In the seminar, I shared some personal perspectives on the future of control systems in the era of Digital Twins and Industrial AI. One of the main messages is on three topics:
-
Data-driven control: the control community has made impressive progress here, leveraging advanced control theory and safety perspectives (stability, robustness, performance) and digital twin applications (i.e. Sim2Real…). Great to see many relevant contributions from the Oxford Control Group as well.
-
Control AI systems: as industry adopts AI-based digital twins (i.e. for large 3D CFD/CAE systems), a challenge emerges: how to design control frameworks with such systems. This direction seems relatively underexplored in my view…
-
Control + Generative/Agentic AI: Foundation language/vision models are powerful and also have connection with control systems, not necessarily for control algorithm performance, but for supporting engineering workflow development, defining objective/reward functions, especially in applications involving subjective human preferences in a feedback loop.
Overall, I believe AI evolution creates new opportunities for the control community to contribute with methodology foundations and more control applications. Hope to see more works in these directions, and if you’re working on them, I would be interested to hear the details ;)
Many thanks to Luca Furieri for the kind invitation, and to Antonis Papachristodoulou, Kostas Margellos, Mark Cannon,… and the group for the warm welcome. I would also like to acknowledge colleagues at Siemens who helped initiate the Siemens-Oxford collaboration (Claire McGlynn, David Moss, Ilaria Carrara Cagni, Yves Lemmens, Katrien Wyckaert, Herman Van der Auweraer…).