SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.

Slides:



Advertisements
Similar presentations
The Average Case Complexity of Counting Distinct Elements David Woodruff IBM Almaden.
Advertisements

16. Mean Square Estimation
Tracking Unknown Dynamics - Combined State and Parameter Estimation Tracking Unknown Dynamics - Combined State and Parameter Estimation Presenters: Hongwei.
ELEC 303 – Random Signals Lecture 20 – Random processes
ELE Adaptive Signal Processing
Cox Model With Intermitten and Error-Prone Covariate Observation Yury Gubman PhD thesis in Statistics Supervisors: Prof. David Zucker, Prof. Orly Manor.
280 SYSTEM IDENTIFICATION The System Identification Problem is to estimate a model of a system based on input-output data. Basic Configuration continuous.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
SYSTEMS Identification
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
Multivariable Control Systems Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
SYSTEMS Identification
LINEAR CONTROL SYSTEMS Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
SYSTEMS Identification
SYSTEMS Identification
Digital Image Processing, 2nd ed. www. imageprocessingbook.com © 2001 R. C. Gonzalez & R. E. Woods 1 Objective To provide background material in support.
Multivariable Control Systems
SYSTEMS Identification
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
Topic4 Ordinary Least Squares. Suppose that X is a non-random variable Y is a random variable that is affected by X in a linear fashion and by the random.
Boyce/DiPrima 9th ed, Ch 11.2: Sturm-Liouville Boundary Value Problems Elementary Differential Equations and Boundary Value Problems, 9th edition, by.
LINEAR CONTROL SYSTEMS Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
Probability theory 2008 Outline of lecture 5 The multivariate normal distribution  Characterizing properties of the univariate normal distribution  Different.
Multivariable Control Systems
SYSTEMS Identification
Multivariable Control Systems Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
Christopher Dougherty EC220 - Introduction to econometrics (chapter 3) Slideshow: prediction Original citation: Dougherty, C. (2012) EC220 - Introduction.
1 10. Joint Moments and Joint Characteristic Functions Following section 6, in this section we shall introduce various parameters to compactly represent.
Adaptive Signal Processing
Linear Prediction Problem: Forward Prediction Backward Prediction
Review of Probability.
Simulation Output Analysis
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
TELECOMMUNICATIONS Dr. Hugh Blanton ENTC 4307/ENTC 5307.
1 Part 5 Response of Linear Systems 6.Linear Filtering of a Random Signals 7.Power Spectrum Analysis 8.Linear Estimation and Prediction Filters 9.Mean-Square.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Deterministic vs. Random Maximum A Posteriori Maximum Likelihood Minimum.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Definitions Random Signal Analysis (Review) Discrete Random Signals Random.
Elements of Stochastic Processes Lecture II
CHEE825 Fall 2005J. McLellan1 Spectral Analysis and Input Signal Design.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
Principles of the Global Positioning System Lecture 12 Prof. Thomas Herring Room ;
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
Charles University FSV UK STAKAN III Institute of Economic Studies Faculty of Social Sciences Institute of Economic Studies Faculty of Social Sciences.
Discrete-time Random Signals
Regression. We have talked about regression problems before, as the problem of estimating the mapping f(x) between an independent variable x and a dependent.
Lecture 5 – 6 Z - Transform By Dileep Kumar.
Geology 6600/7600 Signal Analysis 28 Sep 2015 © A.R. Lowry 2015 Last time: Energy Spectral Density; Linear Systems given (deterministic) finite-energy.
Chapter 2-OPTIMIZATION G.Anuradha. Contents Derivative-based Optimization –Descent Methods –The Method of Steepest Descent –Classical Newton’s Method.
Joint Moments and Joint Characteristic Functions.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory.
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
LINEAR CONTROL SYSTEMS Ali Karimpour Associate Professor Ferdowsi University of Mashhad.
LINEAR CONTROL SYSTEMS Ali Karimpour Assistant Professor Ferdowsi University of Mashhad.
Eeng360 1 Chapter 2 Linear Systems Topics:  Review of Linear Systems Linear Time-Invariant Systems Impulse Response Transfer Functions Distortionless.
Week 21 Statistical Model A statistical model for some data is a set of distributions, one of which corresponds to the true unknown distribution that produced.
11-1 Lyapunov Based Redesign Motivation But the real system is is unknown but not necessarily small. We assume it has a known bound. Consider.
LINEAR CONTROL SYSTEMS
LINEAR CONTROL SYSTEMS
Ali Karimpour Associate Professor Ferdowsi University of Mashhad
LINEAR CONTROL SYSTEMS
§1-2 State-Space Description
SYSTEMS Identification
Linear Systems Review Objective
LINEAR CONTROL SYSTEMS
LINEAR CONTROL SYSTEMS
16. Mean Square Estimation
LINEAR CONTROL SYSTEMS
Presentation transcript:

SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory For The User” Lennart Ljung

lecture 1 Ali Karimpour Nov Lecture 8 Convergence & Consistency Topics to be covered include: v Conditions on the Data Set v Prediction-Error Approach v Consistency and Identifiability v LTI Models: A Frequency-Domain Description of the Limit Model v The Correlation Approach lecture 8

lecture 1 Ali Karimpour Nov Introduction lecture 8 Previously different methods to determine models from data are described. To use these methods in practice, we need insight into their properties: How well will the identified model describe the actual system? Are some identification methods better than others? How should the design variables be chosen? The questions depend to mapping: 1- Simulation. 2- Analysis.

lecture 1 Ali Karimpour Nov Lecture 8 Convergence & Consistency Topics to be covered include: v Conditions on the Data Set v Prediction-Error Approach v Consistency and Identifiability v LTI Models: A Frequency-Domain Description of the Limit Model v The Correlation Approach lecture 8

lecture 1 Ali Karimpour Nov lecture 8 Conditions on the Data Set Data generation configuration Let Analysis means:

lecture 1 Ali Karimpour Nov Conditions on the Data Set lecture 8 D1 A Technical Condition D1

lecture 1 Ali Karimpour Nov Conditions on the Data Set lecture 8 Remind:

lecture 1 Ali Karimpour Nov Conditions on the Data Set lecture 8 S A True System S

lecture 1 Ali Karimpour Nov Conditions on the Data Set lecture 8 Exercise1: Prove the lemma. When S1 holds, a more explicit version of conditions D1 can be given

lecture 1 Ali Karimpour Nov Conditions on the Data Set lecture 8 Information Content in the Data Set. Whether the data set Z  allows us to distinguish between different models in the set? The data set is informative if it is capable of distinguishing between different models. Definition: A quasi-stationary data set Z  is informative enough with respect to the model set M * if, for any two models W 1 (q) and W 1 (q) in the set, Definition: A quasi-stationary data set Z  is informative if it is informative enough with respect to the model set L *, consisting of all linear, time invariant models. Recall:

lecture 1 Ali Karimpour Nov Conditions on the Data Set lecture 8 Remember Theorem: A quasi-stationary data set Z  is informative if the spectrum matrix for z(t)=[u(t) y(t)] T is strictly positive definite for almost all . Proof: We need to show 2-65

lecture 1 Ali Karimpour Nov Lecture 8 Convergence & Consistency Topics to be covered include: v Conditions on the Data Set v Prediction-Error Approach v Consistency and Identifiability v LTI Models: A Frequency-Domain Description of the Limit Model v The Correlation Approach lecture 8

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8 In PEM Clearly it depends to For quadratic criterion and a linear, uniformly stable model structure M, we have Using D1 Condition: Uniformly Stable since M is stable

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8 Since D M is compact It may happen that does not have a unique/global minimum.

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8 Example 8.1 Bias in ARX Structures Suppose that the system is given by Where {u(t)} and {e 0 (t)} are independent white noises with unit variances. Let the model structure be given by The prediction-error variance is Exercise2 : Proof I

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8 Example 8.1 Bias in ARX Structures The prediction-error variance is But for true value of parameters When we apply PEM the estimates will converge, according to pervious theorem. But one of the parameters is biased. It is clear that bias is beneficial for the prediction performance of the model. So it gives a strictly better predictor. Exercise3 : Proof and explain the example by suitable simulation.

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8 Example 8.2 Wrong Time Delay Consider the system {e(t)} and {w(t)} are independent white-noise sequence with unit variances. Let the model structure be given by The associated prediction-error variance is: Exercise4 : Proof I

lecture 1 Ali Karimpour Nov Prediction-Error Approach lecture 8 Example 8.2 Wrong Time Delay Consider the system Hence Now the predictor Is a fairly reasonable one for

lecture 1 Ali Karimpour Nov Lecture 8 Convergence & Consistency Topics to be covered include: v Conditions on the Data Set v Prediction-Error Approach v Consistency and Identifiability v LTI Models: A Frequency-Domain Description of the Limit Model v The Correlation Approach lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov The first condition: lecture 8 Ali Karimpour Nov 2009 Consistency and Identifiability Exercise5: Prove the above-mentioned theorem. Suppose that assumption S1 holds so that we have a true system. Will PE identification recover the true plant? So

lecture 1 Ali Karimpour Nov lecture 8 Ali Karimpour Nov 2009 Consistency and Identifiability

lecture 1 Ali Karimpour Nov Consistency and Identifiability lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov Consistency and Identifiability lecture 8 Ali Karimpour Nov 2009 Example 8.3 First Order Output Error Model Suppose that the true system is given by Let we define a first-order output error model as: By pervious theorem the estimates of and will converges to the true value of and.

lecture 1 Ali Karimpour Nov Lecture 8 Convergence & Consistency Topics to be covered include: v Conditions on the Data Set v Prediction-Error Approach v Consistency and Identifiability v LTI Models: A Frequency-Domain Description of the Limit Model v The Correlation Approach lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 An Expression for Under assumption S1 we have

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 An Expression for Monic Independent So Overbar is complex conjugate.

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 An Expression for Let So

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 We now have a characterization of in the frequency domain. with indicated weightings.

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 Open Loop Case

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 Open Loop Case Consider an independently noise model with Spectral factorization

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model Open Loop Case =

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model Open Loop Case

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 Closed Loop Case

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model Closed Loop Case =

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 Example 8.5 Approximation in the Frequency Domain

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model Example 8.5 Approximation in the Frequency Domain

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model Example 8.5 Approximation in the Frequency Domain

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model Example 8.5 Approximation in the Frequency Domain

lecture 1 Ali Karimpour Nov Lecture 8 Convergence & Consistency Topics to be covered include: v Conditions on the Data Set v Prediction-Error Approach v Consistency and Identifiability v LTI Models: A Frequency-Domain Description of the Limit Model v The Correlation Approach lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 In section 7.5 we defined the correlation approach to identification,with the special cases of PLR and IV methods.

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 Basic Convergence Result Consider the function And correlation vector ξ(t,θ) is obtained by linear filtering of past data: where

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009 Instrumental-variable Methods

lecture 1 Ali Karimpour Nov The Correlation Approach lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov The Correlation Approach lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov The Correlation Approach lecture 8 Ali Karimpour Nov 2009

lecture 1 Ali Karimpour Nov LTI Models: A Frequency-Domain Description of the Limit Model lecture 8 Ali Karimpour Nov 2009