Top challenges of robotics in precision agriculture: Digitalization & Automation


Today, after 20 years of research in precision agriculture, there are many types of sensors for recording agronomically relevant parameters and many farm management systems. Electronically controlled machines are state-of-the-art. Indeed, by networking different machines, technology can now automate cyber-physical systems. We call this “4.0 agriculture.”

However, it still can not be claimed that agricultural precision has been widely established in crop production. Why isn’t it? Data alone isn’t enough. Automatic data recording only helps farm results where it takes less time to analyze the collected material and makes it possible to make more profit compared to the right, intestinal-based management decisions.

Machinery today, not agricultural products, is the most significant portion of the added value deriving from new technology. For example, futures trading on the commodity market has a much faster and more direct influence on agricultural product value development than it does, e.g., by applying site-specific management techniques, product quality or yield is increased in single-figure percentages. Time-saving benefits remain. The task of developing agricultural robotic engineering is creating so-called “smart” systems that are smart and easy to operate. We call smart products that seem brighter than the user to give answers even before the question is asked.

Example: fitness bracelets recording and analyzing the wearer’s movements. The smartness of the equipment is the value analysis. Step count and heart rate below average. It results in a treatment recommendation: exercise more! But the user still has to implement these recommendations — the second example, from agricultural precision this time.

The most successful crop plant sensor systems are easily those that analyze, recommend, and then apply a one-go treatment like the so-called N-Sensors. Despite the highly complex analytical procedure of the system, operating such sensors is very easy. For example, yield mapping is an off-line approach requiring additional processing steps to analyze PC data. The yield information gained from one harvest can only be used in the next growing season, representing a long-term investment with a lot of manual input and hard-to-evaluate benefits.

The development of sensors and agricultural robotic technology is challenged by the required high temporal and spatial resolution data, very different and challenging to measure parameters under most unfavorable conditions. New analytical methods aim to combine data and fuse different layers of information to generate new knowledge. Besides, automation of data collection tasks is a requirement in the development of “smart sensor” systems for agricultural applications in the sense that decision-making is embedded in the sensor so that the results are directly applicable to the robot to perform specific management actions.

But what exactly does the above-mentioned fitness bracelet do in terms of content with the N-Sensor? Both sensors analyze data and process the material so that the result can be used to deduce direct action recommendations. Both analytics are also based on indicators not directly related to target values. The plant’s “fitness” can be evaluated adequately by foliage chlorophyll or green color. But the cause of the problem is not an inadequate supply of nitrogen, but a lack of moisture, the system must have this additional information.

In this regard, intuitive man-robot interaction is needed, which is also a significant development challenge for sensors and crop automation technology. The more understanding of detailed agronomic relationships is appreciated, the more information is needed to understand these relationships better. The more information available, the more in-depth understanding, requires more data collection.

Therefore, the situation is a loop where more and more data were collected, and increasingly intensified agronomic knowledge was developed, especially in recent years. It stagnated the practical application of directly usable agronomic knowledge. Fully applying agricultural precision technology still requires a considerable mass of statistics and software expertise. Relevant information must be integrated into multi-causal decision-making systems to develop smart sensors further.

The goals are complex systems with easy to use systemic, comprehensive, and transparent concepts, good “usability” and simple application. Practical experience must also flow into these integrated systems so that farmers can further develop their expertise using technology. The step from data storage to information retrieval to knowledge management involving large amounts of data is a crucial theme for Ag-robotics decision support systems.

Currently, opportunities to analyze agricultural data and sensor data fusion are expanding by applying multivariate statistical methods and machine learning techniques. It is increasingly expanding the system limits, and holistic concepts for complete value-added networks are already in focus nowadays, whereby mobile data transmission is an essential technology for establishing fully integrated systems that enable real-time data fusion from different sources. We talked about knowledge management and smart systems. But with all this “high-tech,” can we clearly focus on our goal?

More efficient crop production is the crux of all technical developments. Automation and networking should serve systemic control of agronomic processes, not vice-versa. It is the environment where, of course, the Leibniz Research Alliances ‘ innovation initiative “Food and Agriculture 4.0” “Sustainable Food Production and Healthy Nutrition” focuses on the process of agricultural production-smartly connected. The initiative aims to develop underlying process technology for Agriculture 4.0 across disciplines where knowledge-based decision-making ensures the satisfaction of social demands and the demands of individual producers and consumers in terms of yields and profits while taking into account local, spatial, environmental heterogeneities and global climate phenomena.

For this purpose, the research goal is, on the one hand, to develop models of agricultural production processes adapted to meet specific conditions and, on the other hand, automation technologies with which methods are to be controlled in such a way that natural resources can be retained or even improved while maintaining product quality. Interoperability of agriculture and digital networking will enable new process control systems and sales models, such as online slurry sales points, exchange platforms where data is traded for advice or direct online marketing.

However, only what is sown can be driven home, even with Agriculture 4.0. For example, the weather risk is no less, although the harvest window may be better positioned by setting the IT to be used. It concludes with the summary that even Agriculture 4.0 will only show modest results if the new technologies do not take care that some of the added value is associated with agricultural products.