Data-efficient AI

© AdobeStock

The basic idea behind machine learning is that an artificial system learns from patterns and relations in data. Thus, machine learning cannot be utilized without an adequate data basis. In this context, the term “adequate data basis” refers to the quantity, quality, relevance, and diversity of data.

However, the high volume of data in production is often a challenge, as the data must be generated in the production environment and annotated extensively. A further problem is that the data rarely covers borderline cases. This can result in the trained model making incorrect decisions when used. Learning on the real system, which would be necessary with reinforcement learning approaches (i.e., the principle of learning by trial and error), is generally not feasible due to the high cost, time, and maintenance involved.

Therefore, the key question is: How can a machine learning model still be trained effectively if only a small amount of real-world data is available? This is where data-efficient AI comes into play. Possible solutions are highly diverse and range from the use of simulation environments and digital twins to data-saving learning methods and approaches for integrating existing knowledge. The work on data-efficient AI primarily concentrates on so-called physics-informed machine learning and data-efficient reinforcement learning.


Physics-Informed Machine Learning

In machine learning, patterns are learned using large data sets. Often, however, prior knowledge about a specific problem already exists. Physics-informed machine learning combines prior knowledge in the form of physical laws and equations with machine learning. The result are more accurate and robust predictions, often with much less data being required.


Data-efficient Reinforcement Learning

Data-efficient reinforcement learning describes a research field that attempts to make data-hungry algorithms more data-efficient through the use of expert knowledge, physical laws, abstraction, or a digital twin.


Future Work Lab

Digitization and Industry 4.0 are significantly changing work processes in industry. The use of AI is also becoming increasingly popular. How can the potential of AI be harnessed for the manufacturing industry and what do possible use cases look like? Visit our Future Work Lab, where you can experience the industrial work of the future live.



The demands on the accuracy of the robots used in production is often very high. In the KIRK project, new AI-based calibration methods are being developed to significantly improve the accuracy of industrial robots. By using physics-informed machine learning, the developed calibration methods are not only accurate, but also robust and to some extent explainable.



Until now, programming robots for assembly tasks has been costly and time-consuming. Therefore, the “Rob-aKademI” project is exploring ways to make programming easier and more autonomous by using simulations and machine learning, or more precisely data-efficient reinforcement learning.