs/2107.04094. 10

    2021 arXiv

  4. [11]

    Image processing algorithms for uav” sense and avoid”

    Ryan Carnie, Rodney Walker, and Peter Corke. Image processing algorithms for uav” sense and avoid”. In Proceedings 2006 IEEE International Conference on Robotics and Automation, 2006. ICRA 2006. , pages 2848–2853. IEEE, 2006. 3

    2006

  5. [12]

    Aircraft detection by deep convolu- tional neural networks

    Xueyun Chen, Shiming Xiang, Cheng-Lin Liu, and Chun-Hong Pan. Aircraft detection by deep convolu- tional neural networks. IPSJ Transactions on Computer Vision and Applications , 7:10–17, 2014. 3

    2014

  6. [13]

    Passive, long-range detection of aircraft: towards a field deployable sense and avoid system

    Debadeepta Dey, Christopher Geyer, Sanjiv Singh, and Matt Digioia. Passive, long-range detection of aircraft: towards a field deployable sense and avoid system. In Field and Service Robotics , pages 113–123. Springer,

  7. [14]

    Centernet: Keypoint triplets for object detection

    Kaiwen Duan, Song Bai, Lingxi Xie, Honggang Qi, Qingming Huang, and Qi Tian. Centernet: Keypoint triplets for object detection. In Proceedings of the IEEE/CVF international conference on computer vision , pages 6569–6578, 2019. 3

    2019

  8. [15]

    Morpho- logical filtering and target tracking for vision-based uas sense and avoid

    Giancarmine Fasano, Domenico Accardo, Anna Elena Tirri, Antonio Moccia, and Ettore De Lellis. Morpho- logical filtering and target tracking for vision-based uas sense and avoid. In 2014 International Conference on Unmanned Aircraft Systems (ICUAS) , pages 430–440. IEEE, 2014. 3

    2014

  9. [16]

    High-order control barrier function based collision avoidance forma- tion tracking of constrained fixed-wing aircraft

    Junjie Fu, Guanghui Wen, and Jinde Cao. High-order control barrier function based collision avoidance forma- tion tracking of constrained fixed-wing aircraft. In 2023 5th International Conference on Industrial Artificial In- telligence (IAI), pages 1–6, 2023. doi: 10.1109/IAI5...

    2023

  10. [17]

    Air- track: Onboard deep learning framework for long-range aircraft detection and tracking

    Sourish Ghosh, Jay Patrikar, Brady Moon, Mi- lad Moghassem Hamidi, and Sebastian Scherer. Air- track: Onboard deep learning framework for long-range aircraft detection and tracking. In 2023 IEEE Interna- tional Conference on Robotics and Automation (ICRA) , pages 1277–1283, 20...

    2023

  11. [18]

    A comprehensive survey on conflict detection and resolu- tion in unmanned aircraft system traffic management

    Asma Hamissi, Amine Dhraief, and Layth Sliman. A comprehensive survey on conflict detection and resolu- tion in unmanned aircraft system traffic management. IEEE Transactions on Intelligent Transportation Systems,

  12. [19]

    Neural control barrier functions for safe navigation, 2024

    Marvin Harms, Mihir Kulkarni, Nikhil Khedekar, Martin Jacquet, and Kostas Alexis. Neural control barrier functions for safe navigation, 2024. URL https://arxiv. org/abs/2407.19907. 2

    2024 arXiv

  13. [20]

    Foundloc: Vision-based onboard aerial localiza- tion in the wild

    Yao He, Ivan Cisneros, Nikhil Keetha, Jay Patrikar, Zelin Ye, Ian Higgins, Yaoyu Hu, Parv Kapoor, and Sebastian Scherer. Foundloc: Vision-based onboard aerial localiza- tion in the wild. arXiv preprint arXiv:2310.16299, 2023. 9

    2023 arXiv

  14. [21]

    Finding tiny faces

    Peiyun Hu and Deva Ramanan. Finding tiny faces. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 951–959, 2017. 3

    2017

  15. [22]

    Aircraft detection using deep convolutional neural network in small unmanned aircraft systems

    Sunyou Hwang, Jaehyun Lee, Heemin Shin, Sungwook Cho, and David Hyunchul Shim. Aircraft detection using deep convolutional neural network in small unmanned aircraft systems. In 2018 AIAA Information Systems- AIAA Infotech@ Aerospace , page 2137, 2018. 3

    2018

  16. [23]

    Pegasus simulator: An isaac sim framework for mul- tiple aerial vehicles simulation

    Marcelo Jacinto, Jo ˜ao Pinto, Jay Patrikar, John Keller, Rita Cunha, Sebastian Scherer, and Ant ´onio Pascoal. Pegasus simulator: An isaac sim framework for mul- tiple aerial vehicles simulation. In 2024 International Conference on Unmanned Aircraft Systems (ICUAS) , pages 91...

    2024

  17. [24]

    Learning to detect aircraft for long-range vision-based sense-and-avoid systems

    Jasmin James, Jason J Ford, and Timothy L Molloy. Learning to detect aircraft for long-range vision-based sense-and-avoid systems. IEEE Robotics and Automation Letters, 3(4):4383–4390, 2018. 3

    2018

  18. [25]

    Below horizon aircraft detection using deep learning for vision-based sense and avoid

    Jasmin James, Jason J Ford, and Timothy L Molloy. Below horizon aircraft detection using deep learning for vision-based sense and avoid. In 2019 International Conference on Unmanned Aircraft Systems (ICUAS) , pages 965–970. IEEE, 2019. 3

    2019

  19. [26]

    Airobject: A temporally evolving graph embedding for object identification

    Nikhil Varma Keetha, Chen Wang, Yuheng Qiu, Kuan Xu, and Sebastian Scherer. Airobject: A temporally evolving graph embedding for object identification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8407–8416, 2022. 3

    2022

  20. [27]

    The traffic alert and collision avoidance system

    JE Kuchar and Ann C Drumm. The traffic alert and collision avoidance system. Lincoln laboratory journal , 16(2):277, 2007. 2

    2007

  21. [28]

    Hidden markov model filter banks for dim target detection from image sequences

    John Lai, Jason J Ford, Peter O’Shea, and Rodney Walker. Hidden markov model filter banks for dim target detection from image sequences. In 2008 Digital Image Computing: Techniques and Applications , pages 312–

    2008

  22. [29]

    Airborne vision-based collision-detection system

    John Lai, Luis Mejias, and Jason J Ford. Airborne vision-based collision-detection system. Journal of Field Robotics, 28(2):137–157, 2011. 3

    2011

  23. [30]

    Characterization of sky-region morphological-temporal airborne collision detection

    John Lai, Jason J Ford, Luis Mejias, and Peter O’Shea. Characterization of sky-region morphological-temporal airborne collision detection. Journal of Field Robotics , 30(2):171–193, 2013. 3

    2013

  24. [31]

    Learning calibrated-guidance for object detec- tion in aerial images

    Dong Liang, Zongqi Wei, Dong Zhang, Qixiang Geng, Liyan Zhang, Han Sun, Huiyu Zhou, Mingqiang Wei, and Pan Gao. Learning calibrated-guidance for object detec- tion in aerial images. arXiv preprint arXiv:2103.11399 ,

    arXiv

  25. [33]

    An introduction to acas xu and the challenges ahead

    Guido Manfredi and Yannick Jestin. An introduction to acas xu and the challenges ahead. In 2016 IEEE/AIAA 35th Digital Avionics Systems Conference (DASC), pages 1–9, 2016. doi: 10.1109/DASC.2016.7778055. 1, 3

    2016

  26. [34]

    Targeted image transfor- mation for improving robustness in long range aircraft detection

    Rebecca Martin, Clement Fung, Nikhil Keetha, Lujo Bauer, and Sebastian Scherer. Targeted image transfor- mation for improving robustness in long range aircraft detection. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 10431– 10438. I...

    2024

  27. [35]

    Detection of aircraft in video sequences using a predictive optical flow algorithm

    Jeffrey W McCandless. Detection of aircraft in video sequences using a predictive optical flow algorithm. Optical Engineering, 38(3):523–530, 1999. 3

    1999

  28. [36]

    Vision-based detection and tracking of aerial targets for uav collision avoidance

    Luis Mejias, Scott McNamara, John Lai, and Jason Ford. Vision-based detection and tracking of aerial targets for uav collision avoidance. In 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 87–

    2010

  29. [37]

    Detection of aircraft below the horizon for vision-based detect and avoid in unmanned aircraft systems

    Timothy L Molloy, Jason J Ford, and Luis Mejias. Detection of aircraft below the horizon for vision-based detect and avoid in unmanned aircraft systems. Journal of Field Robotics , 34(7):1378–1391, 2017. 3

    2017

  30. [38]

    Molnar, Suresh K

    Tamas G. Molnar, Suresh K. Kannan, James Cunning- ham, Kyle Dunlap, Kerianne L. Hobbs, and Aaron D. Ames. Collision avoidance and geofencing for fixed- wing aircraft with control barrier functions, 2024. URL https://arxiv.org/abs/2403.02508. 2

    2024 arXiv

  31. [39]

    Real-time safety and control of robotic manipulators with torque saturation in operational space

    Muhammad Ali Murtaza, Sergio Aguilera, Vahid Azimi, and Seth Hutchinson. Real-time safety and control of robotic manipulators with torque saturation in operational space. In 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 702–708,

    2021

  32. [40]

    Aerial object tracking from an airborne platform

    Andreas Nussberger, Helmut Grabner, and Luc Van Gool. Aerial object tracking from an airborne platform. In 2014 international conference on unmanned aircraft systems (ICUAS), pages 1284–1293. IEEE, 2014. 3

    2014

  33. [41]

    Owen, Sean M

    Michael P. Owen, Sean M. Duffy, and Matthew W. M. Edwards. Unmanned aircraft sense and avoid radar: Surrogate flight testing performance evaluation. In 2014 IEEE Radar Conference , pages 0548–0551, 2014. doi: 10.1109/RADAR.2014.6875652. 2

    2014

  34. [42]

    Challenges in close- proximity safe and seamless operation of manned and unmanned aircraft in shared airspace

    Jay Patrikar, Joao Dantas, Sourish Ghosh, Parv Kapoor, Ian Higgins, Jasmine J Aloor, Ingrid Navarro, Jimin Sun, Ben Stoler, Milad Hamidi, et al. Challenges in close- proximity safe and seamless operation of manned and unmanned aircraft in shared airspace. arXiv preprint arXiv:...

    2022 arXiv

  35. [43]

    Image, speech, and ads-b trajectory datasets for terminal airspace operations

    Jay Patrikar, Joao Dantas, Brady Moon, Milad Hamidi, Sourish Ghosh, Nikhil Keetha, Ian Higgins, Atharva Chandak, Takashi Yoneyama, and Sebastian Scherer. Image, speech, and ads-b trajectory datasets for terminal airspace operations. Scientific Data, 12(1):468, 2025. 3

    2025

  36. [44]

    Learning to detect aircraft at low resolutions

    Stavros Petridis, Christopher Geyer, and Sanjiv Singh. Learning to detect aircraft at low resolutions. In Inter- national Conference on Computer Vision Systems , pages 474–483. Springer, 2008. 3

    2008

  37. [45]

    Detection and tracking of large number of targets in wide area surveillance

    Vladimir Reilly, Haroon Idrees, and Mubarak Shah. Detection and tracking of large number of targets in wide area surveillance. In European conference on computer vision, pages 186–199. Springer, 2010. 3

    2010

  38. [46]

    Fly- ing objects detection from a single moving camera

    Artem Rozantsev, Vincent Lepetit, and Pascal Fua. Fly- ing objects detection from a single moving camera. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 4128–4136, 2015. 3

    2015

  39. [47]

    Minimum Operational Performance Standards for Airborne Collision Avoidance System X (ACAS X) (ACAS Xa and ACAS Xo), Volume I and Volume II

    RTCA. Minimum Operational Performance Standards for Airborne Collision Avoidance System X (ACAS X) (ACAS Xa and ACAS Xo), Volume I and Volume II . DO 385A. RTCA, 2023. Approved by the RTCA Program Management Committee (PMC) as documented in June

    2023

  40. [48]

    Robust reg- istration and filtering for moving object detection in aerial videos

    Falk Schubert and Krystian Mikolajczyk. Robust reg- istration and filtering for moving object detection in aerial videos. In 2014 22nd International Conference on Pattern Recognition , pages 2808–2813. IEEE, 2014. 3

    2014

  41. [49]

    Complex opti- cal remote-sensing aircraft detection dataset and bench- mark

    Tianjun Shi, Jinnan Gong, Shikai Jiang, Xiyang Zhi, Guangzhen Bao, Yu Sun, and Wei Zhang. Complex opti- cal remote-sensing aircraft detection dataset and bench- mark. IEEE Transactions on Geoscience and Remote Sensing, 61:1–9, 2023. 3

    2023

  42. [50]

    Avoidds: Aircraft vision-based intruder detection dataset and simulator

    Elysia Smyers, Sydney Katz, Anthony Corso, and Mykel J Kochenderfer. Avoidds: Aircraft vision-based intruder detection dataset and simulator. Advances in Neural Information Processing Systems , 36:8076–8091,

  43. [51]

    Constructive barrier certificates with applications to fixed-wing aircraft collision avoidance

    Eric Squires, Pietro Pierpaoli, and Magnus Egerstedt. Constructive barrier certificates with applications to fixed-wing aircraft collision avoidance. In 2018 IEEE Conference on Control Technology and Applications (CCTA), pages 1656–1661, 2018. doi: 10.1109/CCTA. 2018.8511342. 3

    2018

  44. [52]

    Composition of safety constraints with applications to decentralized fixed-wing collision avoidance

    Eric Squires, Pietro Pierpaoli, Rohit Konda, Samuel Coogan, and Magnus Egerstedt. Composition of safety constraints with applications to decentralized fixed-wing collision avoidance. CoRR, abs/1906.03771, 2019. URL http://arxiv.org/abs/1906.03771. 3

    1906 arXiv

  45. [53]

    A method for detection of small moving objects in uav videos

    Vladan Stojni ´c, Vladimir Risojevi ´c, Mario Mu ˇstra, Ve- dran Jovanovi ´c, Janja Filipi, Nikola Kezi ´c, and Zdenka Babi´c. A method for detection of small moving objects in uav videos. Remote Sensing, 13(4):653, 2021. 3

    2021

  46. [54]

    Robust aircraft detection in imbalanced and sim- ilar classes with a multi-perspectives aircraft dataset

    Bo Yang, Dongjian Tian, Songliang Zhao, Wei Wang, Jun Luo, Huayan Pu, Mingliang Zhou, and Yangjun Pi. Robust aircraft detection in imbalanced and sim- ilar classes with a multi-perspectives aircraft dataset. IEEE Transactions on Intelligent Transportation Systems,

  47. [55]

    Efficient motion planning for manipulators with control barrier function- induced neural controller, 2024

    Mingxin Yu, Chenning Yu, M-Mahdi Naddaf-Sh, Devesh Upadhyay, Sicun Gao, and Chuchu Fan. Efficient motion planning for manipulators with control barrier function- induced neural controller, 2024. URL https://arxiv.org/ abs/2404.01184. 2

    2024 arXiv

  48. [56]

    Safety index synthesis via sum-of-squares programming, 2023

    Weiye Zhao, Tairan He, Tianhao Wei, Simin Liu, and Changliu Liu. Safety index synthesis via sum-of-squares programming, 2023. URL https://arxiv.org/abs/2209. 09134. 10

    2023

  49. [57]

    Tracking objects as points

    Xingyi Zhou, Vladlen Koltun, and Philipp Kr ¨ahenb¨uhl. Tracking objects as points. In European Conference on Computer Vision, pages 474–490. Springer, 2020. 3, 5

    2020

  50. [2021]

    doi: 10.1109/IROS51168.2021.9636794. 2

    2021

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.