Lectures, relevant list of papers / book chapters.
FEL3751 Scientific Machine Learning for Modeling and Control of Dynamical Systems 7.5 credits

The number of complex and large systems is increasing with the challenges we are facing today. Traditional control tools often have limited scaling capabilities and require manual tuning. There is a growing need for updated approaches that can automatically manage the design and control of such systems. Scientific Machine Learning (SciML) is an emerging and promising approach that bridges the gap between data-driven models and traditional physical principles. By integrating machine learning algorithms with established physical laws and mathematical models from science and engineering, SciML enables the creation of models that are accurate and consistent with a dynamical system and fundamental principles such as energy and mass conservation.
Unlike conventional machine learning models, SciML leverages additional knowledge of physical systems, which enhances interpretability, stability, and robustness. Recently, it has begun to transform control theory, offering new methods for system identification, observation, and control. This course will provide an introduction to some SciML techniques and its application to the modeling and control of dynamical systems. It will cover their fundamentals, explore their advantages over traditional techniques, and focus on practical use cases.
Information per course offering
Choose semester and course offering to see current information and more about the course, such as course syllabus, study period, and application information.
Course syllabus as PDF
Please note: all information from the Course syllabus is available on this page in an accessible format.
Course syllabus FEL3751 (Spring 2025–)Content and learning outcomes
Course disposition
Course contents
Part I: Modelling and analysis of dynamical systems
Lecture 1: Course outline. Introduction to SciML.
Lecture 2: Tutorial on using Tensorflow for SciML in Python.
Lecture 3: Learning system's dynamics with SciML.
Lecture 4: Stability analysis.
Part II: Observation and control of dynamical systems.
Lecture 5: Learning-based observer
Lecture 6: Learned Model Predictive Control
Lecture 7: Neural-network controllers
Part III: Advanced topics
Lecture 8: Operators and their applications to dynamical systems.
Lecture 9: Gaussian processes and conclusion.
Lecture 10: Project presentations
(Lecture 11: Project presentations)
Intended learning outcomes
After the course, the student should be able to:
- Apply SciML techniques to classical dynamical systems
- Understand SciML algorithms and adapt them to special cases
- Select, design and develop a SciML model for a practical control problem
- Critique the use of a SciML method in a given application
Literature and preparations
Specific prerequisites
Doctoral students at the School of Electrical Engineering and computer Science. External participation by admission of the examiner.
Literature
Examination and completion
Grading scale
Examination
- TEN1 - Written exam, 7.5 credits, grading scale: P, F
Based on recommendation from KTH’s coordinator for disabilities, the examiner will decide how to adapt an examination for students with documented disability.
The examiner may apply another examination format when re-examining individual students.
If the course is discontinued, students may request to be examined during the following two academic years.
Other requirements for final grade
Passing Grade based on weekly homework projects and final presentation. The student will work in group with independent projects. After each lecture, the student will receive a coding assignment which will make use of the concepts explored during the lecture. Each student will submit its assignment and will review the code and the report of the 2 others in its group and give written feedbacks and assessments. There will be 5 homework.
The final project presentation will either be a presentation of the weekly results with a special addition on some uncovered materials during the course or a presentation of an article which is related to the student research topic and scientific machine learning.
Examiner
Ethical approach
- All members of a group are responsible for the group's work.
- In any assessment, every student shall honestly disclose any help received and sources used.
- In an oral assessment, every student shall be able to present and answer questions about the entire assignment and solution.
Further information
Course room in Canvas
Offered by
Education cycle
Supplementary information
Readings and discussion of relevant seminal research articles in connection with each session.