Risk-aware and Robust Approaches for Machine Learning-supported Model Predictive Control for Iterative Processes - Doctoral Thesis

Aug 3, 2026

Upload by Eliana

This doctoral thesis investigates risk-aware and robust control strategies for model predictive control (MPC) in iterative processes using machine learning. It addresses the challenge that learned models may become inaccurate when connections to underlying physical laws are only partially preserved. The work proposes two data-driven uncertainty modeling methods: one combining Gaussian processes for uncertainty and neural networks for the nominal model, and another using tube-based MPC with a “safe set” concept to guarantee constraint satisfaction under stated assumptions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages178

Reading Time7 h 29 min

AccessFree to download

Acknowledgements

Zusammenfassung

Motivation

Proposed methods

| Risk-aware and Robust Approaches for Machine Learning-supported Model Predictive Control for Iterative Processes | | --- | | Zur Erlangung des akademischen Grades Doktor-Ingenieur (Dr.-Ing.) Genehmigte Dissertation von Bruno Morabito aus Reggio Calabria Tag der Einreichung: 27.07.2023, Tag der Prüfung: 20.12.2023
1. Gutachten: Prof. Dr.-Ing. Rolf Findeisen
2. Gutachten: Prof. Dr.-Ing. Sergio Lucia
Darmstadt, Technische Universität Darmstadt, 2023 | |
Electrical Engineering and Information Technology Department
Institute of Automatic Control and Mechatronics
Control and Cyber-Physical Systems | Risk-aware and Robust Approaches for Machine Learning-supported Model Predictive Control for Iterative Processes Accepted doctoral thesis by Bruno Morabito Date of submission: 27.07.2023 Date of thesis defense: 20.12.2023 Darmstadt, Technische Universität Darmstadt, 2023 Bitte zitieren Sie dieses Dokument als: URN: urn:nbn:de:tuda-tuprints-264993 URL: [http://tuprints.ulb.tu-darmstadt.de/26499](http://tuprints.ulb.tu-darmstadt.de/26499) [Jahr der Ver](Jahr der Ver)ö[ffentlichung auf TUprints: 2024](ffentlichung auf TUprints: 2024) Dieses Dokument wird bereitgestellt von tuprints, E-Publishing-Service der TU Darmstadt [http://tuprints.ulb.tu-darmstadt.de](http://tuprints.ulb.tu-darmstadt.de)[ ](http://tuprints.ulb.tu-darmstadt.de)[tuprints@ulb.tu-darmstadt.de](tuprints@ulb.tu-darmstadt.de) Die Veröffentlichung steht unter folgender Creative Commons Lizenz: Namensnennung – Weitergabe unter gleichen Bedingungen 4.0 International [https://creativecommons.org/licenses/by-sa/4.0/](https://creativecommons.org/licenses/by-sa/4.0/) This work is licensed under a Creative Commons License: Attribution–ShareAlike 4.0 International [https://creativecommons.org/licenses/by-sa/4.0/](https://creativecommons.org/licenses/by-sa/4.0/) To my aunt Brunella Erklärungen laut Promotionsordnung § 8 Abs. 1 lit. c PromO Ich versichere hiermit, dass die elektronische Version meiner Dissertation mit der schriftlichen Version übereinstimmt. § 8 Abs. 1 lit. d PromO Ich versichere hiermit, dass zu einem vorherigen Zeitpunkt noch keine Promotion versucht wurde. In diesem Fall sind nähere Angaben über Zeitpunkt, Hochschule, Dissertationsthema und Ergebnis dieses Versuchs mitzuteilen. § 9 Abs. 1 PromO Ich versichere hiermit, dass die vorliegende Dissertation selbstständig und nur unter Verwendung der angegebenen Quellen verfasst wurde. § 9 Abs. 2 PromO Die Arbeit hat bisher noch nicht zu Prüfungszwecken gedient. Darmstadt, 27 .07.2023 Bruno Morabito v Acknowledgements As I present this Ph.D. thesis, I am reminded of the numerous individuals who have contributed to this journey, each playing a vital role. First off, I would like to express my sincere appreciation to Prof. Rolf Findensen for granting me the opportunity to embark on this project. I am very gratefull to all my colleagues, especially to my closest fellow academic travelers: Johannes Pohlodek, Sebastián Espinel Ríos, Rudolph Kok, Hoang Hai Nguyen, Petar Andonov, Mohammed Soliman and Mohammed Ibrahim. We shared more than just research ideas and deadlines. From brainstorming sessions and arabic language classes, you made sure this journey was anything but dull. Our combined intellectual horsepower made for some memorable meetings and even more memorable pub gatherings. I am also grateful to the international Max Planck Research School for its supportive and dynamic academic environment. On a personal note, my deepest gratitude goes to my family, whose unwavering support and encouragement have been a constant source of strength. To my girlfriend, Izel Avcı, Iowe special thanks for her understanding and unwavering support. And to my friends, particularly the Forró community of Magdeburg, who proved that Ph.D. students can indeed dance – thank you for the much-needed breaks. In conclusion, this thesis is a culmination of not just my

FAQs

Why does the thesis focus on risk-aware control for machine learning-supported MPC?

Machine learning can simplify modeling, but it may produce highly inaccurate results when physical law connections are lost, so uncertainty must be explicitly accounted for in the control design.

What is the first proposed uncertainty modeling method?

It uses Gaussian processes to learn model uncertainty and neural networks to learn the nominal model, summarizing uncertainty into a single parameter for a risk-aware MPC decision process.

How does the second method guarantee constraint satisfaction?

It is based on tube-based MPC and the concept of a “safe set,” where a feasible solution exists, and it shows that the safe set can expand under certain assumptions across process iterations.

Aug 3, 2026

Upload by Eliana