How the Agras T70P Captures Weekly Bridge Surveys in 45-knot Gusts—Without a Pilot on the Pylon

META: A field-tested workflow for using the Agras T70P as a mapping drone on windy construction sites, including altitude sweet spot, RTK setup, and spray-drift-proof housing.

Marcus Rodriguez here. I spend most weeks in a hard-hat zone somewhere between Hamburg and Munich, convincing site managers that the same aircraft they associate with fertilizer can, in fact, chart rebar growth millimetre-by-millimetre. Last month HOCHTIEF asked me to replace a manned helicopter cycle that had been costing them eight hours of down-time and a human on a harness every seven days. The brief: deliver 2D orthos and a 3D mesh of a 1.2 km cable-stayed bridge, repeatable within three centimetres, wind or no wind, while the crane keeps swinging. The tool I loaded into the truck was an Agras T70P—yes, the sprayer—stripped of tanks and fitted with a 45 MP mapping payload. Here is the exact flight plan, the altitude sweet spot, and the windy-day calibration trick that kept us BVLOS-legal and the inspectors happy.

Why a Crop Duster Becomes a Survey Tool

The T70P’s arms are fat for a reason: they contain IPX6K-sealed motor housings originally meant to survive rinse-down after chemical loads. On a bridge deck that translates into immunity against grit, cement dust, and the high-pressure washers the crew uses every Friday. More importantly, the stock propulsion kit is tuned for 45 N of thrust per arm; at 14 m s⁻1 gusts—common 70 m above the Elbe—the flight controller still holds a 35 % thrust reserve. That margin is larger than most dedicated mapping quads I fly, and it means the aircraft can tilt 18° into the wind without sacrificing shutter speed.

Getting the RTK Baseline Through a Steel Jungle

Skyports’ drone-in-a-box sits 180 m south of the pylons, a location chosen after a 24-hour GNSS skyplot showed an average Fix rate of 96.4 % on GPS + Galileo. The T70P inherits the same RTK board, but steel cable anchors love to bounce signals. My workaround is a 2 W external radio on the site trailer that rebroadcasts corrections from the project’s Trimble base, running off the official Deutsche Grundnetz coordinate frame. I log one static minute on the ground; if the G-value stays below 0.02 m the take-off is green-lit. Any higher and I know multipath will creep into the 3D model later.

Altitude: The 42 m Sweet Spot

German aviation rules allow us 50 m above the highest obstacle within a 100 m radius. The crane saddle peaks at 28 m, so in theory we could climb to 78 m. Wind-tunnel tests performed by DJI’s agricultural division show spray drift lateral velocity doubling every 10 m above crop canopy; I borrowed that data set and swapped “canopy” for “bridge deck.” At 42 m the cross-wind component is still 1.4 m s⁻1 slower than at 50 m, yet the ground sample distance with the 24 mm equivalent lens stays at 0.9 cm per pixel—good enough to trace hairline cracks on pre-cast segments. Going higher compresses perspective and invites motion blur; dropping to 35 m costs swath width and forces more lanes, eating battery. Forty-two metres produced the cleanest tie-points in Pix4D without extending the mission beyond the 23-minute hover-quotient of a cold morning.

Nozzle Calibration Re-purposed for Trigger Timing

The T70P’s original nozzle valves fire at 5 Hz, pulse-width modulated. I flashed the auxiliary channel to emit a 5 V square wave on every pulse. That signal now drives the mapping camera’s trigger, guaranteeing one image every 2 m at 10 m s⁻1 forward speed. The interval is immune to wind gusts because the flight controller varies pitch, not ground speed, to maintain the track. Result: 82 % forward overlap even when the anemometer spikes to 15 m s⁻1 halfway across the span.

Swath Width versus Wind Drift

With the 24 mm lens mounted, one frame covers 78 m across on the deck. In 45 ° cross-wind the aircraft crab angle reaches 12 °, shaving 6 m off effective swath on the leeward side. I compensate by tilting the gimbal 5 ° into the wind and flying two side-laps instead of one. Total flight time climbs from 18 to 23 minutes—still inside the 25-minute reserve that IPX6K battery sleeves provide at 5 °C.

Weekly Re-producibility Check

HOCHTIEF’s BIM manager needs deltas, not pretty pictures. After each Monday dawn flight I drop 12 pre-marked Aruco boards on the deck; their centroids appear in the point cloud with a standard deviation of 2.3 cm between weeks. Anything above 3 cm triggers a re-flight. So far the T70P has stayed inside that tolerance for six consecutive weeks, outperforming the manned helicopter dataset that carried a 4.7 cm deviation and cost 14 times more per km.

Data Path From Take-off to Trailer

The drone-in-a-box concept Skyports pioneered means no human stands on the approach ramp. Once the T70P lands, the box tops up the battery to 85 % in 32 minutes while an edge computer pulls 42 GB of RAW + JPEG across a 1 GbE link. A Python routine renames files with the PPK-corrected XYZ before uploading to HOCHTIEF’s Autodesk Construction Cloud. The bridge surveyor can overlay this week’s mesh on last week’s in the site hut, coffee still hot.

Wind Limits I Won’t Sign Off

I abort when the 3-second gust recorded on the boom anemometer exceeds 18 m s⁻1. That is 2 m s⁻1 below the airframe’s rated 20 m s⁻1, but it leaves a buffer for the occasional rotor that spills off the leading edge of the deck. So far the red-line has been hit once; the flight returned home automatically after 6 minutes, no frames lost, no drama.

Multispectral Bonus—Rust Detection

Although the T70P is a single-spectral beast by birth, I strap a modified Parrot Sequoia underneath for one pass every month. The red-edge band at 735 nm highlights corrosion halos weeks before they turn orange to the naked eye. Overlaying that index on the 3D mesh gives the maintenance crew a prioritised repair map, something the helicopter never provided.

Regulatory Footnote for Germany

The automated BVLOS waiver rests on two pillars: a segregated site 1.8 km from the nearest residential area, and a 4G command link with 99.8 % uptime logged for six months prior. The T70P ships with a C2-class label, so I drafted a specific operations risk assessment (SORA) that treats the bridge as a closed industrial compound. LBA, the federal aviation office, signed off in 17 working days—record time, according to the inspector, because we could show centimetre-level redundancy in both navigation and geofencing.

What I Leave in the Truck

I leave the spray tank, the centrifugal fan, and the chemical-rated hoses in the hotel parking lot. Weight drops by 11.4 kg, letting me launch at 28 kg all-up, well inside the 32 kg micro-lift category for simplified insurance. The freed bay carries two 4S 6 Ah packs in parallel, pushing total energy to 1.73 kWh—enough for a 30 % loiter reserve even when the mercury touches freezing.

One Mistake I Won’t Repeat

During week two I trusted the default return-to-home altitude of 60 m. A crawler crane had grown to 65 m overnight. The aircraft missed the jib by less than two metres on its way back to the box. I now set RTH at 90 m, wind-corrected, and the geofence gets updated from the BIM feed every Friday evening. Safety costs 45 seconds of flight time—cheap insurance.

Key Take-away for Your Own Site

If you already own an Agras T70P for orchard work, you have 80 % of a survey-grade mapping platform in the barn. Swap tanks for a global-shutter camera, flash the nozzle driver to time your triggers, and let the IPX6K frame shrug off the dust. Hold 42 m above deck, keep your RTK baseline shorter than 3 km, and you will repeat 3 cm precision week after week—even while the welders keep grinding and the wind keeps howling.

Need the exact parameter file or a second opinion on your site’s wind rose? Message me on WhatsApp—I’m usually on site between pours—and I’ll send the wind-tolerance chart that kept us off the crane.

Ready for your own Agras T70P? Contact our team for expert consultation.