From Beta-testing to Integration: How Viticulture is Adopting Robotics

Naïo's TED in the Nursery: Four Seasons to Adapt, Build and Achieve an Optimised System

Sébastien Viguier is a nursery grower in Béziers (south of France), third generation, specialising in fruit tree production and varietal development. He has been using the Naïo's TED robot for four seasons and is now starting his fifth.

He chose robotics out of conviction first "in this business, research and innovation are the future" and out of necessity second. Weeding in a nursery is an thankless task, difficult to delegate to staff yet absolutely essential: before the robot, Sébastien carried out between five and seven chemical weeding treatments per year. "At this point, after four years, on my first-year shoots, I'm at zero chemical weeding."

He did not seek to compare the robots available on the market. It was his distributor, T3M Lavail (Naïo's south-west of France dealers), who suggested he try the TED. He tested it, adopted it, and never looked elsewhere. What this journey illustrates is the crucial importance of the relationship with the distributor in a sector where decision-making remains difficult without real hands-on experience.

The first season was devoted to getting to grips with the machine and adapting the implements. The TED is delivered with standard implements, but the specificities of the nursery quickly required modifications: plantings are at 1.60 m spacing, the mapping of annual crops has to be redone each year, and there was a need to work the entire row and not just the 30-40 cm at the base of the tree. Sébastien built his own electric intercepts, added tines, and kept his own discs. "I built my setup gradually." One hundred hours in the first year, 150 in the second, and by the fourth season, over 300 hours with two employees dedicated to operating the robot. "I no longer need to deal with it myself."

The progression is clear: "Third year, we start to get a good handle on it. We adapt the crop, the irrigation, we put the headlands in place." It is not just the robot that evolves, the entire cropping system reorganises itself around it. Today, the TED works 700,000 trees over a year, covering 3 to 3.5 hectares per day, which means 60 to 70,000 trees weeded at the base daily.

The estimated economic return: around €20,000 in annual labour savings, plus €10,000 in eliminated chemical costs, not counting tractor fuel. But as with other speakers during the day, the labour saved has not been cut — it has been redirected towards improving tree quality. "I haven't saved it because I'm doing something else with it." And the agronomic benefit is equally tangible: "We work the soil much more, much more regularly, and as a result the plants grow better."

Perhaps what best summed up the journey was the reaction of the employees. Initially sceptical "at the beginning, they struggled to accept the robot, they told me: we won't have any work left" they now ask to use it themselves. "Sometimes when we're on plots where the men are working ahead of it, they say to me: can't you bring the robot through, it's easier."

⇒ Watch the replay here
Discover Ted in demo

Pellenc RX20: A Robot That Adapts to Multiple Contexts

A Large Estate in the Languedoc: Towards a Robot Fleet

Stéphane Verdaime, technical manager at Pellenc SAS, supported a beta-tester client on a 200-hectare wine estate in the plain between the Corbières and the Minervois (south of France), of which 130 to 140 hectares are potentially suitable for robotisation. Winter and early spring were devoted to soil cultivation and mechanical under-row weeding across 120 hectares. Late spring provided an opportunity to test mowing, managing grass height to prevent it from blocking the frames when too tall. In autumn, a 20-hectare mission was launched over two days to validate all the improvements made in the meantime by the manufacturer.

The client's objective is clear: to run the robot as a fleet "at least two, perhaps three robots." The reason is as much agronomic as human. "A driver doing mechanical weeding, if he's a little distracted or turns his head slightly, he's no longer centred in the row. A robot is always centred." And faced with the growing difficulty of recruiting and retaining drivers, the logic is compelling: "The few drivers who remain on estates will carry out more delicate tasks, where the human eye is needed. Everything that adds no value will be done by robots. It's obvious."

Narrow-row Vines in the Médoc: The Robot to Preserve Soil Structure

Clément Cazeau, distributor at Pellenc Bordeaux Charentes (south of France), conducted his beta-test on around thirty hectares of narrow-row vines with 1.50 m row spacing in the Médoc area (south of France). The use was exclusively mowing, a strong expectation from the client, whose straddle tractors create ruts in the clay soil, making any subsequent pass difficult or even impossible. "The robot was brought in to manage all of this mowing area."

Fuel consumption recorded: between 2 and 2.5 litres per hour. This test also helped to refine the specific constraints of narrow-row vines: "Old vine stocks, the shape of the stocks, the thickness of the trellis wire, vines that taper to a point" all particularities that fed into the development work on mission generation and guidance.

A Cuma in the Sarthe: Testing Shared Use

Pierre Pichet, agricultural equipment coordinator at the Union des Cuma des Pays de la Loire (west of France), led the trial as part of the European Azimut project, working with a Cuma in southern Sarthe bringing together wine growers with varied practices: organic and non-organic, cultivated and uncultivated soils, with and without slopes. The objective was not to test the robot in ideal conditions, but to put it through its paces "to the very end" in conditions representative of a Cuma: several producers, several plots, several soil types.

The choice of the RX20 was down to its compactness and manoeuvrability, particularly well suited to century-old vines "which are not necessarily planted by GPS" and whose rows can be curved. Its ability to lower the safety roll bar for loading into a van was highlighted as a decisive logistical advantage for the Cuma model.

The organisation started with a collective meeting: who wants to test it? Then Ludovic Patte, the local distributor (LVVD),carried out the field surveying plot by plot, accompanied by Pierre. "Depending on the length of the plot and the parallelism of the rows, we sometimes redid the surveying in the middle of the row so that the robot could find its bearings." A crucial step, on which everything else depends.

More info about the CUMA's trial here - in French

A Well-established Protocol Starting with Plot Surveying

Surveying is one of the key elements in ensuring the robot stays within the plot. "When we survey the plot, we first survey the perimeter, then the rows, the alleys, the exclusion zones," explains Clément Cazeau. This geo-fencing, developed in partnership with Agriculture Intelligence, guarantees that the robot operates within a known and secure space, independently of the onboard safety sensors (LIDAR, obstacle detection). As a result, the robot does not exceed the virtual boundaries defined during this step and cannot come within one metre of fences or plot edges.

In the field, implementation at Stéphane Verdaime's client followed a well-established protocol: an employee dedicated to the robot brought it to the plot each morning, selected the mission in the application, launched the robot, and went about other tasks in the vineyard.

Strengths and Limits of the RX20: Three Witnesses Agree

Feedback converges on several strengths: the compactness and weight of the RX20, ease of transport, and reasonable fuel consumption, around 1.5 to 1.8 litres per hour depending on implements and conditions. "It is capable of climbing steep slopes. That was really impressive," notes Pierre Pichet, who also highlights the contribution of the three-cylinder combustion engine coupled with the electric generator in demanding situations.

The limitations also align. On heavy soil, on a slope with a cross-fall headland, if the robot slips more than around ten centimetres off its line, it stops and waits for human intervention. Clay-induced ruts can disrupt the heading. And insufficiently lifted vine shoots slow progress by triggering the LIDARs. These are, as Florent Georges summarises, "fairly common hazards in the agricultural environment."

The Economic Question: Leasing Rather Than Buying?

On profitability, Stéphane Verdaime puts forward a proposition that reflects the reality of the market: "The robots of today will not be the same in three or four years. Leasing allows the investment to be spread over three years through rental payments, while ensuring renewal towards technology that is always up to date." A logic that avoids ending up with obsolete equipment still with several years of depreciation ahead of it. For the Cuma, the announcement of a leasing option from Pellenc was, according to Pierre Pichet, "a relief", a step towards committing to the activity without too heavy an upfront investment. Clément Cazeau, for his part, observes that price is not the primary obstacle expressed by clients: it is rather unfamiliarity with the technology and apprehension about its complexity that demonstrations progressively help to dispel.

⇒ Watch the replay here
Discover RX20 in demo - video in French with English subtitles

From Promise to Field Reality: The Real Challenge of Integration

These testimonials confirm an underlying trend: viticultural robotics is no longer a prospective subject. It is a real tool, with its strengths and its limits, that demands thoughtful integration.

Surveying is an essential step. It requires time, rigour, and a detailed knowledge of the plots, it is often the distributor who takes charge of it, and their proximity to the grower becomes a key success factor.

Building expertise is gradual and non-negotiable. Sébastien Viguier took four years to fully optimise his system. The RX20 beta-testers adjusted their missions, speeds, and headlands over successive seasons. "Just because you have a robot doesn't mean you've cracked it. There is an enormous amount of work involved."

Finally, the question of shared use through Cuma remains open. The model exists, the desire is there, but organising work between several producers across several plots is a specific challenge that neither the robot nor its software can resolve alone. It is as much a human undertaking as a technical one, and probably one of the sector's next frontiers.