Research Projects

Optimal feedback control is one of the areas where deep learning has an enormous impact. Deep Reinforcement Learning, one of the methods for obtaining optimal feedback laws and arguably one of the most successful algorithms in artificial intelligence, stands behind the spectacular performance of artificial intelligence in games such as Chess or Go, but has also manifold applications in science, technology, and economy. This project explores the mathematical foundation of this success.

Joint graduate college of the Universities of Bayreuth and Jena, combining polymer chemistry with computer science and robotics. The aim is to design copolymers in a targeted manner and analyze them during processing.

Based on first principles simulations of the barriers hindering charge transport, we employ machine learning to engineer the grain boundaries of organic semiconductor thin films to optimise them for use in organic field effect transistors. Our models point out novel molecules which are then synthesized and tested by our experimental collaborators, providing feedback for the ML models. This novel integrated approach will unveil more efficient organic transistor materials by minimising transport losses.

This project aims at discovering novel lead-free perovskite analogue materials for photovoltaic applications. To this end, we employ first principles calculations of the efficiency and stability of the materials, which we then use to train explainable ML models to not only extend the reach of the (otherwise quite expensive) calculations, but also to identify more general design principles leading to the desired materials' properties. In the framework of the IRTG OPTEXC our predicted materials are synthesized and characterised by an automated high-throughput lab at Monash University in Melbourne Australia, not only testing our predictions but also providing further data for our ML models.

The project is funded by the Deutsche Forschungsgemeinschaft and is part of the IRTG OPTEXC.

The Data Protection Sandbox (german: "Datenschutz-Sandbox") project is developing a regulatory testing environment where companies and public authorities can test new digital applications under real-world conditions in compliance with data protection regulations. The goal is to identify data protection issues early on and allow legally compliant innovation. Supported by the University of Bayreuth and the Rheinland-Pfalz Commissioner for Data Protection and Freedom of Information (LfDI Rheinland-Pfalz), the project analyzes legal and technical frameworks, establishes a sandbox, and develops a guideline for data protection authorities.

Find more information here!

Funding statement: Bundesministerium für Forschung, Technologie und Raumfahrt

The Data4Collar project focuses on two processes that are intertwined in the industrial manufacturing of many functional components: shearing and collar drawing. State-of-the-art data AI and analysis technologies and process analytics are efficiently combining aiming to early detect errors in collar forming.

Find more information here!

Funding statement: DFG

As part of the Smart-VHB project, new learning modules on the topic of "Generative AI" are being developed. Once completed, these modules will be accessible via the Virtual University of Bavaria (vhb).

Find more information here!

Funding statement: vhb, virtuelle digitale Hochschule

The proposed research unit – SOURCED – Process Mining on Distributed Event Sources will develop the methodological foundations for novel process mining techniques for distributed event data. Its contributions will leverage online analysis of data from distributed event sources under consideration of infrastructure, data and user concerns.

Find more information here!

Funding statement: DFG

The purpose of this project is to develop data-driven process analysis for context-sensitive management of functional areas to reduce emissions and promote animal welfare in naturally ventilated pig barns with outdoor access. For this purpose, sensors, video data, and AI algorithms form a monitoring system that enables real-time analysis of pig behavior. Deviations from expected behavioral patterns are detected and visualized as key indicators. This provides farmers with information about potential problems and allows them to adjust the animals' living conditions, particularly with regard to the barn's climate. By using AI-supported algorithms for behavioral recognition, animal welfare can be significantly improved and environmental protection considerably promoted.

Find more information here!

Funding statement: Europäische Innovationspartnerschaften (EIP)

The objective of the NFDIxCS consortium is to identify, define, and deploy services for storing complex, domain-specific data objects from across the field of computer science, thereby implementing the FAIR principles comprehensively. This includes the production of reusable data objects that contain not only various types of computer science data but also the associated metadata, as well as the corresponding software, context, and execution information in a standardized format.

Find more information here!

Funding statement: DFG

The aim is to investigate how AI systems can be made explainable in order to enable responsible decision-making processes, trust, and traceability—taking into account legal, philosophical, and psychological perspectives.

Investigates how AI can be used legally and ethically in law enforcement—e.g., for deanonymization in cryptocurrency systems, with particular consideration given to data protection and AI regulations at the EU and federal levels.

Generative AI enables the creation and manipulation of highly realistic synthetic media, known as deepfakes. These technologies are increasingly exploited for fraud and disinformation, posing a dilemma for law enforcement: while authorities face new challenges in evidence assessment and establishing probable cause, deepfakes also offer novel investigative opportunities, such as using digital clones to infiltrate criminal networks.