Contact



 

Electrical Engineering



People & Organization »

Research »

Education »

News

Centers

Publications

Open positions

Intranet

KTH / Electrical Engineering

FEL 3200 System Identification (2011)

12 credits
 


News: - The deadlines for homework exercise sets 5 and 6 have been postponed. (2011-04-26)


 

Course objectives

The aim of this course is to provide the fundamentals of data based modeling, for students specializing

in related areas, such as control theory and signal processing.

Personnel

Cristian Rojas, cristian.rojas@ee.kth.se, 08-790 7427

Course literature and software

The course follows the book

Lennart Ljung. System Identification: Theory for the User, Second Edition, 1999.

An excellent complement is

Torsten Söderström and Petre Stoica. System Identification, 1989 (available for download here).

The Matlab toolbox on system identification will be used throughout the course.

Venue

The lectures will in general be held in one of the seminar rooms of Floor 3, Osquldas väg 10, as stated in the schedule below.

Schedule

Note: The schedule is preliminary. It will be updated as the course advances.

 

 

Day

Time

Place

Content

References

Lec 1

Wed

Mar 2, 13.00-15.00

Conference Room Registry

(floor 3)

Introduction

Presentation of basic ideas and concepts, review of probability and statistics.

Course book: 1, App. I, II

Optional: (Ljung,2010)

Lec 2

Tue

Mar 8, 13.15-15.00

Conference Room SIP

(floor 3)

Signals and Systems

Linear time invariant systems and their uses (prediction, simulation).

Course book: 2, 3

Lec 3

Tue

Mar 15, 13.15-15.00

Conference Room Registry

(floor 3)

Models and Nonparametric Methods

Models of linear time invariant systems. Nonparametric time- and frequency-domain methods.

Course book: 4, 6

Optional: Lemma 2.1, (Ljung,1985)

Lec 4

Tue

Mar 22, 13.15-15.00

Conference Room Registry

(floor 3)

Parametric Estimation Methods

Least squares, prediction error methods, instrumental variables and subspace methods.

Course book: 7, 10.6

Optional: Slides 1, Slides 2, Slides 3, Slides 4, Slides 5, Max-Max problem, (Ljung,1978)

Ex 1

Fri

Mar 25

 

 

2E1,2E5,3E2

Lec 5

Tue

Mar 29, 13.15-15.00

Conference Room Registry

(floor 3)

Asymptotic Properties

Consistency, convergence and asymptotic normality of estimation methods.

Course book: 8, 9

Optional: Slides 6, Slides 7

Ex 2

Fri

Apr 1

 

 

4E5,6G1,6E3

Note: For Ex. 6E3, conditions (a) and (b) should hold simultaneously

Lec 6

Tue

Apr 5, 13.15-15.00

Conference Room Registry

(floor 3)

Experiment Design and Closed-Loop Identification

Bias/variance issues, experiment design, closed-loop identification

Course book: 12, 13

Optional: Slides 8, Slides 9

Ex 3

Fri

Apr 8

 

 

7E2,7E8,7G1

Lec 7

Tue

Apr 12,

12.45-

15.00

Conference Room Registry

(floor 3)

Data Preprocessing and Choice of Identification Criterion

Course book: 14, 15

Ex 4

Fri

Apr 15

 

 

8G4,8E3,9E3 (ignore question about PLR method)

Lec 8

Tue

Apr 19, 13.15-15.00

Conference Room Registry

(floor 3)

Model Structure Selection and Validation

Course book: 16

Optional: Slides 10, Slides 11

Ex 5

Fri

Apr 29

 

 

13G3,13E2,13E5

Ex 6

Tue

May 3

 

 

14E1,15E1

Ex 7

Fri

May 6

 

 

16E2,16E3

 

Additional reading

M. S. Arulampalam, S. Maskell, N. Gordon and T. Clapp. “A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking”. IEEE Transactions on Signal Processing, Vol. 50(2), pp. 174-188, 2002.

E. Chaumette, J. Galy, A. Quinlan and P. Larzabal. “A New Barankin Bound Approximation for the Prediction of the Threshold Region Performance of Maximum Likelihood Estimators”. IEEE Transactions on Signal Processing, Vol. 56(11), pp. 5319-5333, 2008.

B. Efron and C. Morris. “Stein’s Paradox in Statistics”. Scientific American, Vol. 236(5),  pp. 119-127, 1977.

S. Kay and Y. C. Eldar. “Rethinking Biased Estimation”. IEEE Signal Processing Magazine, Vol. 25(3), pp. 133-136, 2008.

L. Ljung. “Convergence Analysis of Parametric Identification Methods”. IEEE Transactions on Automatic Control, Vol. 23(5), pp. 770-783, 1978.

L. Ljung. “On the Estimation of Transfer Functions”. Automatica, Vol. 21(6), pp. 677-696, 1985.

L. Ljung. “Model Validation and Model Error Modeling”. In B. Wittenmark and A. Rantzer (Eds.), The Åström Symposiium on Control, pp. 15-42, 1999.

L. Ljung. “Perspectives on System Identification”. Annual Reviews in Control, Vol. 34(1), pp. 1-12, 2010.

B. Ninness. “Some system identification challenges and approaches”. 15th IFAC Symposium on System Identification, Saint Malo, France, 2009.

B. Ninness and G. C. Goodwin. “Estimation of Model Quality”. Automatica, Vol. 31(12), pp. 1771-1797, 1995.

J. Mårtensson. Geometric analysis of stochastic model errors in system identification. PhD Thesis, KTH, 2007.

J. Rissanen. “Modeling By Shortest Data Description”. Automatica, Vol. 14(5), pp. 465-471, 1978.

T. Söderström. “On Model Structure Testing in System Identification”. International Journal of Control, Vol. 26(1), pp. 1-18, 1977.

T. Söderström. Computational Methods for Evaluating Covariance Functions. Notes for a System Identification course, Uppsala Universitet, 2003. Available here.

T. Söderström and P. Stoica. On Covariance Function Tests Used in System Identification”. Automatica, Vol. 26(1), pp. 125-133, 1990.

P. Stoica. “A Test for Whiteness”. IEEE Transactions on Automatic Control, Vol. 22(6), pp. 992-993, 1977.

P. Stoica, P. Eykhoff, P. Janssen and T. Söderström. ”Model-Structure Selection by Cross-Validation”. International Journal of Control, Vol. 43(6), pp. 1841-1878, 1986.

P. Stoica and B. Ottersten. “The evil of superefficiency”. Signal Processing, Vol. 55, pp. 133-136, 1996.

B. Wahlberg and L. Ljung. “Design variables for bias distribution in transfer function estimation”. IEEE Transactions on Automatic Control, Vol. 31(2), pp. 134-144, 1986.

 

 

 

Projects

To be announced.

 

Examination

  • 72 h take home exam.
  • 80% on weekly home-work problems. They should be handed in on Friday two weeks after the corresponding chapter has been discussed in class (except for Exercise Set 5; see schedule).
  • 1 project, preferably on a problem related to your own research. Feel free to come with suggestions.