Objectives

In filtering and control design, systems are affected by uncertainty so that their behavior is not deterministically know and possibly it is time varying. To counteract uncertainty adaptive schemes can be adopted, where design is not based on a fixed model of the system, but modeling is repeatedly updated based on the actually seen system response. This course illustrates some of the main techniques in adaptive control and filtering. Correspondingly, advance identification techniques to learn the system dynamics from measured data will be introduced and discussed.

 

Contents

1.         Adaptation and learning

 

Uncertainty, robustness, and adaptation in control. Principles of adaptive control. On-line and off-line schemes. Direct and indirect adaptive control.

 

2.         On-line adaptive control

 

The self-tuning paradigm. Certainty equivalence principle. Identification resumed: least-squares identification, recursive identification, forgetting factor. Main properties of the self-tuning scheme: convergence of the model estimate, stability and optimality. Examples.

 

3.         Off-line adaptive control

 

Direct design of the controller and the VRFT (Virtual Reference Feedback Tuning) approach. Frequency interpretation of PEM identification methods. Instrumental Variable identification. Optimality and sub-optimality of the VRFT controller. The VRFT toolbox of MATLAB.

 

4.         Advance topics on system identification and learning  

 

Identification of nonlinear systems by means of neural networks. Statistical learning theory. Data-mining and classification.

Prerequisites

Model identification and data analysis

Final exam is a written exam, constituted by five questions on the whole course programme. The duration is 2h.

Additional information available on the course web-site http://corsi.dei.polimi.it/IMAD/SAA/