display
Farmer Michael Polster has usually found out too late whether one of his cows is doing badly.
She then ate too little for a few days, an infection could spread, she became weak.
The farmer only becomes aware of the animal when the milking robot signals that there is not enough milk.
Then antibiotics are often needed - and that means for the dairy farmer: milk blockage.
He is then not allowed to sell his product for at least ten days.
The milk flows down the drain and with it a lot of resources and money.
Michael Polster places his hopes on his farm in Claussnitz in Saxony
therefore on artificial intelligence (AI).
Algorithms should recognize patterns earlier than the milking machine - and sound the alarm.
To do this, he uses the “Cow Control” program from the Dutch company Nedap.
With Cow Control, it says there, you help the farmers "to be the best farmers in the world".
Sensors on the cow's neck measure how often it gets up, how many steps it takes, how long it stands, how much it ruminates.
The movement and eating behavior provides information about the state of health of the cow.
It takes the AI a few weeks to get to know each cow and its individual rhythms.
But then Polster is automatically informed if a cow is suspicious.
display
According to the industry association Bitkom, 80 percent of farms in this country are already using digitization in the barn, known as smart farming.
Nine percent even rely on AI.
When upholstery is in operation, a sensor determines the nutrients in the slurry tanker.
Satellite data shows where the plants need more fertilizer and where they need less fertilizer.
The pump output is automatically regulated accordingly.
The robot knows the udder of every single cow.
After milking, he also knows from the milk whether an animal is sick or not
Source: Getty Images / Edwin Remsberg
An automatic feeder recognizes by the ear tag of each calf how much milk it is still allowed to drink.
Researchers are working on drones, which are supposed to detect plant diseases earlier than the naked eye, and on indoor fields that automatically extract their nutrients from the sewage treatment plant.
With the new technologies, farmers expect more efficiency, lower costs, better used resources, more animal welfare, less environmental pollution.
Thanks to the technology.
For years, many have spoken of a “milestone” in agriculture, comparable to the tractor or the chemical pesticides that made yields soar in the 1960s.
Those who leave their harvests to chance are not competitive
display
Michael Polster, a farmer for eight years, also believes that nothing works nowadays without digitization.
Those who leave their harvests to chance these days will quickly no longer be competitive.
Employees have to be paid decently, the price pressure is enormous.
Milk is as cheap today as it was in the early 1990s.
And Germans are used to spending little money on groceries.
At the same time, the cost of machinery and agricultural chemicals is increasing by around five percent year on year.
The farmers have to become more productive, save employees, enlarge areas, use fertilizers and seeds more efficiently.
Thomas Herlitzius, however, sees this rather critically: “AI does not save us back and forth,” says the professor for agricultural systems technology at the Technical University of Dresden.
He has been working with intelligent machines and algorithms for 40 years, helping to develop them himself for harvesting or cultivating soils.
In the meantime he says: “The possibilities of these technologies are generally overestimated.” So are the intelligent systems in agriculture just hot air?
AI is "totally hyped" in agriculture, says Herlitzius.
Sure, individual processes could be automated and thus cheaper.
In some cases, AI could also help to save chemicals, which is "very promising" for crop protection.
Herbicides, for example, have so far been distributed across the entire field.
If weeds were to be precisely localized using image recognition, local spraying could be used.
The use of herbicides could be reduced by 90 percent.
display
But that doesn't change anything about the core problem, the overuse of the soil and the resulting deteriorating harvests, says Herlitzius.
Furthermore, large machines are used that damage the soil.
In a short crop rotation, fields are operated in monoculture, chemicals are used.
That is not sustainable.
More ecological methods alone would really turn things around here.
Before that happens, however, there is still a lot to be done.
Most of all, it needs data.
So far, the algorithms have not been trained enough.
“Agriculture works with nature,” says Herlitzius.
Large fluctuations are the order of the day, and weather and climate are also difficult to forecast.
In no other economic area in which AI is already used, so many uncertain variables collide, says the researcher.
However, the more uncertainties there are, the more data is needed for reliable forecasts.
So far, many processes could hardly be summarized in models.
He estimates that a sufficient data basis is still five to 15 years away.
Machines speak different languages
The main reason why things are progressing so slowly is due to the handling of the data.
They accrue when the systems are used in the field and stables, but remain with the providers, be it John Deere, Bayer or BASF.
As a result, the algorithms improved only slowly.
In addition, according to Herlitzius, the systems are not compatible with one another.
“We want to do Industry 4.0 and we can't even make the individual machines talk to each other,” he says.
The data collection from the suppliers is also criticized for another reason: The suppliers could foresee the product needs of the individual farmer and charge him a higher price.
Modern agriculture: data collection with solar cells in the field
Source: Getty Images / Kilito Chan
No data exchange, no compatibility: AI is only making slow progress in the field.
The Federal Ministry of Economics therefore presented the “Agri-Gaia” platform in January, on which the agricultural data is to be brought together independently of the large corporations.
In the future, data from all machines and companies should be able to converge here - and thus improve the training status of the algorithms.
In spite of all their enthusiasm for technology, researchers and farmers see another problem: it is possible, says farmer Michael Polster, to have “a maximum amount of milk per animal” and “maximum yields at the moment”, in the barn and in the field.
However, nothing is designed for the long term and environmental concerns.
The systems fail when he tries to use slightly different varieties such as grass or rye.
So far, the systems are designed for wheat and maize alone, but they are growing all over the world.
display
And researcher Herlitzius believes that AI is primarily something that industry wants to advance.
The necessary change to a more sustainable agriculture is not taken into account.
“This also creates demand that is actually not always there,” he says.
You know what sells well.
The German agricultural machinery manufacturers alone have an annual turnover of almost ten billion euros.
A sizeable market that will develop in the future in such a way that providers sell the entire “healthy field” package.
They then not only provide fertilizer or pesticides, but also provide all-round care for the field, from seed to harvest.
The first projects are already running.
background
Drones for potatoes
Phytophthora infestans, the causative agent of potato rot, once caused famine.
Today it is fought with large quantities of pesticides.
Because if it reaches a field, the harvest is
ruined
within
two weeks
.
In the future, artificial intelligence will help: a drone, equipped with a camera, will spot the first plant from the air as soon as it is infested.
“The
drone can
cover the entire field with every single plant within half an hour,” explains Elmar Berghöfer from the German Research Center for Artificial Intelligence in Oldenburg.
Drones and spectral analyzes have been used for many years to identify infected plants.
So far,
multispectral cameras have been used
.
This allowed you to get an overall picture of the field from a great height.
With the help of new
hyperspectral cameras
, however, much more can be seen.
"If the plant is under stress, it changes its heat radiation," explains Berghöfer.
“You can measure that before the damage is clearly visible on the leaves.” Once you have spotted the first infected plants, the pathogens can be targeted.
Initial experiments on test fields show that this system can save more than
20 percent pesticides
.
Indoor farming
The important fertilizer phosphorus could be used up in 50 to 100 years.
That is why the fabric is recycled from wastewater.
Ansgar Bernardi from the German Research Center for Artificial Intelligence calls it "Turn shit into gold".
Phosphorus is mainly dissolved in wastewater, which is why
heads of lettuce
should now
thrive
in
hydroponics
in basins that are filled with nutrient solution from the sewage treatment plant.
The plants can grow on top of each other on several levels and need little space for their roots, which are located directly in the nutrient-rich water.
Artificial intelligence should now learn when there is a particularly high amount of phosphorus in the wastewater, for example during the halftime breaks of
football games
.
Knowing patterns is particularly important in hydroponics because they are sensitive to fluctuations.
The first indoor field connected to a sewage treatment plant went into operation in California at the end of 2018.
Here, fruit and vegetables are raised by robots.
They get the information from a
cloud
, sensors and cameras monitor growth.
According to the manufacturer, the system achieves 30 times more yield than comparable conventional farms.
A major problem, however, is the price.
This is where you will find third-party content
In order to interact with or display content from third-party providers, we need your consent.
Activate external content
I consent to content from third parties being displayed to me.
This allows personal data to be transmitted to third-party providers.
This may require the storage of cookies on your device.
More information can be found here.