Hilde Gerold receives the 2026 Hans Uhde Prize for innovative research into AI-supported process control

In her master's thesis "Enhancing Control in Chemical Processes using Reinforcement Learning with Human Feedback", which was written at the Chair of Process Automation Systems, she investigates the integration of human process knowledge into data-based control strategies. The focus is on the Reinforcement Learning from Human Feedback (RLHF) approach, in which the preferences of process experts are used to develop safe and efficient controllers for complex chemical processes.
The developed methodology was applied to a bio-batch reactor for penicillin production and a polymerization reactor, among others. The results show that safety-critical factors and long-term effects can be successfully taken into account through human feedback. In addition, the work combines analytical optimization approaches with neural networks and shows new possibilities for combining different optimization goals.




