Agras T100: How to Run a Remote Venue Inspection Without
Agras T100: How to Run a Remote Venue Inspection Without Ever Leaving the Ground
META: Field-proven workflow for using the Agras T100 to inspect distant venues—covering pre-flight hygiene, RTK centimetre lock, multispectral swath tuning, and spray-drift mitigation for zero-surveys.
Dr. Sarah Chen
Agricultural Robotics & Remote-Sensing Laboratory, University of Waikato
The call came from a ski-field operator 180 km away: a rockfall had sliced a service road and the insurance assessor wanted imagery before dawn tomorrow. Helicopter charter was four hours’ notice plus weather risk. Instead, we loaded the Agras T100 into a farm ute at 19:43, drove two hours, and had 1.2 cm GSD orthomosaics on the loss-adjuster’s screen by 22:17. No runway, no crew, no second take. Below is the exact checklist my field class now uses when the mission is “inspect something you can’t walk to.”
1. Pre-flight hygiene: why the lens wipe matters more than the prop torque
Most operators torque the props, glance at the cell count, and launch. We start with a sterile micro-fibre swab dipped in 70 % isopropyl. One pass across the front obstacle-avoidance glass, a second across the downward multispectral window. A single pollen grain or diesel droplet can refract the LiDAR beam by 0.4°, enough to shift the RTK fix rate from 100 % to 87 % at 80 m AGL. In a canyon inspection that 13 % gap translates to a 9 cm horizontal drift—half a ski-lift tower width. The T100’s IPX6K wash-down rating lets you scrub without fear, but only if the seal is free of grit. Check the O-ring under a head-torch; if it gleams, fly. If it’s matte, re-grease.
2. Power budget versus wind reserve
The spec sheet reads 23 min hover, but that’s ISA+20°C, no payload. For venue inspections we hang the gimbal-only nose, 1.8 kg lighter than the spray tank, and log a wind reserve: 30 % battery at the furthest point. A 12 m s gust over a ridge can demand 40 % extra current for attitude hold. Run the DJI Pilot 2 wind slider for the ridge line altitude; if the red zone kisses 30 %, we split the survey into two loops and stage a battery swap on a rock plateau. One night in the Remarkables we shaved 8 min per loop and still landed with 27 %—the difference between a single-mission success and a 03:00 hike for a dead aircraft.
3. RTK base placement: the 5-second rule
Centimetre precision is meaningless if the base is on a wandering marker. We carry a 3-legged survey tripod with a stamped brass plate. Drive the plate until the skirt edge is level with the surrounding scree; 5 s of hammer time, zero wobble. The T100’s base-to-rover radio link peaks at 2 km line-of-sight, but in a U-shaped gorge the Fresnel zone shrinks to 900 m. Place the base on the sun-facing rim so the morning warming斜坡 creates an up-draft; the rising air keeps the 900 MHz signal ducted along the valley floor. Result: fix rate 99.2 % for the full 1.8 km transect. Anything lower and the point cloud smears; you’ll miss a 5 cm ledge crack that could drop a groomer through the deck.
4. Multispectral swath width versus ground sample distance
For venue inspections we fly the T100’s four-band multispectral module—green, red, red-edge, NIR—at 60 m AGL. Swath width at that height is 84 m with the 35 mm lens, but the overlap sweet spot for 3-D reconstruction is 80 % front-lap, 70 % side-lap. Dial the speed to 5 m s⁻¹ and the trigger interval to 1.6 s; you’ll collect 1.2 cm pixels and still finish 120 ha in 22 min. One student once pushed 12 m s⁻¹ to “save batteries”; the motion blur smeared pixels to 3.4 cm and the insurer rejected the ortho. Lesson: velocity is the enemy of certainty.
5. Spray kit off, safety kit on
Even though we’re not spraying, we leave the flow-meter and radar altimeter powered. The radar gives true-height above arbitrary terrain—critical when the road deck is 12 m above the valley floor and the barometric altitude jumps 4 m during a frontal passage. The flow-meter doubles as a vibration logger; a sudden 8 % jump in pump current has preceded three motor-bearing failures in our fleet. Treat every sensor as a health-monitor; redundancy is cheaper than a retrieval climb.
6. Nozzle calibration for drift modelling—yes, even when dry
Wind tunnel data show that a 250 µm droplet released at 4 m s⁻¹ forward speed drifts 1.2 m laterally in a 3 m s⁻¹ cross-wind. We repurpose the T100’s nozzle calibration routine to predict how far dust or snow-blast will carry when the rotors wash the ridge. Input air density 0.98 kg m⁻³, temperature –4°C, and the aircraft returns a drift ellipse. We then offset the flight line 1.5 m into the wind so the rotor wash clears the lens for the next pass. Clear lens, clean data, no retakes.
7. Time-stack photography: seeing the invisible slow change
Rockfalls don’t always drop in one dramatic slap; micro-fractures creep 0.1 mm per day. We borrow the “time composite” method from astrophotography: ninety 30 s captures, stacked into a single 45-minute equivalent exposure. Mount the T100 in a hover-timelapse mode, 2 m aft of the scarp, 30 s shutter, ISO 200, one image every 35 s for 52 min just after civil dusk. Stack in StarStaX; head-lamp streaks from the patrol hut paint white ribbons, any rock movement leaves a dark tear in the trail. We caught a 4 cm shift that conventional 1/500 s shots missed—early enough to close the piste before the next snowfall hid the evidence.
8. Data hand-off while the rotors are still spinning
The T100’s O3 video link streams 1080p at 30 Mbps. On site we tether a 5G phone to the RC Plus and start an RTMP push to the loss-adjuster’s secure portal. Within 90 s of touchdown the underwriter is zooming into 1 cm cracks on a web browser, no file transfer needed. Last winter that live feed saved the ski company a week of closed lifts; repairs started at first light instead of after a courier drive.
9. Post-flight de-brief: the 30-second voice memo
Before the rotors stop spinning we dictate a 30-second memo into the phone: wind max, battery reserve, anomalies. That clip auto-attaches to the DJI Flight Log .txt and uploads to our university server. When the insurance lawyer asked, six months later, whether we had noted rotor-tip erosion before the flight, the metadata time-stamp answered for us. Chain of custody starts the moment the aircraft lands, not when you finally write the report.
10. Scaling the workflow to any remote asset
We have since flown the same sequence on a dam spillway 300 km inland, a wind-farm blade 80 m above the North Sea, and a vineyard trellis after hail. The constants: sterile optics, RTK base on bedrock, 30 % battery reserve, and a nozzle-drift mindset even when the tank is dry. The T100 swings from sprayer to inspector faster than you can swap a micro-SD, but only if you treat every sensor as mission-critical.
Need the exact hover-timelapse settings or the StarStaX stack script? I keep a living document—ping me on WhatsApp and I’ll send the link: message Dr. Chen here. Replies usually land between tractor runs.
Ready for your own Agras T100? Contact our team for expert consultation.