System Identification
6.435
SET 6 – Parametrized model structures – One-step predictor – Identifiability
Munther A. Dahleh
Lecture 6 6.435, System Identification 1 Prof. Munther A. Dahleh Models of LTI Systems
• A complete model
u = input y = output e = noise (with PDF).
Lecture 6 6.435, System Identification 2 Prof. Munther A. Dahleh • A parametrized model
u = input y = output e = noise (with PDF of e). e white noise
Lecture 6 6.435, System Identification 3 Prof. Munther A. Dahleh One Step Linear Predictor
•
• Want to find that minimizes
•
• Minimum Prediction Error Paradigm
Data:
Estimate of at time N
Lecture 6 6.435, System Identification 4 Prof. Munther A. Dahleh One Step Linear Predictor (Derivation) The predicted output has the form
Both & have a delay. Past inputs and outputs are mapped to give the new predicted output.
Notice that:
Lecture 6 6.435, System Identification 5 Prof. Munther A. Dahleh Now:
has at least has at least one delay one delay
The lower bound is achieved if z = 0
Equivalently
Lecture 6 6.435, System Identification 6 Prof. Munther A. Dahleh Result
Lecture 6 6.435, System Identification 7 Prof. Munther A. Dahleh Examples of Transfer Function Models
–ARX (Autoregressive with exogenous input) • Description
•Matched with the model
Lecture 6 6.435, System Identification 8 Prof. Munther A. Dahleh Examples … ARX
•One step predictor
(a function of past data)
• Prediction error
Lecture 6 6.435, System Identification 9 Prof. Munther A. Dahleh Examples …. ARMAX
–ARMAX (Autoregressive moving average with exogenous input) • Description
• Standard model
• More general, includes ARX model structure.
Lecture 6 6.435, System Identification 10 Prof. Munther A. Dahleh Examples …. ARMAX
•One step predictor
or
• Pseudo-linear Regression past predictions
where or simply Not linear in
Lecture 6 6.435, System Identification 11 Prof. Munther A. Dahleh Examples …. OE
• OE (Output Error) • Description
•One step predictor
•Standard
Lecture 6 6.435, System Identification 12 Prof. Munther A. Dahleh Examples …. OE
• Nonlinear Regression Vector
Define
Lecture 6 6.435, System Identification 13 Prof. Munther A. Dahleh Examples …. Box-Jenkins
An even more general model
Lecture 6 6.435, System Identification 14 Prof. Munther A. Dahleh Examples …. State-Space
–State-Space Models • Description
disturbance •Noise: two components Output noise Usual assumptions
w, v are white.
Lecture 6 6.435, System Identification 15 Prof. Munther A. Dahleh • One Step Predictor = Kalman Filter
is a positive semi-definite solution of the steady-state Riccati equation:
(error covariance)
Lecture 6 6.435, System Identification 16 Prof. Munther A. Dahleh Innovation Representation
In general
Lecture 6 6.435, System Identification 17 Prof. Munther A. Dahleh General Notation
Define
It follows
Describes the model structure
Predictor:
Lecture 6 6.435, System Identification 18 Prof. Munther A. Dahleh For a given
• Notice: can be stable even though is not!
•
Lecture 6 6.435, System Identification 19 Prof. Munther A. Dahleh Predictor Models
Def: A predictor model is a linear time-invariant stable filter that defines a predictor
Def: A complete probabilistic model of a linear time-invariant system is a pair of a predictor model and the PDF associated with the prediction error (noise).
In most situations, is not complete known. We may work with means & variances.
Lecture 6 6.435, System Identification 20 Prof. Munther A. Dahleh Stability Requirements
Example ARX
If has zeros outside the disc, then the map from is unstable.
is always stable.
Lecture 6 6.435, System Identification 21 Prof. Munther A. Dahleh Model Sets
Def: A model set is a collection of models
Index set
Examples all linear models all models where is FIR of fixed order a finite set of models nonlinear fading memory models
Comment: These are “big” sets that are not necessarily parametrized in a nice way.
Lecture 6 6.435, System Identification 22 Prof. Munther A. Dahleh Model Structures
Model structures are parametrizations of model sets. We require this parametrization to be smooth.
Let a model be index by a parameter We require to be differentiable with respect to
Lecture 6 6.435, System Identification 23 Prof. Munther A. Dahleh Notice that
Def: A model structure is a differentiable map from a connected open subset to a model set such that the gradient, is defined and stable.
is the map is one particular model
Lecture 6 6.435, System Identification 24 Prof. Munther A. Dahleh Example: ARX model structure
stable
Lecture 6 6.435, System Identification 25 Prof. Munther A. Dahleh General Structure
Lecture 6 6.435, System Identification 26 Prof. Munther A. Dahleh You need
to be differentiable
is stable
(not sufficient).
Lecture 6 6.435, System Identification 27 Prof. Munther A. Dahleh Model Structures
Proposition: The parametrization
(= set of parameters of A , B , C , D , F )
restricted to the set
has no zeros outside the unit disc
is a model structure
Lecture 6 6.435, System Identification 28 Prof. Munther A. Dahleh Proof: Follows from the previous general derivative. Notice that H may be unstable!
Proposition: The Kalman Filter parametrization is a model structure if
(The stability property is equivalent to being detectable).
Lecture 6 6.435, System Identification 29 Prof. Munther A. Dahleh Independent Parametrization
A model structure m is independently parametrized if
Useful for giving a frequency domain interpretation of the estimate.
• We can define a model set as the range of a model structure:
• We can define unions of different model structures:
Useful for model structure determination!
Lecture 6 6.435, System Identification 30 Prof. Munther A. Dahleh Identifiability
Central question I: Does the identification procedure yield a unique parameter ? Data Identification procedure Informative Identifiability Question
Def: Model structure is identifiable (globally!) at if It is locally identifiable at if there exists an such that implies
Def: Model structure is strictly identifiable (local or global) if it is identifiable (local or global) for all
Lecture 6 6.435, System Identification 31 Prof. Munther A. Dahleh Central question II: Is the identified parameter equal to the “true parameters” ?
Parametrized structure: true system
Case I:
Case II:
Define:
Let for some . If is identifiable at , then
Lecture 6 6.435, System Identification 32 Prof. Munther A. Dahleh Identifiability of Model Structures
General:
Identifiable at
ARX:
If &
&
ARX is strictly identifiable
Lecture 6 6.435, System Identification 33 Prof. Munther A. Dahleh OE:
Suppose . Then
& iff (B, F ) are coprime.
(To do this cleanly, need to consider where & are the delay powers in both B & F )
Lecture 6 6.435, System Identification 34 Prof. Munther A. Dahleh Identifiability
Theorem:
is identifiable at iff
1) There are no common factors of 2) ” 3) ”
Lecture 6 6.435, System Identification 35 Prof. Munther A. Dahleh