- Introduction to dynamical systems and their representation
- Linear time-invariant systems: solutions, computational aspects and structural properties
- State-feedback controllers for linear systems
- Linear observer (Luenberger) and dynamical feedback compensator
- Optimization-based regulator: model predictive control
- Methods for data-driven system identification
- Training of recurrent neural networks for system identification
- Neural networks-based control design
- Astrom, Murray, "Feedback Systems: An Introduction for Scientists and
Engineers", Princeton U.P. 2008 (available online)
- Lygeros, Ramponi, "Lecture Notes on Linear SystemTheory", ETH Zurich 2013 (available online)
- Soderstrom, Stoica, "System Identification", Prentice Hall 1988
- Bishop, "Pattern Recognition and Machine Learning", Springer Science 2006
- Pham, Liu, "Neural Networks for Identification, Prediction and Control",
Springer-Verlag 1995
Obiettivi Formativi
-Familiarize with state-space representation as a modelling
formalism of linear differential/difference equations
- Analyze structural linear system properties: stability,
controllability and observability
- Design state observers: Luenberger, Kalman filters
- Design state-feedback controllers: pole placement, LQR, MPC
- Apply machine learning-based approaches to solve standard problems in system/control theory: critical analysis of pros and cons
Prerequisiti
Linear algebra, mathematical analysis and calculus
Metodi Didattici
Slides+blackboard, paper reading sessions. Integrate theoretical discussion with hands-on sessions
Altre Informazioni
From March 6, 2024, to May 10, 2024:
- Wednesdays, 8.45-11.10, Learning Center "Morgagni", aula 110
-Fridays, 13.30-15.00, Learning Center "Morgagni", aula 110
Modalità di verifica apprendimento
- Oral exam, or
- Project discussion
Programma del corso
- Introduction to dynamical systems and their representation, formalism of differential/difference inclusions
- Linear time-invariant systems: solutions and computational aspects
- Linear time-invariant systems, structural properties: stability, controllability, observability
- State-feedback controllers for linear systems: pole-placement, linear-quadratic regulator
- Linear (Luenberger) and stochastic (Kalman) observer for dynamical feedback compensator
- Intermezzo: convex optimization
- Optimization-based regulator: (linear) model predictive control
- Methods for data-driven system identification
- Training of recurrent neural networks for system identification
- Neural networks-based control design