Market Gardening and Field Crops: When Farmers Share Their Experience with Robots

AgXeed: Robotics to Optimise Arable Farming Operations

Laure Figuéreu-Bidaud, a farmer from Normandy (north of France) - AgXeed user, runs 550 hectares with her husband in a mixed arable and dairy farming operation. Wheat, barley, rapeseed, sugar beet, flax, maize, and alfalfa, a diverse crop mix that generates intense workload peaks and heavy equipment requirements.

Taking over the family farm was an opportunity to rethink everything from scratch. Wide working widths were causing recurring agronomic problems and required levels of tractive power that were never quite sufficient. It was at an open day organised by the local dairy control body, Littoral Normand — “which is quite ironic,” she admits — that she first saw the AgXeed in action. An economic analysis confirmed her instinct: targeting 800 hours of use per year, the robot comes at a cost comparable to a 250 horsepower tractor with 6-metre implements. But where a tractor requires an available driver, the robot works autonomously, “it doesn't need a toilet break or to answer its phone.” At just 3 to 4 metres wide, it covers the same area, simply by working longer.

The machine was acquired for around €300,000, including field mapping, first-year warranty and subscriptions. A “full pack” maintenance contract at €12,500/year over five years brings the cost down to €12.50 per hour, all expenses included. Laure draws a parallel with her Lely milking robot: the same logic of predictable fixed costs, “we budget for it in our cash flow and there are no nasty surprises.”

In the field, the robot is currently dedicated exclusively to soil cultivation. Delivered in July, it has already clocked up 600 hours. The most tangible advantage: being able to seize weather windows that would be impossible to exploit with large machinery. “They were forecasting 30 mm of rain in a single day. We had it working all night beforehand so we could sow right ahead of the rain.” The autumn 2024 sowings were better than with previous solutions.

The logistics require more planning: a 30-hectare field represents 12 hours of robot work, so planning needs to be done on a weekly basis. When asked about the machine's limits after 600 hours, Laure is direct: “So far, we're still looking for them.”

The most unexpected effect came from neighbours. People living near the fields whom Laure had never met stopped to watch the machine working, which opened up conversations about farming practices. “Our big machines frighten them. Their conclusion this time was: oh, it's like my lawn-mowing robot. Just bigger.”

⇒ Watch the replay here
Watch a demo of the AgBot | T2 & W4 - videos in French with English subtitles

Cyclair: A Maize Trial That Invites You to Rethink Your Intervention Windows

Nicolas Bastien, network partner manager at Fermes Leader, tested the Cyclair robot last June on his group's experimental farm near Fontainebleau (near Paris - France). The trial focused on inter-row weeding in maize, on wide strips, with one to three passes of the robot. The objective: to evaluate the machine under real farming conditions.

Overall, the assessment was positive. The optical row detection, one of Cyclair's distinguishing features, raised no issues. “It went perfectly, we were really satisfied with the overall detection,” he confirms. However, a nuance emerged regarding mechanical efficacy: at 4–5 km/h, the robot moves more slowly than a tractor, which reduces the uprooting effect on weeds. On well-developed weeds — which was the case in this trial, conducted slightly late — the difference is noticeable. Nicolas Bastien does qualify this, however: “it's not really a problem in itself, you just need to adapt your intervention windows,” since the robot's low ground pressure actually makes early intervention easier.

A technical glitch was also noted: the front-facing LIDAR sensors were triggered in strong wind, as the maize straightened up in front of the machine and caused unwanted safety stops. The fix is simple, raising the sensors, but the incident illustrates the importance of field feedback in helping machines improve. “These small technical hiccups are important to share, and they're also important for seeing how well the company responds,” notes Nicolas Bastien.

Mickaël Rapin, Project and Digital Solutions Manager at Cyclair, took the opportunity to announce the forthcoming GW model: 5.40 m effective working width, 1 ha/hour output, 20 hours of autonomy. Target price around €200,000 all-inclusive, with the first models expected in the field in spring 2027.

⇒ Watch the replay here
Discover the Rover GW in demo - video in French with English subtitles

Naïo (Oz): Ten Years in the Field, a Relationship Built Over Time

Matthias Carrière, Managing Director of Naïo, retraced ten years of collaboration with Bejo Production, a seed company operating in the Val de Loire and Castelnaudary regions (south of France). A journey that says a great deal about what integrating a robot truly demands.

At the outset, the entry point was not profitability, it was physical strain. Bejo, a salaried-worker enterprise producing organic seeds — onion, leek, and seed-bearing crops among others — was looking to lighten exhausting weeding tasks for its teams. The Oz robot, guided at the time by laser sensors and cameras, was already capable of adequate inter-row weeding on well-developed crops. Matthias Carrière describes the machine as “a blind person with a cane in a vegetable corridor that changes every day” functional, but with limitations.

Everything changed in 2020 with the arrival of RTK GPS. Suddenly, the range of uses expanded: sowing under glass, weeding at very early crop stages, end-to-end crop monitoring. Some crops are now directly established with the robot, which creates a precise map enabling accurate hoeing at a later stage. “We tend to take a cobot approach, that is, a human-robot duo that works in such a way that, at the end of the day, the total work accomplished is greater than if the employee had worked alone,” explains Matthias Carrière.

What this story illustrates above all is the need for mutual adaptation. Bejo has evolved its crop management practices to get the most out of the machine, and Naïo has evolved the machine thanks to feedback from Bejo. “Sometimes each party has to take a step towards the other to ensure the machine runs as much as possible.” Today, the robots work almost every day, two hours a day, covering 10 to 20 hectares of weeding per year.

The financial returns are beginning to take shape. An aromatic herb grower estimates a gain of €2,000 per tunnel per year: switching from transplanting to direct sowing, one additional cut per cycle, less labour. But behind the figures, there are also human stories: that farmer who “was on the verge of giving up, who could no longer get down on all fours to weed,” and who, five years later, is still farming.

The Oz robot is priced between €30,000 and €40,000 depending on the specification, with a second-hand market beginning to take shape for earlier models.

⇒ Watch the replay here
Discover Oz in demo

Ecorobotix (ARA): Spray Less, But Exactly Where It's Needed

Originally developed in Switzerland to address thistle and dock problems in grassland, Ecorobotix's ARA technology quickly found other areas of application. Aurélien Garnier, Regional Sales Manager for France and Belgium at Ecorobotix, and Dirk Den Bakker, sales representative at Stecomat (ARA distributor) and a farmer in the Tarn-et-Garonne (south of France) department, presented its uses during the session.

The principle relies on millimetre-level canopy vision: cameras coupled with artificial intelligence identify each plant in real time and trigger a localised application to a 6 × 6 cm zone. This level of precision allows the use of non-selective herbicides in the immediate vicinity of crops, thanks to a calculated safety zone around each plant. The result: up to 95% savings in product volume, depending on the situation.

But beyond the reduction in spray volume, it is crop health that stands out as the primary benefit. “The reduction in spray volume is the first indicator. But today, the most important thing is plant health,” insists Aurélien Garnier. On onions — the entry crop in France, with now 80% of surfaces treated using the technology — yield gains reach 15%. On chicory, a cooperative in northern France recorded up to 3 additional tonnes per hectare during the 2025 campaign, by reintroducing productive varieties that had previously been abandoned because they were too sensitive to phytotoxicity from conventional herbicides.

For 2026, new algorithms are in development (garlic, potato, tomato, nuts edge, groundsel) and a 12-metre version has been announced to improve throughput in arable farming. The current machine retails at €160,000, with an annual licence fee of €6,500 covering three algorithms of the user's choice and all updates, “which ensures farmers are always up to date, whether they bought the machine in 2026 or in 2021.”

⇒ Watch the replay here
Discover ARA in demo - video in French with English subtitles

Sabi Agri: The kit Robot, for Those Who Want to Understand What They're Buying

Alexandre Prévault Osmani from Sabi Agri presented an approach that stands apart from the others: the SRBC robot is sold in parts, on a pallet, via their online shop torque.works. The farmer assembles it themselves.

The idea is not purely economic — although the kit starts at €13,500 — it is philosophical. “One of the drivers of agricultural robotics adoption is that the entire ecosystem understands how it is used, what its limitations are, and what the technology is based on.” For Sabi Agri, a farmer who has assembled their own robot knows where to look when something goes wrong and can replace a part without putting the machine out of action for days on end.

Assembly takes one day for a beginner and between 2.5 and 3 hours for someone familiar with the machine. A step-by-step guide, YouTube videos (on the Torkworks channel), and foolproof connectors make assembly accessible without any specific training. Once assembled, the robot runs through a series of tests via the mobile app, validated remotely by Sabi Agri before use is unlocked.

For positioning, the machine can connect to the Centipede network — a free, community-based centimetre-level GPS correction network — or to private services. The warranty covers parts, and all parts are available individually on the website.

⇒ Watch the replay here
Discover the SRBC robot in demo

What These Testimonials Have in Common

Beyond the specific features of each robot, several common threads run through all of these accounts.

A robot is not a tractor you start up and leave to run. It requires an adaptation period, sometimes a lengthy one, a revision of crop management practices, and more forward-looking organisation. But in return it opens up possibilities that conventional machinery cannot offer: intervening early, working at night, seizing a narrow weather window, reducing physical strain without eliminating jobs.

Physical strain recurs in almost every account as the primary driver of adoption, often ahead of financial returns. And the economic benefits, when they do materialise, build up over time as the farm adapts its practices to the machine and the machine evolves thanks to feedback from the farm.

Finally, several contributors said it in their own way: the robot also changes how others perceive farming. Neighbours, employees, and sometimes even fellow farmers who were initially sceptical “and then thought to themselves: well, I'll be damned, the thing actually works.”