Autonomous tractor workflows for row-by-row variable-rate
Set the rate decision, build management zones, assign rates, link prescription to position, then plan row passes and test control.
Variable-rate work starts with an agronomic choice, not a driverless machine. The operator must decide which input will change, where it will change, and what evidence supports that change.
Autonomy comes later. It helps the tractor follow the planned path and repeat the task, but it cannot repair a weak prescription.
Define the rate decision before planning the route
Variable-rate application means changing an input rate as the applicator moves through a field. The first task is therefore to define what will vary by location.
- DJI Agras T100 100 L
- DJI Agras T70P 70 L
- XAG P150 70 L
- DJI Agras T55 50 L
- XAG P100 Pro 50 L
- DJI Agras T50 40 L
| Model | Value |
|---|---|
| DJI Agras T100 | 100 L |
| DJI Agras T70P | 70 L |
| XAG P150 | 70 L |
| DJI Agras T55 | 50 L |
| XAG P100 Pro | 50 L |
| DJI Agras T50 | 40 L |
Sources: ag.dji.com, ag.dji.com, xa.com, ag.dji.com, xa.com, ag.dji.com
That input might be fertiliser, lime, weed control or seed. The source data and the control method may differ, but the core question stays the same: what should the machine change?
Site-specific crop management raises several types of question. Some concern economics, while others concern agronomy, the environment or the equipment. These questions should be settled before an autonomous route becomes the main focus.
A useful starting brief covers the following points:
- the input that will vary;
- the field condition used to set the rate;
- the areas where the rate should change;
- the prescription or sensor rule that will control the change;
- the way machine position will be tied to that decision.
This brief keeps the job concrete. “Apply fertiliser by management zone” gives the workflow a clear purpose. “Use an autonomous tractor for variable rate” only names the tools.
The distinction matters because variable-rate work can be map-based or sensor-based. Map-based VRA uses machine position and a prescription map to change the input. Sensor-based VRA measures soil or crop traits on the move and uses those readings to control the applicator.
| VRA method | What sets the rate | What must be ready before work |
|---|---|---|
| Map-based | Machine position linked to a prescription map | A sound map, desired rates and a way to locate the machine |
| Sensor-based | Soil or crop readings taken on the move | A defined sensing rule and a control link to the applicator |
| Combined | Map data and live sensor readings | Clear rules for how both sources affect the rate |
For a row-by-row autonomous job, map-based planning gives the clearest route from zone design to machine action. A combined system may add live readings, but it still needs clear control rules.
Management zones must come before autonomy
The first step in variable-rate fertiliser application is to set proper management zones. Those zones become the reference layer for the rest of the job.
A management zone groups field areas that should receive the same treatment under the chosen plan. The operator remains responsible for the rate assigned to each area. The tractor’s task is to apply that decision in the correct place.
Map-based strategies can draw on several sources. These include soil type, soil colour and texture, topography, crop yield, field scouting and remotely sensed images. A strategy may use one source or a mix.
Drone maps can contribute to that evidence base, but an image is not yet a prescription. Our guide to turning drone imagery into decisions covers the gap between collecting data and making a field choice. The broader precision agriculture mapping with drones guide explains the mapping side of that work.
For nutrient VRA, the documented map-based process starts with systematic soil sampling and laboratory analysis. The operator then maps the soil nutrient properties of interest. An algorithm turns those mapped properties into a site-specific nutrient prescription.
The prescription map then controls the variable-rate fertiliser applicator.
That chain must remain intact:
| Layer | Its job in the workflow | Failure to avoid |
|---|---|---|
| Field evidence | Show the soil or crop pattern of interest | Treating an image as a rate decision |
| Management zones | Group areas that need a common response | Drawing zones with no link to the objective |
| Prescription | Assign the desired input rate | Leaving the machine to infer the agronomy |
| Position | Place the machine within the prescription | Applying the right rate in the wrong area |
| Applicator control | Deliver the selected rate | Following the route without changing output |
The key test is simple: every zone should lead to an explicit machine instruction. If a zone has no treatment rule, it is only a map feature.
Build the mission from soil, water and topography
The mission plan should turn the zone layer into a field task the machine can repeat. Soil, water and topography are the main factors highlighted in the supplied mission-planning research.
These factors belong near the start of the plan because they help explain field variation. They should inform the zones and rates before the operator lays out the autonomous path.
Water data may also sit within a wider field record. The guide to drones in irrigation and water management is relevant when aerial observations feed that record.
The mission plan acts as the bridge between the prescription and the route. It tells the machine where to travel, while the variable-rate system determines what to apply at each location.
For a row-by-row task, build that bridge in a fixed order:
- settle the treatment objective;
- approve the management zones;
- assign the desired rate to each zone;
- relate the prescription to field position;
- place the planned passes against that field layer;
- confirm that the applicator can read and act on the rate command.
This order prevents route design from driving the agronomy. A neat set of passes does not show that the rate map is sound.
The operator should also decide whether the workflow is map-based, sensor-based or combined. That choice changes what the mission must carry into the field.
A map-based mission needs a prescription and reliable position data. A sensor-based mission needs a continuous stream of readings, a control calculation and a link to the applicator.
The mission is ready only when path and rate can be read together. At any point in the field, the workflow should identify both the intended route and the intended input decision.
Autonomy should execute the plan, not invent it
Autonomous tractor systems bring automation, robotics, guidance and repeated operation to field work. Their strongest role in VRA is consistent execution.
Automation lets a tractor carry out tasks without a person directing every movement. Some systems still need an operator in the seat, while a true driverless system can work without one.
Autosteer shows the narrower form clearly. Once engaged, it takes over driving along a precise path while the operator manages the implement. At the end of a pass, the operator may take control for the turn and then engage guidance again.
This is why autonomy should not be treated as the source of the variable-rate decision. It does not decide whether a management zone is agronomically sound. It repeats the instructions supplied by the wider system.
The best automation jobs are bounded. The grower defines the field, plan and operating limits, while the machine handles repeatable execution. This approach narrows the task until the machine can perform it well.
That principle also helps when choosing the level of automation. A driver-optional system and a driverless system may place different demands on supervision, but both need a valid prescription. Neither removes the need to connect field evidence, zones, rates and position.
For more on how guidance fits within the wider equipment mix, see farm automation beyond drones. The useful question is not how autonomous the tractor appears. It is how reliably the whole workflow carries the planned rate into the correct row and zone.
Test the whole stack before trusting repeat runs
A variable-rate autonomous workflow should be checked as one connected stack. Testing guidance and rate control in isolation can miss a fault at the point where they meet.
Start with the zone map. Each mapped area must have a defined treatment decision based on the selected field evidence. Then check that the prescription carries those decisions in a form the application system can read.
Next, check the position link. Map-based VRA depends on the machine finding its location and relating that position to the desired rate. Without that link, a valid prescription cannot control the correct field area.
The final check is the hand-off to the applicator. The workflow must change the input as the machine crosses the relevant prescription areas. Route following alone is not proof of variable-rate control.
| Check | Question to answer |
|---|---|
| Objective | Is the input to be varied clearly defined? |
| Evidence | Do the zones follow the chosen soil, crop, water or topography data? |
| Prescription | Does every zone carry a desired rate? |
| Position | Can the system relate machine location to that rate? |
| Path | Do the planned passes follow the intended rows and field limits? |
| Control | Does the applicator respond to the prescription or sensor rule? |
| Record | Can geo-referenced data support later review or future work? |
Treat any break in this chain as a workflow fault, not merely a mapping or steering issue. A tractor can follow every pass correctly and still deliver the wrong plan.
The practical setup priority is therefore clear. Define the variable-rate aim, build sound management zones and turn them into a usable mission. Add autonomous execution only after those parts agree.