PRF4Mechanical robotic weeding

What it actually is

You grow a row crop, roots, field vegetables, salads, or soft fruit outside the permanent kinds, and the weeding in it is done by a mechanical robot: camera or GPS guided hoes that physically remove the weeds from the ground, between the rows and within them. A pre-emergence residual herbicide is allowed at or within days of the crop's establishment; after that, no herbicides on the crop from sowing to harvest. The action applies to at least one cash crop sown in each of the three years, followed from sowing to harvest. GOV.UK's stated aim is weed control with minimal herbicide use, supporting productivity, water and soil quality, biodiversity and integrated pest management, and the page notes it will not usually be possible to do this action on combinable crops.

Source: GOV.UK, PRF4 action page, checked 23/08/2026.

What it pays, and what that comes to

The rate is in the panel at the top of this page. Multiplied out:

Source: arithmetic on the payment rate shown at the top of this page, which comes from Tilth's copy of the GOV.UK action catalogue and carries its own last-checked date. The 15 hectares is an illustration.

Is it worth it on your ground?

Step one: what the technology costs. Like the rest of the precision family, the payment was never sized to buy the machine. Camera-guided robotic hoes are serious capital or a contractor service with rates to match, and the honest way to read the rate is as a contribution towards running one, not a route into owning one.

Step two: what it replaces. In a veg rotation the herbicide programme this action removes is itself expensive, and hand-weeding, the other alternative in salads and roots, is more expensive still where labour can be found at all. The robot's real competitor on most veg farms is the gang, not the sprayer, and that comparison is entirely in your own cost records.

Step three: where that leaves it. For growers already running or hiring robotic hoes, the action pays for what the system already does, plus the discipline of the no-herbicide window. For everyone else, the sum starts with a machinery or contracting decision this page cannot make, and the payment is the smaller term in it.

Source: the payment rate at the top of this page. The judgements are Tilth's own reading rather than scheme guidance.

Does it work, and what makes it work better

This section is what we have found in published sources about the practice this action pays for. It is not scheme guidance, it is not Defra's or the RPA's view, and it adds no requirement beyond what the GOV.UK action page sets out. We have named every source and given the date we checked it. What it means for your ground is your decision.

Robotic weeding is too new for the graded evidence assessments this site leans on elsewhere, and the best recent trial we have found tests the neighbouring technology rather than this one: laser weeding, which is PRF3's method, not the mechanical hoes this action requires. For what it is worth across the family, those 2024 trials in New Jersey and New York state, in beet, spinach and pea, found robotic laser weeding as effective as or superior to label-rate herbicides on erect annual weeds, with weed biomass down at least 97 per cent against untreated ground by season's end, crop stunting no more than 1 per cent, and crop biomass up at least 30 per cent where the robot replaced the herbicide programme. The weakness found was sequentially emerging weeds like purslane and annual grasses with protected growing points, and both weaknesses would face a hoe as they faced the laser, since the machine can only kill what has emerged where it can reach.

Source: Sosnoskie and others (2025), Deep learning-based laser weed control compared to conventional herbicide application across three vegetable production systems, Pest Management Science, checked 23/08/2026. The transfer of the weakness to mechanical hoes is Tilth's own reading, not a finding of the paper.

Mechanical inter-row hoeing itself is hardly new; what is new is the guidance and the in-row reach, and we have not found a peer-reviewed trial of the current camera-guided in-row machines to set beside the laser study. One trial, one season, one country either way: a promising file, not a settled one.

What you actually have to do

Apply the action to at least one cash crop of root crops, field vegetables, salads or non-permanent soft fruit sown in each of the three years, following each crop from sowing to harvest, the third year's crop to harvest or the action's end date, whichever comes first. Where standing crops make the first year impossible, apply it to a crop sown within 12 months of the start date. Weed the crop with a mechanical robot guided by cameras or GPS whose hoes physically remove weeds, between and within the rows. Use no herbicides on the crop from sowing to harvest, other than a pre-emergence residual applied at or within days of establishment.

Source: GOV.UK, PRF4 action page, checked 23/08/2026.

Where it can go

Arable land below the moorland line growing the eligible crops as cash crops, or lying fallow with the FA01 code, registered with arable land cover and compatible land use codes. SSSIs need Natural England's consent before agreement. Land with historic or archaeological features needs an SFI HEFER first. Rotational or static: the action can move with the rotation or stay put.

Source: GOV.UK, PRF4 action page, checked 23/08/2026.

What you can stack it with

The GOV.UK page lists compatible actions across SFI26, SFI 2024 and SFI 2023, plus Countryside Stewardship options, with the usual four boundary actions, CHRW2, BND1, BND2 and WBD2, on the parcel edges. The overwinter stubble and cover crop actions appearing on the same-area lists fit the veg rotation's gaps rather than its cropped months; read the lists on the page for your own combination.

Source: combinations, GOV.UK PRF4 action page, checked 23/08/2026. The fitting note is Tilth's own reading.

Things worth watching

This section is Tilth's own reading of where this action is easy to get wrong. It is not scheme guidance, it is not sourced from Defra or the RPA, and none of it adds a requirement beyond what the GOV.UK action page sets out.

The no-herbicide window is absolute after establishment. One pre-emergence residual at or within days of establishment, then nothing until harvest. A flush the robot cannot catch in a wet fortnight has no chemical backstop; the plan for that fortnight is the robot going round again, and the machine capacity to do it is worth confirming before the crop is entered.

One crop a year is the floor, and it must be a cash crop. The action follows a commercial crop from sowing to harvest in each year. A trial strip does not qualify, and the year the rotation gives you no eligible crop on the entered land is a problem to see coming at application time.

Combinable crops are effectively out. The page says it will not usually be possible on combinable crops, which matches what the machines are: row-crop tools. This is an action for the veg and roots end of the farm, not the wheat.

The robot's log is the compliance record. The required evidence is built around the machine's job recording: weeder identity, dates, areas. A contractor's machine means the contractor's data, and getting that export agreed before the season starts is easier than reconstructing it after.

Keeping the evidence

Also Tilth's own reading. GOV.UK sets the requirements; what follows is a practical suggestion about record keeping.

The tractor or weeder job data with machine, dates and areas, field operations records with parcel references and sowing and harvest dates, and contractor invoices where hired. As with the rest of the precision family, the machine writes most of the record itself; the work is filing it against the right parcels and keeping the spray record that shows the window stayed clean.

Source: evidence requirements, GOV.UK, PRF4 action page, checked 23/08/2026. The suggestion is Tilth's own.

The part only your own ground can answer

What the robot saves you is the whole sum: the herbicide bill, the hand-weeding gang, the passes, against the machine or contractor cost across your hectares. Those numbers live in your own operations and cost records, crop by crop, which is what Tilth keeps. Request an invite.