Escolar Documentos
Profissional Documentos
Cultura Documentos
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 1/17
Lecture 4: Well-Posedness and Internal Stability
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 1/17
One Degree of Freedom Feedback Control System
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s)up + d,up = u + di
u = K(s) r − n − y
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
u = K(s) r − n − P (s) u + di + d
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
u = r − n − P (s) u + di + d
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
u + P (s)u = r − n − P (s)di − d
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
−1
u = 1 + P (s) [r − n − P (s)di − d]
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
−1
u = 1 + P (s) [r − n − P (s)di − d]
s−1
−1
s−1
u= 1− r−n+ di − d
s+2 s+2
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
−1
u = 1 + P (s) [r − n − P (s)di − d]
−1
s + 2 − (s − 1) s−1
u= r−n+ di − d
s+2 s+2
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Example of a not realizable system
Suppose that
s−1
P (s) = − , K(s) = 1
s+2
y = P (s) u + di + d
u = K(s) r − n − y
−1
u = 1 + P (s) [r − n − P (s)di − d]
−1
s + 2 − (s − 1) s−1
u= r−n+ di − d
s+2 s+2
s + 2 s−1
u= r−n−d + di
3 3
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 3/17
Concept of Well-Posedness:
r, n, di , d
y and u
are proper (the degree of the numerator is not bigger than the
degree of the denominator).
If this is not true, the system is not physically realizable.
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 4/17
Combining external signals
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 5/17
Combining external signals
K̂(s) = −K(s), w1 = d i , w2 = n + d − r
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 5/17
One Degree of Freedom Feedback Control System
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 6/17
One Degree of Freedom Feedback Control System
e1 = w1 + K̂(s)e2 , e2 = w2 + P (s)e1
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 6/17
One Degree of Freedom Feedback Control System
e1 = w1 + K̂(s)e2 , e2 = w2 + P (s)e1
−1 h i
e1 = w1 +K̂ w2 + P e1 ⇒ e1 = I − K̂(s)P (s) w1 + K̂(s)w2
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 6/17
One Degree of Freedom Feedback Control System
e1 = w1 + K̂(s)e2 , e2 = w2 + P (s)e1
−1 h i
e1 = I − K̂(s)P (s) w1 + K̂(s)w2
−1
e2 = I − P (s)K̂(s) w2 + P (s)w1
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 6/17
Criteria for Well-Posedness
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 7/17
Criteria for Well-Posedness
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 7/17
Criteria for Well-Posedness
is invertible.
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 7/17
Criteria for Well-Posedness
is invertible.
It is sufficient to have either K̂(s) or P (s) strictly proper.
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 7/17
Lecture 4: Concepts of Well-Posedness and Internal
Stability
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 8/17
Concept of Internal Stability of Feedback System:
The well-posed feedback system
e1 = w1 + K̂(s)e2 , e2 = w2 + P (s)e1
or the well-posed feedback system
−1 h i
e1 = I − K̂(s)P (s) w1 + K̂(s)w2
−1
e2 = I − P (s)K̂(s) P (s)w1 + w2
belong to RH∞ .
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 9/17
Example
Suppose that
s−1 1
P (s) = − , K(s) =
s+1 s−1
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 10/17
Example
Suppose that
s−1 1
P (s) = − , K(s) =
s+1 s−1
Then
" # −1 −1 " #
e1 I − K̂(s)P (s) I − K̂(s)P (s) K̂(s) w1
= −1 −1
e2 I − P (s)K̂(s) P (s) I − P (s)K̂(s) w2
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 10/17
Example
Suppose that
s−1 1
P (s) = − , K(s) =
s+1 s−1
Then
" # −1 −1 " #
e1 I − K̂(s)P (s) I − K̂(s)P (s) K̂(s) w1
= −1 −1
e2 I − P (s)K̂(s) P (s) I − P (s)K̂(s) w2
s+1 −(s + 1)
" #
s+2 (s − 1)(s + 2) w1
=
s−1 s+1 w2
s+2 s+2
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 10/17
Example
Suppose that
s−1 1
P (s) = − , K(s) =
s+1 s−1
Then
" # −1 −1 " #
e1 I − K̂(s)P (s) I − K̂(s)P (s) K̂(s) w1
= −1 −1
e2 I − P (s)K̂(s) P (s) I − P (s)K̂(s) w2
s+1 −(s + 1)
" #
s+2 (s − 1)(s + 2) w1
=
s−1 s+1 w2
s+2 s+2
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 10/17
Co-Prime Factorization
Two polynomials
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 11/17
Co-Prime Factorization
Two polynomials
If p(s) and q(s) are co-prime, then there are two polynomials
x(s) and y(s) such that
x(s)p(s) + y(s)q(s) = 1
and the pair ( x(s) , y(s) ) can be found from the (reversed)
Euclid’s algorithm.
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 11/17
Co-Prime Factorization
Two transfer functions p(s) ∈ RH∞ , q(s) ∈ RH∞ are
called co-prime if there are two new transfer functions
x(s) ∈ RH∞ and y(s) ∈ RH∞ such that
x(s)p(s) + y(s)q(s) = 1
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 12/17
Co-Prime Factorization
Two transfer functions p(s) ∈ RH∞ , q(s) ∈ RH∞ are
called co-prime if there are two new transfer functions
x(s) ∈ RH∞ and y(s) ∈ RH∞ such that
x(s)p(s) + y(s)q(s) = 1
such that
then
h−1 (s) ∈ RH∞
i.e. h(s) has neither zeros no poles in the closed right-half
plane (such transfer functions are called minimum-phase).
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 12/17
Co-Prime Factorization
Two matrix transfer functions M (s), N (s) ∈ RH∞ are called
right co-prime over RH∞ if
• they have the same number of columns
• there are matrices Xr (s) , Yr (s) ∈ RH∞ such that
" #
M (s)
Xr (s), Yr (s) = Xr (s)M (s)+Yr (s)N (s) = I
N (s)
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 13/17
Co-Prime Factorization
Two matrix transfer functions M (s), N (s) ∈ RH∞ are called
right co-prime over RH∞ if
• they have the same number of columns
• there are matrices Xr (s) , Yr (s) ∈ RH∞ such that
" #
M (s)
Xr (s), Yr (s) = Xr (s)M (s)+Yr (s)N (s) = I
N (s)
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 13/17
Defining a Co-Prime Factorization
Given a matrix transfer function G(s) , then
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 14/17
Defining a Co-Prime Factorization
Given a matrix transfer function G(s) , then
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 14/17
Defining a Co-Prime Factorization
Given a matrix transfer function G(s) , then
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 14/17
Computing a Co-Prime Factorization " #
A B
Given a matrix transfer function G(s) = with
C D
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 15/17
Computing a Co-Prime Factorization " #
A B
Given a matrix transfer function G(s) = with
C D
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 16/17
Using a Co-Prime Factorization
Given plant and controller matrix transfer functions
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 16/17
Using a Co-Prime Factorization
Given plant and controller matrix transfer functions
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 16/17
Using a Co-Prime Factorization
Given plant and controller matrix transfer functions
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 16/17
Using a Co-Prime Factorization
Given plant and controller matrix transfer functions
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 17/17
Next Lecture / Assignments:
(s−1)
Problem: Let G(s) = (s+2)(s−3)
. Find a stable co-prime
n(s)
factorization G(s) = m(s)
and x, y ∈ RH∞ such that
xn + ym = 1 .
c A. Shiriaev/L. Freidovich. January 29, 2010. Optimal Control for Linear Systems: Lecture 4 – p. 17/17