Smart farming

In the service of more sustainable management

  • Forschung

Agriculture ought to become more efficient, environmentally friendly and sustainable. In order to come closer to this goal, various groups of the Institute for Agricultural Sciences (IAS) are driving the development of modern technologies. At ETH, weeding robots, independent cows, hovering plant breeders and emission-measuring barns are already a reality.

Agriculture today faces many challenges. The population, as well as experts, are demanding more animal welfare and more environmentally friendly production. At the same time, management must be adapted to the changing climate.

This puts pressure on both researchers and farmers to find alternative production methods. “Modern technologies such as image recognition, artificial intelligence and automation can contribute to this,” says Achim Walter, Head of the Crop Science research group and smart farming expert. And he continues: “The development of new methods and machines has always been a strength of ETH. In the field of smart farming, this strength can be put to excellent use in the service of more sustainable management of our fields. Together with farmers, authorities and companies, ETH researchers are trying to find out which technologies can be used in practice in the near future.”

Machines learn to evaluate images

The term “smart farming” refers to the use of computer-assisted technologies in agriculture. Smart farming can be practised in very many areas, which is also reflected in the diversity of research projects at ETH. A pioneering role is played by Achim Walter’s Crop Science group. It is involved in several projects at once. The DeepField project is developing a way to use satellite imagery to make very precise fertiliser recommendations for nitrogen in the future. Two ongoing projects aim to make grain breeding more efficient: For the Global Wheat Head Detection Challenge, algorithms are being created to determine the number of wheat heads per area in photos – an important indicator for breeding. The second project, TraitSpotting, uses the unique Field Phenotyping Platform in Lindau-Eschikon and drones to select the best varieties for further breeding.

The FIP in action: the sensor head can position itself directly over the individual plots to document the characteristics of each wheat variety.
(Photo: Tom Kawara, ETH Zurich)
Frederic Kislinger gets Matrice 600 ready to start at the FIP site.
(Photo: Andreas Hund, ETH Zurich)
Matrice 600 during a low-level flight over the wheat trial in the FIP.
(Photo: Norbert Kirchgessner, ETH Zurich)

Reducing pesticides with robots

In addition to the development of image recognition software, ETH is also investing in various other research projects aimed at optimising crop production, for example in the development of new sensors or robots. The latter both in the form of drones and autonomous robots that can navigate a field and be used for sowing, pest control or harvesting. There is a lot of potential here for reducing the use of pesticides or fertilisers. The robots are intended to reliably detect nutrient deficiencies or bioaggressors and combat them in a very targeted manner. This means they only apply fertiliser or pesticides where absolutely necessary. Or, in the case of weeds, they dispense with herbicides altogether and carry out control mechanically. The aim is to not only reduce the amount of additives in the environment, but also lighten the burden on farmers.

This is roughly what the drones of the Environmental Robotics group will look like, which will be able to explore tree canopies at close range in the future.
(Photo: Research group for environmental robotics)
Logo of the recently founded Environmental Robotics group.
(Photo: Research group for environmental robotics, ETH Zurich)

Tailored to the specifics of Switzerland

Many of the existing technologies are geared towards large farms, which is an obstacle to their use in Switzerland with its small-scale structures. Often an investment is not worthwhile from a purely economic point of view. This is where the InnoFarm project comes in, involving three research groups and the farmers’ association of the canton of Solothurn. Research is being conducted into how drones and the latest sensor technology can be used to apply nitrogen and pesticides in a more targeted manner and thus reduce them. Based on the resulting data, the Agricultural Economics group is reviewing whether and how these technologies can actually be used in practice, for example whether each farmer procures a drone for himself, several farms share one or whether it is a job to be done by contractors. The project also investigates the role that agricultural policy can play.

The hand-sized prototype of the Digit Soil sensor. A new membrane (in blue) is used for each measurement.
(Photo: Sonia Meller, Digit Soil)
A part of the team: agronomist Hélène Iven (left) and Digit Soil founder Sonia Meller (right).
(Bild: Jasmin Fetzer)

An important element in the study programme

The topic of smart farming is also becoming increasingly important in teaching. In recent years, new lectures such as “Crop phenotyping” or “Innovation in precision agriculture” have been launched. Here, students are introduced to the world of digitised agriculture, learn about the latest technologies and their limitations, apply the technologies themselves, develop their own ideas, have the opportunity to found start-ups and benefit from the exchange with students from other departments.

The development team presents its functional weeding robot, Rowesys.
(Photo: Immanuel Denker, Rowesys)
The Rowesys robot removes the weeds between the rows with a four-pronged grubber.
(Photo: Immanuel Denker, Rowesys)

Applied smart farming in farm animal sciences

An example of applied smart farming is the dairy barn at AgroVet-Strickhof in Eschikon. It is equipped with modern sensor technology that allows a high degree of animal welfare. The cows are largely independent. They operate milking robots and automatic feeders themselves based on their own needs. In addition, the activity of each animal is recorded. The regularity of chewing movements, number of milkings per animal and day and visits to the automatic feeder are recorded. This makes it possible, for example, to detect possible diseases at a very early stage.