Cognitive Internet Technologies for "Smart Agriculture" and "Precision Farming"
By 2050, almost 10 billion people will be living on Earth. However, there is only limited space available for growing food, and climate change is also increasing the pressure. Consequently, only sustainable and innovative agriculture can feed mankind. It has to produce more food and at same time protect the environment. Digital concepts such as "smart agriculture" and "precision farming" play a central role in this. Therefore Fraunhofer is developing cognitive internet technologies.
In an era of a growing global population, increasing environmental pollution and changing climatic conditions, concepts in digital technology are the key to more sustainable and efficient agriculture. They make an essential contribution to ensuring that the world is fed. At the same time, industry, politics and society have recognized the digitization of agriculture and cultivation research as a major challenge and are seeking appropriate solutions.
Optimizing resource use and yields
Thus, together with users, Fraunhofer is pushing forward applied research projects for the digitization of agriculture. In close cooperation with the Fraunhofer lighthouse project Cognitive Agriculture (COGNAC), the Fraunhofer CCIT is working on an intelligent IT and electronics system that is intended to support agricultural enterprises in optimizing the cultivation of their plants. In so doing, the Fraunhofer CCIT approach combines intelligent sensor technology, machine learning and technologies for protected data spaces. Different types of sensor technologies will independently monitor the growth and condition of the plants and, on the basis of the data, identify the optimum use of water, fertilizers and pesticides as well as the optimum harvesting time. This leads to greater sustainability, optimum use of resources and more reliable yields. The system will be autonomous and react adaptively to variable and dynamic constraints. The sensor technology has already been developed. In the next step, scientists will apply machine learning algorithms to the data obtained.
This applied research by the Fraunhofer CCIT will benefit not only seed producers and farmers. After successful use in food production, the system will be applied to the entire value chain – from cultivation to transportation and marketing to consumption by the end customer. Automated, intelligent condition monitoring, combined with data exchange in secure infrastructures, results in better quality and more efficient handling of the products.
Machine learning with small data sets
Machine learning allows agricultural production requirements to be recognized much earlier than is possible with conventional methods. Sensors record the condition of plants and soils. The data from sensors distributed over great distances can be aggregated. Through patterns identified in this way, it is possible to predict, for example, where pests or diseases will spread. This is an important basis for an economical and preventive use of crop protection agents. However, the training of intelligent analysis systems, such as neuronal networks, requires large amounts of sample data. These are not always available in agriculture. The technological approach of the Fraunhofer CCIT therefore also leverages farmers’ expertise and experience. Hybrid machine learning based simultaneously on expertise and recorded data thus makes reliable analyses possible even with small amounts of data.