The conventional starting point for control system problems is a given representation of the system. Consequently, the classical formulation of control
problems is based on input/state/output representations, leading to the so-
called model-based approaches.
In this course, we introduce an alternative formulation of the problem,
namely the learning-based control. Here, the aim is to exploit data directly
collected from the system to autonomously learn a controller capable of per-
forming a specific task, with no human intervention.
Therefore, we will first introduce some classical optimal model-based control approaches, examining the underlying principles and methodologies and
introducing all the basic concepts required for a full understanding of the sub-
ject. Next, we will explore the evolution from model-based to model-free control
strategies. Hence, we will present the main learning-based control strategies in
order to provide the students with a solid grounding as well as an overall picture
of the current developing technologies. Students will also grasp the advantages
and limitations of the presented control strategies.
During the course, attendees will be involved in a series of practical sessions
in which theoretical concepts will be applied to existing dynamic systems in simulation. Through practical applications, the students will gain a comprehensive
perspective on the capabilities and challenges of implementing learning-based
control approaches.
problems is based on input/state/output representations, leading to the so-
called model-based approaches.
In this course, we introduce an alternative formulation of the problem,
namely the learning-based control. Here, the aim is to exploit data directly
collected from the system to autonomously learn a controller capable of per-
forming a specific task, with no human intervention.
Therefore, we will first introduce some classical optimal model-based control approaches, examining the underlying principles and methodologies and
introducing all the basic concepts required for a full understanding of the sub-
ject. Next, we will explore the evolution from model-based to model-free control
strategies. Hence, we will present the main learning-based control strategies in
order to provide the students with a solid grounding as well as an overall picture
of the current developing technologies. Students will also grasp the advantages
and limitations of the presented control strategies.
During the course, attendees will be involved in a series of practical sessions
in which theoretical concepts will be applied to existing dynamic systems in simulation. Through practical applications, the students will gain a comprehensive
perspective on the capabilities and challenges of implementing learning-based
control approaches.
- Teacher: ERICA SALVATO