New Lecture at the IAT – Model Predictive Control and Machine Learning

Goal and Content of the Lecture

The students will understand the basics concepts of model predictive control (MPC). Furthermore, they are familiarized with machine learning approaches that can support model predictive controllers and possibly enhance the controller performance. This entails knowledge about theoretical questions such as stability in the nominal case, as well as extensions to the case of uncertain and disturbed systems. The students are enabled to design and implement model predictive controllers based on first principle/physical or data-based/machine learning based models. This entails the setup and design of the control structure as well as the tuning and identification of suitable parameters and cost functions of the controller.

Topics

• Introduction and basics of optimal control
• Linear Quadratic Regulator (LQR) discrete-time/continuous-time
• Basics of model predictive control (MPS) (cost functions, constraints, receding horizon)
• Nominal model predictive control of linear systems
• Robust and stochastic model predictive control of linear systems
• Control of nonlinear systems with model predictive control
• Basics of machine learning
• Combination of machine learning approaches with model predictive control

The lecture is held in English.

Further details: see TUCaN and the Moodle page of the course

Click here to go to the lecture page