System Identification

linma2875  2022-2023  Louvain-la-Neuve

System Identification
5.00 credits
30.0 h + 30.0 h
Q2
Teacher(s)
Bianchin Gianluca;
Language
Prerequisites
This courses assumes familiarity with transfer functions, as taught in
LINMA1510 (Linear Control) or LEPL1106 (Applied mathematics : Signals
and systems)
Main themes
This class is an introduction to system identification, which consists in finding an appropriate representation of a dynamical system using appropriate measurements. It will cover some of the main parametric and nonparametric methods for identifying dynamical systems, including in closed loop. It will also cover the properties of signals and model classes that are relevant for system identification. A realistic identification project will give students the opportunity to apply and implement the techniques that they will have learned.
Learning outcomes

At the end of this learning unit, the student is able to :

1 With respect to the L.O. framework, this class contributes to the developpement of the following learning outcomes
  • AA1.1, AA1.2, AA1.3
  • AA2.1, AA2.4
  • AA3.2
  • AA5.3, AA5.5
More precisely, by the end of the class, the student will be able to :
  • recognize a problem of system identificaiton
  • propose and implement solutions to simple identification problems
  • identify a dynamical systems using input-output data
  • validate a model of system that has been identified, and compare different simple models
  • design an experiment to identify a simple system
  • develop a deeper understanding of system identification by him/herself if necessary  in order to solve more complex problems
Transversal learning outcomes :
  • Handling unforeseen technical issues that appear when treating a real-world problem
  • Making reasonable hypothesis for a given problem, and evaluating them a posteriori
  • Taking part to a technical class in English
 
Content
The following topics will be covered
  • Nonparametric methods: temporal analysis, frequential analysis, including Fourier and spectral analysis
  • Main classes of LTI systems and their properties, including the notions of identifiability and predictors
  • Certain parametric methods: linear regression, instrumental variables, prediction errors, and some statistical methods including the maximum likelihood method
  • The properties of (input) signal, including the notion of information content of the signals and the level of persistence of excitation.
  • The convergence of the method seen
  • Identification techniques for systems controlled in closed loop
Teaching methods
  • Regular lectures.
  • Resolutions of simple problems under the supervison of teaching assistant in order to get familiar with new concepts.
  • Problem sets to be solved in small group in order to develop a deeper understanding of the concepts.
The activities above take place in a classroom, but may be organized partly or entirely remotely if required by the sanitary situation or by practical constraints.
Evaluation methods
The grade will be based on
  • Written exam. The exam may be replaced by a remote oral exam in case required by sanitary situation or by practical constraints.
  • Projects and problem sets assigned during the year. These are graded based on the student's demonstration of his/her understanding of the problem and its solution. 
To constitute the final grade, the weighting given to the assessments is:
• 75% to the written exam;
• 25% to the project(s) and or homework(s).
Other information
The lectures and problem sessions are in English, and all documents are in English. Homework, exams, and project reports can be written in English or French.
Students are expected to be familiar with dynamical systems and transfer functions.
 
Bibliography
Le cours s'appuie sur un syllabus disponible sur Moodle
Des livres de références sont également proposés :
  1. L. Ljung System Identification - Theory for the user, Prentice Hall, 1999. (disponible en bibliothèque)
  2. T. Soderstorm and P Stoica, System Identification (http://user.it.uu.se/~ts/sysidbook.pdf)
  3. P. van Overschee and B. de Moor  - Subspace Identification for Linear Systems: Theory, Implementation, Applications, Springer, 2011.
Faculty or entity
MAP


Programmes / formations proposant cette unité d'enseignement (UE)

Title of the programme
Sigle
Credits
Prerequisites
Learning outcomes
Master [120] in Biomedical Engineering

Master [120] in Mechanical Engineering

Master [120] in Electrical Engineering

Master [120] in Electro-mechanical Engineering

Master [120] in Mathematical Engineering

Master [120] in Data Science Engineering

Master [120] in Data Science: Information Technology