Difference between revisions of "Frequency theorem"
(Importing text file) |
m (fix tex) |
||
(One intermediate revision by one other user not shown) | |||
Line 1: | Line 1: | ||
+ | <!-- | ||
+ | f0417101.png | ||
+ | $#A+1 = 47 n = 0 | ||
+ | $#C+1 = 47 : ~/encyclopedia/old_files/data/F041/F.0401710 Frequency theorem | ||
+ | Automatically converted into TeX, above some diagnostics. | ||
+ | Please remove this comment and the {{TEX|auto}} line below, | ||
+ | if TeX found to be correct. | ||
+ | --> | ||
+ | |||
+ | {{TEX|auto}} | ||
+ | {{TEX|done}} | ||
+ | |||
A theorem that states conditions for the solvability of the Lur'e equations in control theory: | A theorem that states conditions for the solvability of the Lur'e equations in control theory: | ||
− | + | $$ \tag{1 } | |
+ | P ^ {*} H + H P + h h ^ {*} = G ,\ \ | ||
+ | H q - h \kappa = g , | ||
+ | $$ | ||
− | where | + | where $ P $, |
+ | $ G = G ^ {*} $, | ||
+ | $ q $, | ||
+ | $ g $, | ||
+ | $ \kappa $ | ||
+ | are given $ n \times n $, | ||
+ | $ n \times n $, | ||
+ | $ n \times m $, | ||
+ | $ n \times m $, | ||
+ | and $ m \times m $ | ||
+ | matrices respectively, and $ H = H ^ {*} $, | ||
+ | $ h $ | ||
+ | are the required $ n \times n $ | ||
+ | and $ n \times m $ | ||
+ | matrices. The Lur'e equations have two other equivalent forms: If $ \mathop{\rm det} \kappa \neq 0 $, | ||
− | + | $$ \tag{2 } | |
+ | H Q _ {0} H + ( P _ {0} ^ {*} H + H P _ {0} ) + G _ {0} = 0 , | ||
+ | $$ | ||
− | where | + | where $ Q _ {0} = Q _ {0} ^ {*} \geq 0 $, |
+ | $ G _ {0} = G _ {0} ^ {*} $, | ||
+ | and in the general case | ||
− | + | $$ \tag{3 } | |
+ | 2 \mathop{\rm Re} x ^ {*} H ( P x + q \xi ) = {\mathcal G} ( x , \xi ) | ||
+ | - | h ^ {*} x - \kappa \xi | ^ {2} \ \ | ||
+ | ( \forall x , \xi ) , | ||
+ | $$ | ||
− | where | + | where $ {\mathcal G} ( x , \xi ) $ |
+ | is a given Hermitian form of two vectors $ x \in \mathbf C ^ {n} $, | ||
+ | $ \xi \in \mathbf C ^ {m} $; | ||
− | + | $$ | |
+ | {\mathcal G} ( x , \xi ) = x ^ {*} G x + 2 | ||
+ | \mathop{\rm Re} ( x ^ {*} g \xi ) + \xi ^ {*} \Gamma \xi , | ||
+ | $$ | ||
− | Moreover, | + | Moreover, $ \Gamma = \kappa ^ {*} \kappa \geq 0 $, |
+ | $ G _ {0} = g \Gamma ^ {-1} g ^ {*} - G $, | ||
+ | $ P _ {0} = P - g \Gamma g ^ {*} $, | ||
+ | $ Q _ {0} = q \Gamma ^ {-1} q ^ {*} $. | ||
− | Let the pair | + | Let the pair $ \{ P , q \} $ |
+ | be controllable: $ \mathop{\rm rank} \| q , Pq \dots P ^ {n-1} q \| = n $. | ||
+ | Then the Lur'e equations reduce to the case where | ||
− | + | $$ | |
+ | P = \mathop{\rm diag} [ \lambda _ {1} \dots \lambda _ {h} ] ,\ \ | ||
+ | \lambda _ {j} + \lambda _ {h} \neq 0 ,\ \lambda _ {j} \in \mathbf R . | ||
+ | $$ | ||
− | If | + | If $ m = 1 $ |
+ | and all the matrices are real, the Lur'e equations in scalar notation take the form | ||
− | + | $$ | |
+ | \sum _ { k=1} ^ { n } q _ {k} | ||
+ | \frac{h _ {j} h _ {k} }{\lambda _ {j} + \lambda _ {k} } | ||
+ | - h _ {j} \sqrt \Gamma = \gamma _ {j} ,\ j = 1 \dots n ; | ||
+ | $$ | ||
− | here | + | here $ h = [ h _ {1} \dots h _ {n} ] $ |
+ | is the required vector. | ||
The frequency theorem asserts that for the Lur'e equations to be solvable it is necessary and sufficient that | The frequency theorem asserts that for the Lur'e equations to be solvable it is necessary and sufficient that | ||
− | + | $$ | |
+ | {\mathcal G} [ ( i \omega I - P ) ^ {-1} q \xi , \xi ] \geq 0 | ||
+ | $$ | ||
− | for all | + | for all $ \xi \in \mathbf C ^ {m} $, |
+ | $ \omega \in \mathbf R ^ {1} $, | ||
+ | $ \mathop{\rm det} \| i \omega I - P \| \neq 0 $ | ||
+ | ($ I $ | ||
+ | is the identity matrix). The frequency theorem also formulates a procedure for determining the matrices $ H $ | ||
+ | and $ h $ | ||
+ | and asserts that if | ||
− | + | $$ | |
+ | \mathop{\rm det} \Gamma \neq 0 ,\ \mathop{\rm det} \| i \omega I - | ||
+ | P \| \neq 0 ,\ {\mathcal G} [ \| i \omega I - P \| ^ {-1} q \xi ,\ | ||
+ | \xi ] > 0 | ||
+ | $$ | ||
− | (for all | + | (for all $ \xi \neq 0 $, |
+ | and all $ \omega $), | ||
+ | then there exist (unique) matrices $ H $ | ||
+ | and $ h $ | ||
+ | such that (except for the case of equation (3)) the following is true: $ P + q \kappa ^ {-1} h ^ {*} $ | ||
+ | is a Hurwitz matrix (see [[#References|[3]]]). | ||
The Lur'e equations in the form (2) are also sometimes called the matrix algebraic Riccati equation. The frequency theorem is used when solving problems on absolute stability [[#References|[2]]], [[#References|[4]]]–[[#References|[6]]], control and adaptation (see, for example, [[#References|[7]]]–[[#References|[9]]]). | The Lur'e equations in the form (2) are also sometimes called the matrix algebraic Riccati equation. The frequency theorem is used when solving problems on absolute stability [[#References|[2]]], [[#References|[4]]]–[[#References|[6]]], control and adaptation (see, for example, [[#References|[7]]]–[[#References|[9]]]). | ||
Line 41: | Line 114: | ||
====References==== | ====References==== | ||
<table><TR><TD valign="top">[1]</TD> <TD valign="top"> A.I. Lur'e, "Some non-linear problems of the theory of automatic control" , Moscow-Leningrad (1951) (In Russian)</TD></TR><TR><TD valign="top">[2]</TD> <TD valign="top"> V.M. Popov, "Hyperstability of control systems" , Springer (1973) (Translated from Rumanian)</TD></TR><TR><TD valign="top">[3]</TD> <TD valign="top"> V.A. Yakubovich, "A frequency theorem in control theory" ''Sib. Math. J.'' , '''14''' : 2 (1973) pp. 265–289 ''Sibirsk. Mat. Zh.'' , '''14''' : 2 (1973) pp. 384–420</TD></TR><TR><TD valign="top">[4]</TD> <TD valign="top"> A.K. Gelig, G.A. Leonov, V.A. Yakubovich, "Stability of non-linear systems with a unique equilibrium state" , Moscow (1978) (In Russian)</TD></TR><TR><TD valign="top">[5]</TD> <TD valign="top"> , ''Methods for studing non-linear systems of automatic control'' , Moscow (1975) (In Russian)</TD></TR><TR><TD valign="top">[6]</TD> <TD valign="top"> D.D. Siljak, "Nonlinear systems. Parameter analysis and design" , Wiley (1969)</TD></TR><TR><TD valign="top">[7]</TD> <TD valign="top"> V.N. Fomin, A.L. Fradkov, V.A. Yakubovich, "Adaptive control of dynamic objects" , Moscow (1981) (In Russian)</TD></TR><TR><TD valign="top">[8a]</TD> <TD valign="top"> J.C. Willems, "Almost invariant subspaces: an approach to high gain feedback design I. Almost controlled invariant subspaces" ''IEEE Trans. Autom. Control'' , '''1''' (1981) pp. 235–252</TD></TR><TR><TD valign="top">[8b]</TD> <TD valign="top"> J.C. Willems, "Almost invariant subspaces: an approach to high gain feedback design I. Almost conditionally invariant subspaces" ''IEEE Trans. Autom. Control'' , '''5''' (1982) pp. 1071–1084</TD></TR><TR><TD valign="top">[9]</TD> <TD valign="top"> W. Coppel, "Matrix quadratic equations" ''Bull. Austr. Math. Soc.'' , '''10''' (1974) pp. 377–401</TD></TR></table> | <table><TR><TD valign="top">[1]</TD> <TD valign="top"> A.I. Lur'e, "Some non-linear problems of the theory of automatic control" , Moscow-Leningrad (1951) (In Russian)</TD></TR><TR><TD valign="top">[2]</TD> <TD valign="top"> V.M. Popov, "Hyperstability of control systems" , Springer (1973) (Translated from Rumanian)</TD></TR><TR><TD valign="top">[3]</TD> <TD valign="top"> V.A. Yakubovich, "A frequency theorem in control theory" ''Sib. Math. J.'' , '''14''' : 2 (1973) pp. 265–289 ''Sibirsk. Mat. Zh.'' , '''14''' : 2 (1973) pp. 384–420</TD></TR><TR><TD valign="top">[4]</TD> <TD valign="top"> A.K. Gelig, G.A. Leonov, V.A. Yakubovich, "Stability of non-linear systems with a unique equilibrium state" , Moscow (1978) (In Russian)</TD></TR><TR><TD valign="top">[5]</TD> <TD valign="top"> , ''Methods for studing non-linear systems of automatic control'' , Moscow (1975) (In Russian)</TD></TR><TR><TD valign="top">[6]</TD> <TD valign="top"> D.D. Siljak, "Nonlinear systems. Parameter analysis and design" , Wiley (1969)</TD></TR><TR><TD valign="top">[7]</TD> <TD valign="top"> V.N. Fomin, A.L. Fradkov, V.A. Yakubovich, "Adaptive control of dynamic objects" , Moscow (1981) (In Russian)</TD></TR><TR><TD valign="top">[8a]</TD> <TD valign="top"> J.C. Willems, "Almost invariant subspaces: an approach to high gain feedback design I. Almost controlled invariant subspaces" ''IEEE Trans. Autom. Control'' , '''1''' (1981) pp. 235–252</TD></TR><TR><TD valign="top">[8b]</TD> <TD valign="top"> J.C. Willems, "Almost invariant subspaces: an approach to high gain feedback design I. Almost conditionally invariant subspaces" ''IEEE Trans. Autom. Control'' , '''5''' (1982) pp. 1071–1084</TD></TR><TR><TD valign="top">[9]</TD> <TD valign="top"> W. Coppel, "Matrix quadratic equations" ''Bull. Austr. Math. Soc.'' , '''10''' (1974) pp. 377–401</TD></TR></table> | ||
− | |||
− | |||
====Comments==== | ====Comments==== |
Latest revision as of 12:59, 13 January 2021
A theorem that states conditions for the solvability of the Lur'e equations in control theory:
$$ \tag{1 } P ^ {*} H + H P + h h ^ {*} = G ,\ \ H q - h \kappa = g , $$
where $ P $, $ G = G ^ {*} $, $ q $, $ g $, $ \kappa $ are given $ n \times n $, $ n \times n $, $ n \times m $, $ n \times m $, and $ m \times m $ matrices respectively, and $ H = H ^ {*} $, $ h $ are the required $ n \times n $ and $ n \times m $ matrices. The Lur'e equations have two other equivalent forms: If $ \mathop{\rm det} \kappa \neq 0 $,
$$ \tag{2 } H Q _ {0} H + ( P _ {0} ^ {*} H + H P _ {0} ) + G _ {0} = 0 , $$
where $ Q _ {0} = Q _ {0} ^ {*} \geq 0 $, $ G _ {0} = G _ {0} ^ {*} $, and in the general case
$$ \tag{3 } 2 \mathop{\rm Re} x ^ {*} H ( P x + q \xi ) = {\mathcal G} ( x , \xi ) - | h ^ {*} x - \kappa \xi | ^ {2} \ \ ( \forall x , \xi ) , $$
where $ {\mathcal G} ( x , \xi ) $ is a given Hermitian form of two vectors $ x \in \mathbf C ^ {n} $, $ \xi \in \mathbf C ^ {m} $;
$$ {\mathcal G} ( x , \xi ) = x ^ {*} G x + 2 \mathop{\rm Re} ( x ^ {*} g \xi ) + \xi ^ {*} \Gamma \xi , $$
Moreover, $ \Gamma = \kappa ^ {*} \kappa \geq 0 $, $ G _ {0} = g \Gamma ^ {-1} g ^ {*} - G $, $ P _ {0} = P - g \Gamma g ^ {*} $, $ Q _ {0} = q \Gamma ^ {-1} q ^ {*} $.
Let the pair $ \{ P , q \} $ be controllable: $ \mathop{\rm rank} \| q , Pq \dots P ^ {n-1} q \| = n $. Then the Lur'e equations reduce to the case where
$$ P = \mathop{\rm diag} [ \lambda _ {1} \dots \lambda _ {h} ] ,\ \ \lambda _ {j} + \lambda _ {h} \neq 0 ,\ \lambda _ {j} \in \mathbf R . $$
If $ m = 1 $ and all the matrices are real, the Lur'e equations in scalar notation take the form
$$ \sum _ { k=1} ^ { n } q _ {k} \frac{h _ {j} h _ {k} }{\lambda _ {j} + \lambda _ {k} } - h _ {j} \sqrt \Gamma = \gamma _ {j} ,\ j = 1 \dots n ; $$
here $ h = [ h _ {1} \dots h _ {n} ] $ is the required vector.
The frequency theorem asserts that for the Lur'e equations to be solvable it is necessary and sufficient that
$$ {\mathcal G} [ ( i \omega I - P ) ^ {-1} q \xi , \xi ] \geq 0 $$
for all $ \xi \in \mathbf C ^ {m} $, $ \omega \in \mathbf R ^ {1} $, $ \mathop{\rm det} \| i \omega I - P \| \neq 0 $ ($ I $ is the identity matrix). The frequency theorem also formulates a procedure for determining the matrices $ H $ and $ h $ and asserts that if
$$ \mathop{\rm det} \Gamma \neq 0 ,\ \mathop{\rm det} \| i \omega I - P \| \neq 0 ,\ {\mathcal G} [ \| i \omega I - P \| ^ {-1} q \xi ,\ \xi ] > 0 $$
(for all $ \xi \neq 0 $, and all $ \omega $), then there exist (unique) matrices $ H $ and $ h $ such that (except for the case of equation (3)) the following is true: $ P + q \kappa ^ {-1} h ^ {*} $ is a Hurwitz matrix (see [3]).
The Lur'e equations in the form (2) are also sometimes called the matrix algebraic Riccati equation. The frequency theorem is used when solving problems on absolute stability [2], [4]–[6], control and adaptation (see, for example, [7]–[9]).
References
[1] | A.I. Lur'e, "Some non-linear problems of the theory of automatic control" , Moscow-Leningrad (1951) (In Russian) |
[2] | V.M. Popov, "Hyperstability of control systems" , Springer (1973) (Translated from Rumanian) |
[3] | V.A. Yakubovich, "A frequency theorem in control theory" Sib. Math. J. , 14 : 2 (1973) pp. 265–289 Sibirsk. Mat. Zh. , 14 : 2 (1973) pp. 384–420 |
[4] | A.K. Gelig, G.A. Leonov, V.A. Yakubovich, "Stability of non-linear systems with a unique equilibrium state" , Moscow (1978) (In Russian) |
[5] | , Methods for studing non-linear systems of automatic control , Moscow (1975) (In Russian) |
[6] | D.D. Siljak, "Nonlinear systems. Parameter analysis and design" , Wiley (1969) |
[7] | V.N. Fomin, A.L. Fradkov, V.A. Yakubovich, "Adaptive control of dynamic objects" , Moscow (1981) (In Russian) |
[8a] | J.C. Willems, "Almost invariant subspaces: an approach to high gain feedback design I. Almost controlled invariant subspaces" IEEE Trans. Autom. Control , 1 (1981) pp. 235–252 |
[8b] | J.C. Willems, "Almost invariant subspaces: an approach to high gain feedback design I. Almost conditionally invariant subspaces" IEEE Trans. Autom. Control , 5 (1982) pp. 1071–1084 |
[9] | W. Coppel, "Matrix quadratic equations" Bull. Austr. Math. Soc. , 10 (1974) pp. 377–401 |
Comments
The frequency theorem is better known as the Kalman–Yakubovich lemma or Kalman–Yacubovich lemma.
References
[a1] | R.E. Kalman, "Lyapunov functions for the problem of Lurie in automatic control" Proc. Nat. Acad. Soc. USA , 49 : 2 (1963) pp. 201–205 |
[a2] | B.D.O. Anderson, S. Vongpanitlerd, "Network analysis and synthesis: a modern systems theory approach" , Prentice-Hall (1973) |
Frequency theorem. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Frequency_theorem&oldid=18832