Classical engineering is characterized by plannability. As early as in the design phase, developers know how the individual components and the complete system will behave later on. Systems containing artificial intelligence (AI) or machine learning (ML) components, however, are not that plannable. As they are data-driven, they develop further during their runtime and will reach their final functionality during operation only. This makes the reliable management of exceptional situations very difficult and also economic profit can hardly be determined in advance. For this reason, use of smart systems by companies is prevented by the fact that they are not as calculable as classical engineering systems.

CC-KING now pools IT and engineering competencies in Karlsruhe to facilitate AI use in practice. In close cooperation with industry, the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB, FZI Research Center for Information Technologies, and KIT study fundamental aspects, methods suited for practice, and concrete application problems.

Fundamental Methodological Aspects …

The (non-) predictability of the behavior of learning systems is a central topic of AI engineering. "AI engineering is aimed at making AI and ML useable by engineers, comparable to classical engineering. It is still a very young discipline that bridges the gap between AI fundamental research and engineering sciences," says Professor Jürgen Beyerer, Scientific Director of the Competence Center, Managing Director of Fraunhofer IOSB, and Professor at KIT. "Research will focus on predictability as well as on the safety of AI-based systems, explanation of their decisions, and incorporation of prior and expert knowledge in data-driven approaches." The goal is to develop a standard procedure for AI engineering to ensure usability of AI technologies by large and heterogenous teams.

"As a technology hub with a long tradition in engineering sciences and IT, Karlsruhe offers optimum prerequisites for the Competence Center," Beyerer emphasizes. The Baden-Württemberg Test Area for Autonomous Driving and the Research Factory that is currently being built in Karlsruhe will be used as real-world labs for applications in the mobility and industrial production sectors. "Under these conditions, AI engineering may become the unique selling point of German AI."

… as well asl Advice and Learning Lab for Small and Medium-sized Enterprises

CC-KING will seek to make the use of AI components manageable by small and medium-sized enterprises. “Even highly innovative medium-sized companies are often lacking AI competence. This gap is difficult to close, because AI experts are rare and often not familiar with typical applications,” Beyerer says. CC-KING offers concrete support to companies that may unbureaucratically order so-called QuickChecks or TransferChecks. A consultation office and an AI Engineering Learning Lab for training staff are currently being established.

If interested, companies can contact the CC-KING coordination office of Fraunhofer IOSB (by mail to: kompetenzzentrum@ki-engineering.eu or by phone to the project assistance 0721/6091-290). In August 2020, the Baden-Württemberg Ministry for Economic Affairs, Labor, and Housing approved funding of CC-KING to the tune of EUR 3 million.

Contributions of the Partners

Fraunhofer IOSB as coordinating consortium partner contributes its wide IT competence in industrial automation and control technology as well as in the areas of AI and ML. “In the past three years, we have already developed a tool-based approach to AI engineering in industrial production within our project “M4P – Machine Learning for Production,” says Dr. Julius Pfrommer, Head of the research group at IOSB and Scientific-technical Director of CC-KING. “This tool helps us apply AI methods in a plannable and reproducible way. The AI algorithms are of crucial importance, but often represent a fraction of the solution only.” Work rather focuses on deep integration of tools existing in engineering in AI methods. ”In this way, AI can also work well in areas where no or only few data are available.”