Maverick weeding robot as temporary worker
16 December 2024
No chemical pesticides. No migrant workers. A clean field. Thanks to the Maverick developed by Dutch start-up Odd.Bot. “He identifies every single weed, and effortlessly removes it, even among high-density crops. Faster and cheaper than weeding by hand.”
A 50×50 metre wasteland along the waterfront. In the Merwe-Vierhavens in the port of Rotterdam which now goes under the trendy name Makers District. This is where the Odd.Bot Maverick does its trial rounds. Having now covered 436 kilometres and 1324 rounds. In the field lab, carrots and onions are grown, the most widely grown vegetables in the world. At the ends, you can see how neatly and smoothly the Odd.Bot turns round.
“Over the past few weeks, the Maverick has been doing rounds to see how we can further optimise its flat-field navigation,” says Odd.Bot founder and CEO Martijn Lukaart, watching the self-driving white box. And not just any box. Equipped with AI, two picking arms and a super-precise camera, the Maverick can remove more than 240,000 weeds per hectare with a precision of 2 millimetres. One arm picks up to two weeds per second and replaces about one and a half people. A robot with two arms does the work of three people. Non-stop, day and night, in all weather conditions. “That’s the advantage of a robot. They don’t complain about bad weather or working overtime.”
Odd.Bot founder and CEO Martijn Lukaart with the Maverick weeding robot
Robot delivery service
The Odd.Bot office is on the other side of the field lab. In a huge brick building, a former 1930s energy house. Lukaart employs ten people. Odd.Bot is an odd name that sticks. “Like a computer called apple,” says Lukaart. He’s an entrepreneur: sees opportunities, seizes opportunities, uses them. Combined with a healthy fascination with technology. At high school, he was one of those boys who spent entire afternoons tinkering with mopeds. Refurbishing and selling old ones. Which is why he chose to study Commercial Economics in Rotterdam. “I wanted to learn how to sell new technological product-market combinations.”
Even as a student, he was developing innovative concepts. From a nationwide radio network using Wimax technology to a neighbourhood delivery service. After graduating, he invented another delivery service, this time not by bike but by robot. In collaboration with the lab for autonomous driving at Delft University of Technology. Legislation got in the way of further development. Where should a robot drive, for example? On the pavement or the cycle path?
Weeding robot
In 2019, he heard about a European Commission tender. There was a desire to be able to grow crops using fewer chemicals. For organic farmers. But try weeding all those weeds by hand. Impossible. It takes days and is anything but sustainable, because it involves flying in migrant workers from Romania or Bulgaria to do the work. You see it in the cleaning industry too, says Lukaart. Dutch people no longer want to do that type of work, so help from abroad is needed. Or smarter technological solutions. These often end up in the graveyard of well-intentioned ideas, he says. Because they don’t reflect the needs of companies or end users. “With robotics, it shouldn’t be about resolving staff shortages, but about making tedious work more attractive.”
Lukaart saw an opportunity. “Robotics in arable farming is the future.” He already had a team in Delft and in 2019 was able to start work developing a weeding robot, with six students taking a minor in robotics. An industrial designer, an aerospace engineer, an electrical engineer and several systems engineers. Five months later, their prototype was ready. A trolley with a delta arm and 3D camera that could use artificial intelligence (AI) to identify which plants are weeds and which are not. Lukaart: “Underneath that delta arm, there was a kind of stick blender that crushed the weeds. We still use a similar delta arm, but now it neatly plucks the weeds from the ground as you would in your garden at home. Right at the stalk, root and all, disturbing as little soil as possible.”
Flevoland and Wageningen University & Research
Some time later, he took his first prototype to the Biobeurs in Zwolle. A delegation from the Province of Flevoland were so enthusiastic about it that they gave him a loan from the TMI Proof of Concept Fund Flevoland. Funding to develop an innovative idea to a proof-of-concept phase. The company then partly moved to Flevoland which was home to many organic farmers. He also launched a 4-year PPS Hands-free Production in Agrifood with Wageningen University & Research (WUR) and Kubota, a manufacturer of agricultural machinery with a focus on smart farming technologies.
At the end of 2021, there was a follow up to the WUR collaboration. His Weed Whacker and later the Quirky, the first two Odd.Bot concepts, were given a place at the WUR Farm of the Future in Lelystad. Vision+Robotics’ Ard Nieuwenhuizen was their mentor from WUR. “The Farm of the Future helped us enormously,” says Lukaart. “They measured the performance and established trial plots where they would re-sow a patch every four weeks, so that we could take photos of all the different crop growth stages. First carrots, later onions and chicory.”
Maverick
Lukaart does not have an agricultural background, nor do most of his technical employees. Which might be an advantage, because it forced them to go out and ask questions more quickly. Lukaart: “From the start, we collected feedback from the end users. A kind of co-creation with organic farmers to tweak and adapt the model as much as possible.” The Weed Whacker and Quirky were followed by the first marketable product: the Weed Whacko. A robot with three wheels. This was successfully tested by five growers on as many soils as possible in Zeeland, Germany and three in Flevoland. With carrots, onions and chicory. “In 2023, we also introduced Weeding as a Service using the Weed Whacko, keeping the fields of organic farmers weed-free for a fee per hectare.”
The Whacko became the four-wheeled Maverick. Lukaart: “He identifies every single weed, and effortlessly removes it, even among high-density crops. Faster and cheaper than weeding by hand.” Over the years, he never felt nervous or wondered what on earth he’d started. His biggest fear: that the robot would falter at an important moment, during a demonstration in front of investors or many customers. Nevertheless, the main challenge related to system integration. Vision, KI and delta arm: “How do you ensure that everything is properly aligned? That requires optimal eye-hand coordination, especially when the Maverick navigates bumpy terrain.”
10 percent more onions
In ‘standard’ arable farming, Lukaart is increasingly seeing a move towards chemical-free weeding. This is largely due to a shortage of labour and the growing opposition to the use of chemical pesticides. Not only are they harmful for humans and animals, they are anything but conducive to biodiversity. Furthermore, weeds are becoming increasingly resistant to pesticides. Conventional arable farmers often spray the whole field with chemicals, sometimes doing it again when the crops come up, six times a season. In many cases, the crops themselves also get a hefty hit, at the expense of some of the yield. Lukaart can now say: “When farmers stop using chemicals and choose our robots, they can boost their onion yield by up to ten percent.”
As crazy as this may sound, economies of scale in agriculture are also playing into his hands. Farmers need bigger and bigger machines to achieve this. These press heavily on the soil and harden the topsoil, says Lukaart. Eventually, water can barely seep through, disrupting life in the soil. Roots can no longer establish themselves, soil fauna and nutrients disappear: the balance in the soil food web is disrupted. “Which is how you cause problems for soil health.”
Soil compaction
Precision spraying saves a lot of chemicals but according to Lukaart, that is not the future. Nor does he get excited by the use of lasers. They may be able to identify weed plants at an early stage, but he does not feel they are very energy efficient. With all those additional tools, agricultural machinery becomes even heavier, which in turn compacts the soil. “Our Maverick is lighter than a weeding bed with eight people.” At Odd.Bot, they have covered carrots onions and chicory. They are now considering the next series of crops. But a farmer can also teach the Maverick to distinguish weeds from a crop. If they take photos of that crop throughout the growing season. Which is necessary to feed the algorithm. Lukaart: “You do that until the Maverick can distinguish between that crop and weeds.”
Fingers in concrete mortar
Lukaart is moving from a start-up to a scale-up. He plans to increase Maverick production from 16 to 48 units, and offer the camera and gripper to other system integrators and Original Equipment Manufacturers (OEMs). If they want to hang those tools behind their own tractor or trolley. He is also working with distributors and resellers, and wants to set up a proper service organisation, so that spare parts can be delivered quickly if anything breaks down. The three weed-picking fingers on the delta arm will last for up to a year, he says, which was demonstrated after extensive durability testing in concrete mortar. Lukaart smiles: “Without water, of course.”
