Vertical aerial view of a multi-level urban highway interchange

Drones and AI for infrastructure inspection

Pavement condition, measured from the air.

A 15-minute drone flight becomes an ASTM PCI report: every crack detected, measured, and mapped. Fly any survey-grade drone, or Nimbus, our long-range VTOL designed and built in Europe.

24hFrom your upload to the report in your inbox
100%Surface coverage, versus 10 to 20% sampled by manual inspection
1mmSmallest crack detected and measured
3,000+Flight test hours behind Nimbus

Show the work

Raw imagery in. Engineering data out.

You are looking at a real runway. Drag the handle.

Raw drone orthoimagery of a runway surface before analysis The same runway surface with HALO AI crack segmentation overlay Raw imagery HALO AI detection
The overlay on the right is what the ASTM tables in the report are computed from.

One real inspection, as HALO AI reported it.

27,000of pavement flown and analyzed in one inspection.
979separate distresses found on that surface, each classified by type and severity and measured.
18the PCI score that came back. On the ASTM 0 to 100 scale, that is Failed.

Will an AI score hold up in an audit? The deduct values and the max CDV are printed in the report, so your auditor can re-run the ASTM math by hand. See the full sample report.

HALO AI detection overlay: annotated parking lot orthomosaic with a table of measured distresses
HALO AI's detection overlay on a real lot: each distress traced, with confidence, area, and thickness in the table.

The platform

HALO AI works with the drone you already fly.

HALO AI traces every crack, measures it, scores the pavement to ASTM D6433 or D5340, and totals the repair quantities. You get back a PCI report, a defect map, and an interactive digital twin.

Crack width, length, and area come back 98% accurate, and the PCI map lands within 95% of what a qualified engineer scores on the same pavement. Two engineers scoring the same surface by hand typically land about 16% apart. One model scores every survey, so the number you get in March is comparable in October.

HALO AI needs high-resolution, georeferenced imagery. DJI, Wingtra, Skydio, and more already produce it, and files upload as they come off the card.

Explore HALO AI
Operator holding the red Nimbus VTOL overhead, full wingspan against the sky
2.82 m of wingspan, carried by one person.

The aircraft

Nimbus flies the whole corridor in one launch.

Nimbus takes off and lands vertically, so a survey launches from any clear area on site. It flies up to 300 km of route between one launch and one landing, and the corridor comes back as one continuous dataset, captured in a single weather window.

The payload margin is sized for a survey-grade head: a full-frame mapping camera, a LiDAR unit with its IMU, or a thermal and RGB pair carried together. Sony, Phase One, YellowScan, Gremsy, and Workswell are among the heads already integrated.

Explore Nimbus

Why Fadron

Two Fadron engineers assembling a fixed-wing UAV on a workbench
Airframe assembly at the Fadron facility in Sofia.

Why Fadron

We build the aircraft, the AI, and the report.

One team stands behind the number in your report, the software that produced it, and the aircraft that captured it. Changes to any of the three are engineered and validated in house.

Our mission is to make critical infrastructure safer, smarter, and more resilient by combining drones and AI into a seamless system for continuous, automated inspection. That leads to condition monitoring on a fixed interval: every asset measured on the same basis, on a schedule, before anything has failed.

Named Innovative Enterprise of the Year 2025 by ARC Fund. Holder of an EIC Seal of Excellence.

About Fadron

News & insights

From the field and the lab.

Technical notes and field stories on drone inspection, long-range mapping, and AI perception, on the Fadron blog. Every post carries at least one original number or image.

  • Runway FOD: what a walk-down finds, what it misses, and what that costs in closure time.
  • The ground sampling distance your survey actually needs: about 5 mm/px for road PCI, about 3 mm/px for airfields.
  • Why catching a crack under 2 mm buys three to five more years of pavement life.
Read the articles

See what HALO AI finds on your pavement.

Meet HALO AI, the analysis platform, and Nimbus, the aircraft. Send imagery of your own site and we will analyze 1 ha of it at no charge.

Request a demo