In Focus: The Introduction of Artificial Intelligence and Machine Learning into Our Everyday Life
Control engineers and mathematicians from Chemnitz University of Technology cooperate in a nationwide research consortium that aims to optimize mobility applications and make them sustainable
Artificial intelligence (AI) and machine learning play an increasing role in industrial processes today. "The technological possibilities for data acquisition and the extraction of useful information from it open up exciting starting points for various research efforts, the results of which will significantly shape future industry as well as social perception," says Prof. Dr. Stefan Streif, Professorship of Automatic Control and System Dynamics at Chemnitz University of Technology. That is why his professorship, together with the Professorship of Numerical Mathematics (holder: Prof. Dr. Oliver Ernst), participated in a very competitive program by the Federal Ministry of Education and Research (BMBF) in the field of Mathematics for Innovations - with success.
AI-based methods for autonomous driving
With a nationwide consortium of research institutions and industrial partners, they are now working on the "Synthesis of optimal controls and adaptive neural networks for mobility applications" (SOPRANN for short) until March 2023. The project, which is funded with 1.2 million euros by the Federal Ministry of Education and Research, focuses on data-driven control and regulation systems in mobility. In particular, their resource efficiency, sustainability and safety are to be improved. "Among other things, the area of autonomous driving will be addressed, for which AI-based methods will be used and whose configuration and use will be examined with regard to mathematically defined criteria, such as optimality," explains Ernst. The research consortium also includes the University of the Federal Armed Forces Munich, the Fraunhofer Institute for Industrial Mathematics Kaiserslautern and the industrial partners Vitesco Technologies GmbH, Limbach-Oberfrohna, Hörmann Vehicle Engineering, Chemnitz, Smart Rail Connectivity Campus, Annaberg-Buchholz, IAV GmbH, Gaimersheim, and Daimler Trucks, Wörth am Rhein.
From the evaluation of optimality criteria to the quantification of uncertainty
The Chemnitz researchers will contribute to the development of adaptive training methods and the evaluation of associated optimality criteria. The analysis of approximation-based optimal control and learning algorithms, one of the core competences of the Chair of Control Engineering and System Dynamics, will contribute significantly to this work. Approximation structures, such as neural networks, form the basis of the SOPRANN project. The main research topic of the Chair of Numerical Mathematics is the development of a concept for quantifying the uncertainty of such models and the development and evaluation of novel approaches.
Useful data for AI-based learning algorithms
The main goal of the nationwide research consortium is also the continuous acquisition and use of data by AI-based learning algorithms, which are implemented and applied by the project partners. "In order to guarantee that safety aspects are taken into account, reliable uncertainty quantification and compliance with clearly defined safety certificates are also essential," says Ernst. The practice-oriented adaptation of these certificates in combination with the application-specific design of neural networks embodies a promising approach that the researchers are pursuing. "The use of captured data and its inclusion in learning algorithms offers numerous advantages for a constantly changing society and the associated growing infrastructural requirements, and also establishes AI as an inevitable element of future technologies," adds Streif.
For further information, please contact Prof. Dr. Stefan Streif, telephone 0371 531-31899, e-mail email@example.com, and Prof. Dr. Oliver Ernst, telephone 0371 531-33742, e-mail firstname.lastname@example.org.
(Author: Mario Steinebach / Translation: Chelsea Burris)