What happened
The FAA's Airport Technology Research and Development Branch, with contractors Woolpert and Neya Systems, built a proof-of-concept workflow to detect foreign object debris on airport surfaces. A small unmanned aircraft photographs the pavement, and a trained deep-learning anomaly detector, FastFlow, flags debris in the images. The team tested the workflow at Cape May County Airport in three stages and then validated it on a runway at Atlantic City International Airport. It placed the debris items that FAA Advisory Circular 150/5220-24 specifies. In its report of September 2024, the branch states that 173 of 180 items (96%) were detected across initial and validation testing, above the circular's 90% requirement. Lug nuts, deep sockets and distorted metal strips were missed most often. The workflow met the circular's location accuracy, inspection frequency and surveillance area requirements. It fell short on false positives, which greatly exceeded the allowed one per day, on processing time, on a software interface for alerts and on low-light and bad-weather operation. The team tested its own workflow. No airport used it in operation, and no debris strike was prevented.
Sources
-
Report, by the makers.
The sUAS-based FOD detection workflow, which used the FastFlow ML deep learning algorithm, was capable of meeting some of the AC 150/5220-24 requirements, including achieving a 96% detection rate for FOD items specified in the AC.
Of 180 FOD items, 173 (96%) were detected overall. The FOD items not detected included the lug nut (12 of 15 detected), the deep socket (14 of 15 detected), and the distorted metal strip (12 of 15 detected).
Details
AI system
- Workflow
-
The team flew a commercial small unmanned aircraft over runways and trained the FastFlow deep-learning anomaly detector on the images. Test items from the advisory circular's list were placed on the pavement at two airports. Images taken from directly above at a 0.2-cm ground sampling distance detected more items than oblique images.
Follow-up
- Conditions for replication
-
Daylight and fair weather, since the workflow did not cover low light or inclement weather. Someone must sort the false positives, which greatly exceeded the circular's allowed rate. Processing time must be acceptable, and the report says that it still needs reduction.
Entry details
- Sector
-
Aviation safety / airport FOD detection.
- Purpose
-
The FAA's Airport Technology Research and Development Branch built the sUAS-based FOD detection workflow to test whether commercial drones and machine-learning algorithms can detect foreign object debris on airport surfaces to the standard of Advisory Circular 150/5220-24.
31 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this benefit
Peter H. Vartanian. (2026). Drone plus FastFlow finds 96% of test runway debris for the FAA (AIBD-1200). Artificial Intelligence Benefits Database, edition of 2 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1200.html (archived at https://doi.org/10.7910/DVN/AADSII, version 9.0)