Machine learning is at the heart of artificial intelligence. Robotics, energy technology and numerous other applications could not function without constant learning. But when it comes to the greatest possible safety during operation, conventional systems still reach their limits, for example in autonomous driving or in the interaction between humans and machines - with potentially fatal consequences.
The DFG research group "Active Learning for Dynamic Systems and Control - Data Informativity, Uncertainties and Guarantees (ALeSCo)" aims to make machine learning of complex dynamic systems and their control more data-efficient and reliable. The scientists’ aim is to develop fundamentally new approaches to active learning in which the learning process is continuously influenced: What needs to be learned when and how?
Professor Karl Worthmann, head of the Optimization-based Control Group at TU Ilmenau, is responsible for combining machine learning with mathematical precision in the ALeSCo project: Prof. Worthmann wants to derive verifiable guarantees about the control quality of complex systems from the analysis of the information content of existing data. He is convinced that innovative approaches to active learning will not only make the application of artificial intelligence in complex dynamic systems safer in the future, but will also further increase their performance: "Active learning and its theoretical foundation are crucial if we want to increase data efficiency. This is how we make the use of machine learning methods flexible and reliable, especially in safety-critical areas such as energy systems and robotics"
Prof. Karl Worthmann
Head of the Optimization-based Control Group+49 3677 69-3624
karl.worthmann@tu-ilmenau.de