Graduate course on Recursive identification, spring 2005
 
Area description
Recursive identification, also known as adaptive filtering, concerns methodologies how to adaptively update, often in real-time, mathematical models of dynamic systems. It has wide applications, perhaps particularly so in control systems and signal processing.
Course contents
Introduction. Different approaches to recursive identification. Tracking time-varying systems. A general framework for models and identification methods. Analysis of properties for time-invariant systems. User choices of algorithms. Implemenation aspects. Algorithmic details. Analysis of tracking time-varying systems.
Prerequisites
The course should be of interest for graduate students in automatic control, signal processing, systems theory, mathematical statistics, etc. The participants are assumed to have a basic knowledge of system identification using parametric methods, or parameter estimation for dynamic models, or time-series analysis.

Structure
The graduate course will be given during period 4, spring semester 2005. There will be one and sometimes two four hour session(s) per week. Each session will comprise a 2 hour lecture (partly of survey character), and a 2 hour part where the participants will demonstrate solutions to the homework assignments.
Venue
The sessions will take place in room 2114 (and twice in room 2244),  house 2, Polacksbacken, Uppsala.

Lecturer
Professor Torsten Söderström, email ts@it.uu.se.
Undergraduate students
If there is an interest, I will be happy to arrange also an abbreviated version of the course, suitable for undergraduate students.  
Literature
  • L Ljung and T Söderström: Theory and Practice of Recursive Identification, MIT Press, Cambridge, MA, 1983. The book is out of print, but we will arrange copies for the participants.
  • T Söderström: Basics about tracking time-varying dynamics.
  • T Söderström: Problems in recursive identification.
For those who are not registrated as graduate students at the Division of Systems and Control, there will be a fee of SEK 200 for the copied book and the additional material.
Registration
In order to arrange appropriate copying of the course literature, those interested to participate in the course are asked to inform Torsten Söderström (email ts@it.uu.se, phone 018-4713075), not later than March 11, 2005.
Examination details
  • The examination is based on both homework assignments, and a final take-home exam.
  • The participants' solutions to the homework assignments are to be presented and discussed during the problem solving sessions. These sessions are therefore an essential part of the course, and are also aimed to be helpful for the understanding of the material.
  • The take-home exam will be for a period of 2 days, and will be scheduled for the end of the period.
  • There are no grades (except pass and fail), as for any other PhD course.
  • To pass, a satisfactory performance of both the homework assignments (including presentations at the problem solving sessions) and the final exam is required.
The course will give 7 units in the graduate program. An undergraduate variant of the course would give 4 or 5 units.

Schedule
 
Pass Week Time, venue Contents
lecture		 Homework
type A		 Homework
type B
1 12 Tue March 22, 
13-17
2114
 Chapter 2
Approaches		 - -
2 13 Mon April 11, 
13-17
2244
 Chapter 3
Models and methods		 A2.1, A2.2 B2.1, B2.2
3 13 Thu April 14, 
13-17
2114
 Chapter 4
Analysis		 A3.1(C) B3.1
4 14 Tue April 19, 
13-17
2114
 Chapter 4, cont'd
Analysis		 A4.1, A4.2 B4.1
5 16 Mon April 25, 
13-17
2114
 Chapter 5
User choices		 A5.1, A5.2 B5.1, B5.2
6 17 Thu April 28,
13-17
2114
 - A6.1, A6.2 B6.1, B6.2
7 19 Tue May 03,
08-12
2114
 Chapter 6
Implementation		 A7.1, A7.2(C) B7.1(C)
8 19 Tue  May 10,
13-17
2244
 Notes
Tracking, accuracy		 A8.1 B8.1, B8.2
9 20 Tue May 17,
13-17
2114
 Notes, cont'd 
Tracking, accuracy		 A9.1(C) B9.1(C)
10 21 Tue May 24,
13-17
2114
 - A10.1, A10.2, A10.3 B10.1(C), B10.2

Last updated: 20 January 2005 by Torsten Söderström