Autonomous drone network

On this page 12

Docked, charged, and out again at first light

Drones launch from charging stations on their own schedule, cover the fields and dock again without a driver.

CategoryOperate
Slugautonomous-network
CadenceDaily, weather permitting, with the calendar published a week ahead
PayloadDock weather station, aircraft sensor payload per mission, redundant RTK
Needs a visit firstNo, but a dock has to be sited and installed

The problem

A pilot in a van is the bottleneck. Travel time between blocks decides how many fields get flown in a day, which means the outlying ones get flown least and the schedule collapses the moment the weather closes a window.

How it works

StepWhat happens
StationWeatherproof docks are placed to cover your blocks within a single battery
ScheduleEach field has its own cadence; the network plans the day around weather, wind and daylight
Fly and dockAircraft launch, fly their route, return and charge without anyone attending
EscalateAnything the models are unsure about goes to a human reviewer before it reaches you

What you get

  • Daily coverage log
  • Missed flight report with the reason
  • Fleet and battery health
  • Escalation queue

What the dashboard measures

  • Flights completed against planned
  • Hectares per station per day
  • Weather cancelled flights
  • Battery cycles and dock uptime

Where it matters most

Cooperatives, contractors, maize, winter wheat, and above all dairy and grazing, where the daily patrol is the product.

Build it

The regulatory work is larger than the engineering work. Read Regulation before you cost this.

What a dock actually has to do

FunctionWhy it is hard
Open and closeA moving lid outdoors, all year, in wind and ice
Precision landGNSS alone is not enough. Fiducial marker plus a downward camera, or IR beacon
ChargeContact pads or wireless. Contact is cheaper and needs alignment
ConditionHeating and drying, or a wet aircraft ices up on the next dawn flight
Report weatherThe go/no-go decision has to come from the dock, not the forecast
Fail safeA dock that cannot close must not launch. A drone that cannot dock must land safely
ConnectLTE, with local autonomy when the link drops

Buy or build

TurnkeyDIY
ExampleDJI Dock 3 via Solectric or AdvexureEnclosure, actuator, charger, marker, compute
Cost€15,000 to €40,000 including aircraft€2,000 to €6,000 plus your aircraft
Time to first flightWeeksMonths
Regulatory supportManufacturer documentation helps the authorisation caseYou write the safety case yourself
Sensor choiceTheir payloads onlyAnything you can build

Buy the first one. The value of the capability is the schedule it keeps, and proving that with somebody else's dock is far cheaper than discovering your lid mechanism fails at -5 °C in February. Build docks later, for payloads the turnkey systems do not carry.

DIY dock hardware

ItemEU sourceUS sourceIndicative
IP66 enclosure, 800 mm classRittal, Fibox via ReicheltDigiKey€300 to €900
Linear actuators for the lidIgusServoCity€150 to €400
Charging contactsSpring pins, or a WiBotic wireless systemsame€50 to €3,000
Precision landing markerPrinted AprilTag plus an IR beaconsame€50
Dock computeRaspberry Pi 5 at BerryBaseThe Pi Hut€90
Weather stationDavis Vantage Pro2, or an anemometer plus SHT45same€80 to €1,200
Router with failoverTeltonika RUTX11SparkFun€200
Solar and battery, off gridVictron via AntratekRobotShop€600 to €2,500
Heater and dehumidifierReicheltDigiKey€60 to €200

In-house software

StageWhat we runLicenceReplaces
AutonomyArduPilot or PX4 with a companion computer running MAVSDKGPL / BSDvendor-locked flight stacks
Precision landingArduPilot's precision landing with AprilTag detectionGPL
Dock controllerState machine: closed, opening, ready, launched, returning, charging, faultown code
Fleet schedulerExtend ScheduleCapabilityFlights with weather gating, battery state and daylightthis repo
Weather gatingWind at altitude, precipitation, visibility and light, from the dock plus DWDown code
Escalation queueLow confidence findings routed to a human before they reach the customerthis repo
Health telemetryBattery cycles, motor current signatures, dock cyclesown code

The scheduler in this repository already does the cadence half of the job. What a network adds is the constraint solver: which aircraft, from which dock, in which weather window, with which battery, and what to drop when the day is short. Model it as a scheduling problem with hard constraints (battery, daylight, airspace) and soft ones (cadence, priority), not as a queue.

Fail-safe posture

Write these down before the first unattended flight, because the authorisation will ask:

  • No launch unless the dock reports closed-and-healthy, wind under limit, precipitation nil, and the return battery reserve is met at the far end of the route.
  • Link loss means continue the route and return, never hold.
  • A failed dock means divert to the next dock in range, or land at a pre-surveyed safe area.
  • Any geofence breach is an immediate return to launch.
  • Every one of these decisions is logged and shows up in the missed flight report with its reason. "Weather cancelled" with a wind figure attached is a report a customer trusts.

Cost efficiency

  • Site the docks by battery, not by field. One dock that reaches four blocks beats four docks. The siting exercise is the highest-value hour in the whole deployment.
  • Grid power if you can possibly get it. Off-grid solar for a dock that charges an aircraft several times a day is a large battery bank and a real cost. A yard, a pump house or a mast base with power is worth a detour.
  • Start with a daily patrol, not a survey. Livestock and fence checks justify a dock on their own and use a light payload. Add the survey capabilities to the same dock afterwards.
  • The escalation queue keeps a human in the loop cheaply. One reviewer can clear a day's ambiguous findings across many farms in an hour, and that is what lets you promise autonomy without promising infallibility.