Agricultural robots need to know the field.
Autonomous tractors, weeders and pickers work inside boundaries and around obstacles that someone has defined. Keeping that field data right through the season is what lets them work alone.
- Boundary v3 loaded · RTK surveyed
- Pole buffer 3 m applied
- Wet patch from 2 Nov marked skip
- Crop rows at 4-leaf stage
- Field model shared with farm team
What agricultural robots do.
Agricultural robots take over farm work that needs many hours or many hands: driving, sowing, weeding, spraying and picking. These are the main types in commercial use.
Autonomous tractors
Tractors that work a field without a driver. AgXeed builds the autonomous AgBot, and John Deere has an autonomy kit for both new and existing tractors.
Weeding robots
Robots that remove weeds without herbicide or with less of it. The FarmDroid FD20 sows and then weeds around the seeds it planted. Carbon Robotics' LaserWeeder uses AI to detect plants and lasers to kill weeds.
Harvesting robots
Robots that pick fruit. Dogtooth harvests strawberries on standard tabletop systems in glasshouses and polytunnels.
Spraying drones and treatment robots
DJI's Agras T50 drone sprays crops and generates maps and flight routes. Saga Robotics' Thorvald treats powdery mildew with UV-C light.
Close to this many agricultural robots were sold worldwide in 2024, 6% fewer than the year before, according to the International Federation of Robotics.
The robot knows the job. The field sets the limits.
A robot can be very good at weeding or driving straight lines. Where the field ends, what it must avoid and what has changed since last week all come from the field.
What the robot brings
- Follows a planned path using satellite positioning
- Detects plants or obstacles with cameras or lidar (some robots)
- Repeats a task across a whole field
- Records where it has worked, and in some cases where it sowed
What the field has to give it
- A closed, accurate field boundary
- Headlands, gateways and room to turn
- Fixed obstacles: poles, ditches, trees and wells
- Areas to skip: wet patches and buffer strips
- Crop rows, and how the crop changes through the season
Each machine gets its own field set-up.
Makers of autonomous farm machines give farmers and dealers solid tools to set up a field. The work happens in each maker's own system.
Survey the boundary
Autonomous machines work inside a boundary drawn before the first job. For AgXeed's AgBot, a certified professional maps the field boundaries on-site, keeping clear of roads, ditches and treelines. The map is then stored in AgXeed's TraXwise portal.
Meet the autonomy rules
A John Deere dealer's guide says autonomy boundaries need a fully closed exterior boundary with no gaps, recorded with RTK correction, and must meet Deere's rules for turn radii, headlands and minimum field size. A common failure is that recording wasn't paused around obstacles.
Record the crop
Some robots record the crop themselves. FarmDroid states that the FD20 marks the position of each plant at seeding using RTK GPS with 8 mm precision, so it knows where every seed is when it comes back to weed.
Do it again
Each autonomy system keeps the field in its own software: TraXwise for AgXeed, Operations Center for John Deere. A second brand of machine means setting up the same field again, and changes during the season are mostly known only to the farmer.
The result: the same field is surveyed and drawn again for each brand of machine, and what changes in the season, such as a wet patch, a new fence or a closed gateway, is rarely written down anywhere a machine can use.
Six things a field robot needs.
None of these depend on which machine you use. Held once and kept current, they can serve every machine on the farm.
Field boundaries
Closed, surveyed boundaries with RTK-level accuracy, held once for every machine and every season, and corrected when a hedge, fence or gateway moves.
field 12 · 14.2 ha boundary v3 · RTK surveyed closed · no gaps
Headlands and access
The strip at the field edge used for turning, plus gateways, tracks and where a machine may enter and leave.
headland · 12 m gateway · north-east corner track · to yard 600 m
Obstacles and buffers
Poles, ditches, trees, wells and drains, each with a safe distance, plus buffer strips next to water and hedges.
pole · 3 m buffer ditch · east edge · 2 m buffer well · centre · 4 m buffer
Crop rows and plants
Row direction, spacing and, where it was recorded at sowing, the position of each plant.
sugar beet · rows 50 cm direction · 78° plants · recorded at sowing
Season and condition
Growth stage, crop health, wet areas and damage, updated from satellite and drone imagery as the season goes on.
stage · 4-leaf wet patch · 0.3 ha · 2 Nov → skip on next pass
Change history
Every change to the field as a new version: what changed, when, and which machines and tasks it affects.
2 Nov · wet patch added affects · weeding pass 3 v2 → v3
Fields, rows and glass.
Arable fields
Large, open fields where boundaries, headlands and obstacles decide where a machine can work alone.
Orchards, vineyards and soft fruit
Rows of trees, vines and bushes. Saga Robotics' Thorvald works on strawberries and grapevines, and Monarch lists dairy and blueberry farms.
Greenhouses and polytunnels
Fixed rows and rails, but a crop that grows taller and wider every week. See our post on greenhouse robotics and crop growth simulation.
A defined working area is becoming a rule.
Safety standards and EU law both expect an autonomous farm machine to know where it may work. None of them provides the field data itself.
ISO 18497
The safety standard for partially automated, semi-autonomous and autonomous agricultural machinery, revised in 2024 into several parts, including one on obstacle protection systems.
EU Machinery Regulation
Applies from 20 January 2027. According to CEMA, the European agricultural machinery association, it requires an autonomous mobile machine to travel safely in a defined working area.
ISOBUS and ISO-XML
Machines and farm software exchange task data through ISOBUS and the ISO-XML format, developed across manufacturers by the AEF. Autonomy boundaries are still set up in each maker's own system.
One current model of every field.
Kaiinos does not make farm machines. We build and maintain the field data they work from.
Field model from your data
Built from RTK surveys, drone flights, satellite imagery and the boundaries you already have.
Boundaries and hazards in one place
Boundaries, headlands, obstacles and buffers held once, rather than redrawn for each machine.
Crop and condition layers
Growth stage, crop health and wet areas from satellite and drone data. We have mapped sowing status and crop health, and predicted yield, from satellite data across 1,762 villages.
Kept current through the season
Changes in the field become a new version of the model, with a history of what changed.
One model, more than one machine
Available through APIs and standard formats, so the field work is done once.
Test before the season
Check paths and clearances in a simulation of your real site, including crops at different growth stages.
Start with one season.
Prove it on one farm and one task, then add fields, machines and seasons.
Pick one farm and one task
For example the fields one weeding robot or autonomous tractor works this season.
Build the field model
We combine your boundaries, survey data and imagery into one field model, or survey the fields first.
Line it up with your machines
Your team checks the model against the boundaries already in the machines' own software.
Season updates
Through the season we add what changes, and your team works from one current version.
Frequently asked questions
What is an agricultural robot?
A machine that does farm work with little or no human control: driving a tractor, sowing, weeding, spraying, picking fruit or treating plants. Most work inside a field boundary and follow a planned path.
Why do agricultural robots need RTK?
RTK is a correction service that makes satellite positioning far more precise. DJI gives the Agras T50's hovering accuracy as ±10 cm with RTK and ±60 cm without it, horizontally, with a strong signal. For work between crop rows that difference matters.
What is a headland?
The strip at the edge of a field where machines turn at the end of each row. Autonomous systems such as John Deere's have rules for headlands and turning space.
Does satellite positioning always work in the field?
Not always. Researchers note that it can be unreliable under the crop canopy, so robots working under tall crops often rely on their own sensors as well.
Which rules apply to autonomous farm machines in the EU?
ISO 18497 covers the safety of autonomous agricultural machinery. From 20 January 2027 the EU Machinery Regulation adds requirements for autonomous mobile machinery, including travelling safely in a defined working area.
Does Kaiinos make agricultural robots?
No. Kaiinos builds and maintains the field data: boundaries, hazards, crop layers and changes. You keep your machines and their software.
The same site, other robots.
Your machines can work alone. We make sure they know the field.
Tell us which farm, which machine and which task.
Start a pilothello@kaiinos.ai