Agriculture 4.0 for more animal welfare in the barn
Artificial intelligence for more animal welfare in the barn. The Education and Knowledge Center Boxberg (LSZ) and the University of Hohenheim are developing data sources for sustainable pig farming. Experimental data on animal behavior, biological data from routine operations, data on the housing environment, animal genetics: on the herd of sows including piglet rearing and pig fattening at the Education and Knowledge Center Boxberg (LSZ) there are immense amounts of data. A treasure that has so far hardly been used. Because the data records, recorded in Excel spreadsheets, paper forms or through specialist applications, are not networked. In the “Agriculture 4.0: Information system for pig farming” project, business IT specialists at the University of Hohenheim in Stuttgart combine this data in a digital platform – and thus open it up for data analysis and machine learning. This enables new insights that benefit animal welfare and a sustainable orientation of pig farming. The Ministry for Rural Areas and Consumer Protection (MLR) supports the project under the direction of the LSZ. With a funding amount of almost 200.000 euros for the University of Hohenheim, it represents a research focus.
Stress with the neighbors in the pens, fighting for access to resources such as water, feed and manipulable materials, health problems, high levels of harmful gases in the compartment – all of these factors encourage pigs to bite their tails. Science and practice assume that a combination of these risk factors plays a role - but there are still many knowledge gaps.
This is where intelligent Big Data analytics come in. "This allows us to analyze large amounts of data on these factors from different sources - and thus gain new information and uncover previously unknown connections," explains Prof. Dr. Stefan Kirn, Head of the Department of Business Informatics II at the University of Hohenheim.
"Animal husbandry offers challenging applications for machine learning methods, e.g. the welfare of the animals can be improved or operational management can be optimized," emphasizes business informatics specialist Martin Riekert, who heads the sub-project at the University of Hohenheim.
Diverse application possibilities in animal husbandry
One topic that the researchers have in their sights is the question of how to identify health risks in the piglets at an early stage using machine learning methods. To do this, they are currently examining around 25 variables and have been evaluating data on around 2011 pigs since 50.000 to check whether it is possible to predict early health risks.
"Another conceivable application would be to monitor animal behavior as part of animal welfare monitoring in order to identify stress at an early stage," says Riekert. The team uses video cameras with deep learning to evaluate the lying behavior of the animals.
Therein lies an important application for general practice. Agricultural animal husbandry faces future-oriented tasks. Many consumers today want to know where the animals come from, how they are kept and fed and that they are doing well. The data from the animal itself and from the housing environment, the husbandry technique and the state of health provide information on the many questions in their combination. Digitization and networking make significant contributions to greater acceptance of animal husbandry in society and a better image.
Many individual data islands at the Boxberg Education and Knowledge Center
The Hohenheim scientists want to put this into practice with the Boxberg Education and Knowledge Center (LSZ). The challenge at the LSZ: "There is a lot of data there, but it cannot be used because it is all isolated solutions. They are not networked," says Dr. Achim Klein, who headed the Knowledge Extraction work area to which the sub-project is assigned until the end of August 2019. “There is an enormous backlog in animal production. Because unlike in plant production, the data sets are hardly accessible for data analysis.”
Very different data is recorded on the sows, piglets and fattening pigs in the training and experimental stables. "We have routinely collected structured data such as sow planning data or fattening and slaughter data," reports Riekert. “In addition, there is further structured data on the housing environment such as compartment temperature, ventilation settings, water flow or feed consumption. In addition, unstructured experimental data on animal behavior, which we receive from more than 50 video cameras, among other things.”
Digital networking instead of isolated solutions
Up until now, this data has been recorded using Excel spreadsheets and specialist applications – so far, data recording has not even been digital everywhere. In the project, the scientists combine this heterogeneous data in a data platform (data warehouse).
They equip the entire barn with WLAN and install industrial computers with touch screens. They integrate existing external systems, such as ventilation and feeding systems. The goal: manual steps are no longer necessary in the paperless barn, and the data goes directly to the data platform via the new input mask. "Data entry is faster and more efficient this way," explains Tobias Zimpel, a researcher on the project. "A plausibility check is carried out on site and the employees can access the information system at any time."
Networking then makes the data available for data analysis. "Thanks to machine learning, the system can recognize the patterns and regularities in the diverse data," explains Riekert. "The aim is to derive previously unrecognized connections and to use them to develop decision-making aids and forecasting models that benefit animal welfare, research and individual farm management."
BACKGROUND: Project "Agriculture 4.0: Information system for pig farming"
The "Agriculture 4.0: Information system for pig farming" project is funded by the Baden-Württemberg Ministry for Rural Areas and Consumer Protection (MLR) as part of the state government's "Agriculture 4.0 sustainable.digital" strategy. The project is managed by the Education and Knowledge Center Boxberg (LSZ). The Department of Business Informatics II at the University of Hohenheim receives 197.648 euros for its project part, the total funding amounts to around 0,3 million euros. The project started on November 1.11.2016st, 31.12.2019 and will end on December XNUMXst, XNUMX.
Background: Research heavyweights
Scientists from the University of Hohenheim acquired 32,5 million euros in third-party funding in 2018 for research and teaching. In loose succession, the series “Heavyweights in Research” presents outstanding research projects with a financial volume of at least 350.000 euros for research using equipment or 150.000 euros for non-equipment research.
Text: Elsner (University of Hohenheim)
AI in the barn: The LSZ Boxberg and the University of Hohenheim open up data sources for better and more economical pig farming | Image source: University of Hohenheim / Sacha Dauphin
