Ekostartup Ullmanna He made himself known for the first time at the end of last year, when with his clever shoulder won a silver spot in the CzechInvest Startup Challenge competition. In the following months, the company worked on filing a patent and got into the Swiss accelerator Mass Challenge.
“We want to scale organic production with artificial intelligence and robotics to make production greener and more sustainable,” explains the company’s co-founder Martin Ullmann. So far, Startup has produced a prototype of a robotic hoe, on which it tests the technology, and is gradually looking for the first customers.
“It is a row hoe that does not pull out the weeds, but with the help of knives it digs up the soil, so it basically hooves, as if someone takes a hoe and digs up the weeds, which then dries out. It is important that the machine does not primarily eliminate weeds, but creates conditions that prevent its formation. At the same time, it supplies other nutrients to cultivated crops. This has been done for centuries. By hoeing around the plants, the soil is soaked, so more oxygen gets into it and at the same time not so many weeds grow in those places. Hoeing also prevents water from evaporating. In addition, the cut weeds remain in the field and form more biomass after their decomposition. Thanks to our machine, the use of herbicides in conventional agriculture can be reduced by up to 80% and pesticides by up to 40%. ”
Weed control for farmers is one of the most important activities. Harmful species suck nutrients out of the field and overshadow crops. The yields from the weedy field are thus much lower than they should be. “Classical and especially smaller farmers solve this problem by using chemistry, which, however, cannot be considered a long-term sustainable method. On the contrary, the mechanical method is sustainable and it is actually a return to what was done forty or fifty years ago, “says Martin Ullmann.
When an ajťák and an agronomist meet
Martin Ullmann co-founded the company in 2019 with his father-in-law Jindřich Ullmann. “I am an Ajťák and my father-in-law is a farmer, an agronomist and at the same time an expert on sugar beet. One of the well-known farmers came to him with the idea that he would like to grow organic sugar beet, which requires mechanical hoeing. They figured out how to do it and that the machines that are available are very expensive and from abroad. And I was thrilled by the idea of connecting IT and agriculture, where we would machine-detect the crop, ”recalls Martin Ullmann. In the end, he came to the conclusion that the problem of weed control could be solved through artificial intelligence that would control the shoulder knives.
“Our equipment connects behind the tractor to a classic three-point hitch, just like any other agricultural machine. Due to the use of our hoeing machine, the farmer does not have to adjust anything complicated and can connect it to the hinge, which he uses for everything else. He just sets the crop he wants to dig on his shoulder, goes around the field and watches how it works. The hoe can be two-, six- or twelve-row and always has a pair of shoulder knives, we call it a cultivation unit, and each row contains a camera, a computer unit for controlling the knives and then the knives themselves. The camera scans the soil in a row and sends the image to a computer, which searches for selected plants, such as sugar beet. The computer evaluates whether it sees beets or where it is located and how fast the machine moves, and with this information in mind, it controls the shoulder knives, which open in front of the plant, bypass it so as not to damage it, and close behind it again. “
To make the device easier to develop and operate, the computer only detects selected crops, not the weeds themselves. “It’s a much simpler solution than smart spraying, which detects weeds and sprays chemicals where they can be found. But this means that they must be able to recognize different types of weeds and select the appropriate spray for each. At the same time, the weeds must be large enough to be captured by the camera. On the other hand, when we hoe, weeds may not have climbed out of the soil yet, but we will suppress it. So it’s much more efficient, and if done right, no weeds should even grow there. “
Crop detection works on the basis of machine vision and machine learning. Startup has created datasets of several crops. He collected the largest number of pictures for the mentioned sugar beet. “But we also work with other crops. We tested cabbage, we have data for pumpkins and then, for example, for garlic. It can be said that we can make any crops that are cherries or are in a row with corresponding gaps. Theoretically, we could also use our hoe on corn if it is sown in 45- to 50-centimeter rows. On the contrary, it is not suitable for cereals or onions. “
“We have over 200,000 annotated photos of beets and we use them to learn when we train a model so that it can find beets in the field. We have taken pictures in two seasons, but it is important that they are as diverse as possible. The model must be able to distinguish a crop from weeds, but at the same time it must be small enough, yet powerful and fast. The hardest part was putting it all together so that it worked very quickly. If we drive at a speed of 2 km / h, which is about half a meter per second, then there are three to four beets at that half a meter. So the machine must recognize three or four crops in a second that the knives must not cut. And we’re able to drive almost twice as fast, so you don’t even have time to watch the knives open and close. “
The hardware part of the hoe runs on Nvidia’s Jetson minicomputer. Pictures of the cultivated line are taken by ordinary industrial cameras. Some integrated circuit boards were then designed by the company itself. He develops the software part of the startup himself in C ++ using the CUDA framework again from Nvidia. Each line has its own camera, computer unit and blade driver.
The concept of the device is designed so that each of the units can be easily replaced in the event of a failure. “On the one hand, there is a solution where each cultivation unit has its own hardware, overall more expensive than a centralized solution, on the other hand, we are talking about prices in the hundreds of dollars. And when one unit breaks down, it is simply replaced and moved on than to make a central solution that is cheaper to manufacture but more complicated to set up and service. This makes it more resilient and easier for farmers to understand. ”
In some situations, the hoe is better than a human
“We have the first season behind us, when we tested it from start to finish. The smaller the plant, the much higher the success rate. When we tested with the ideal plant size, we drove for half an hour at a time and the result was perfect, we did not destroy a single plant. But it was a field without weeds, where we only distinguished crops, so it went great. Once there is a weed, it depends on how much the plant can be seen. When it is totally overgrown with weeds, it is of course very difficult to recognize, but even so, I would say that the machine achieves such qualities that in some situations it is better than a human and it handles it extremely quickly. There are situations where it doesn’t work 100% yet and we have to improve it, but we’re talking about 99 percent success, for example, ”explains Ullmann, how reliable his shoulder is when working in the field. Every accidentally cut crop is a financial loss for farmers and farmers are reluctant to make mistakes.
“But we are trying to explain to them that many chemicals will die when chemistry is applied, especially the smaller ones, because they get a high dose. But farmers do not see that. It is important when communicating with farmers whether the ajťák or the farmer talks to them. Fortunately, we have farmers in the team, so the discussion is much easier. But not all farmers are open to modern technology. “
If farmers are to think about switching to hi-tech solutions, they have to achieve higher profits thanks to technology, according to Ullmann, otherwise they are said to not believe much in modern solutions. “In addition, the European Union will gradually ban the use of chemical sprays using the sugar and whip method and subsidize organic farming more. This will gradually force farmers to use more sustainable cultivation methods. They will then look at what is available on the market and see that our solution is not a magic box that sticks into a field and then counts something somewhere in the cloud, tells the farmer ‘you should water the field’ and he says’ I know, it’s dry ‘. Modern technology must make it much easier for farmers to work. And our six-line machine can basically replace 60 people. ”
The company has now completed a prototype and is finishing various mechanical improvements after this season. And it is gradually developing the specification of the second version of the prototype, which it wants to complete in September. Next year, Ullmann wants to produce at least two six-line machines. “One for the future customer we already have, the other for further testing with a large global partner in Germany, on whose farm we will test it. We also want to present ourselves at the big agricultural fair in Hanover next year. “
“We don’t have any investors at the moment, but we’re having fun and we’re open. We would need to expand our team with more people so we can start working on other solutions. Our intention is not to be another factory and machine building, but rather to license our technology and know-how to other manufacturers. There are many manufacturers who started long ago as blacksmiths and then began to specialize in agricultural machinery. These are often family businesses that cannot grasp information technology at all. They have a lot of knowledge and experience in the field of metal processing, but not in IT, and we see an opportunity there. “