Advanced MPC
Description
Model predictive control (MPC) has established itself as a powerful control technique for complex systems under state and input constraints. This course discusses the theory and application of recent advanced MPC concepts, focusing on system uncertainties and safety, as well as data-driven formulations and learning-based control.
Tentative content:
- Review of Bayesian statistics, stochastic systems and Stochastic Optimal Control
- Nominal MPC for uncertain systems (nominal robustness)
- Robust MPC
- Stochastic MPC
- Set-membership Identification and robust data-driven MPC
- Bayesian regression and stochastic data-driven MPC
- MPC as safety filter for reinforcement learning
Requirements
Basic courses in control, advanced course in optimal control, basic MPC course (e.g. 151-0660-00L Model Predictive Control) strongly recommended.
Background in linear algebra and stochastic systems recommended.
Literature
Lecture notes will be provided.
Exam
Final oral exam during the examination session, covers all material.