Master Graduation Project – Physics-informed data-driven modeling of dynamical networks: A system-theoretical perspective

Project Description
Dynamical networks are spatially distributed dynamical systems and consist of a large number of
individual subsystems that interact with each other to achieve sophisticated tasks. They are widely
employed in human daily life and in emerging applications, such as robotic swarms, power grids,
transportation systems, satellite networks, and other infrastructure networks. Given the importance of
these systems, the demand for their automation has pushed the development of advanced modeling and
control strategies. Particularly, due to the advanced sensing capability in these systems, data-driven
methods have received considerable interests to aid the modeling procedure and the controller synthesis.

The focus of this thesis project is on developing a novel learning algorithm for data-driven modeling of
dynamical networks. More importantly, the algorithm will incorporate essential physical features of the
network, characterized by system-theoretical properties, such that the learned network model preserves
important physical properties. The algorithm will be validated by theoretical analysis and simulation case
studies in the field of energy networks or robotic networks.

Project Application

Experiences in the following aspects are preferred:

1. Data-driven modeling, e.g., machine learning or system identification
2. Systems and control theory, e.g., robust control

This project will be conducted in the research group of Prof. Bart De Schutter. If you are interested in
this project, please contact me via the following email: Dr. ir. Shengling Shi, s.shi-3@tudelft.nl