On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic...
Transcript of On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic...
![Page 1: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/1.jpg)
On algebraic simplifications of linear functional systems
Thomas Cluzeau
XLIM (UMR CNRS 6172), University of Limoges (France)
DMI, CANSO Project, ENSIL
joint work with Alban Quadrat (INRIA APICS Project)
ACA 2008, RISC-Linz, Castle of Hagenberg, Austria, July 27-30, 2008
AADIOS Session
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 1 / 31
![Page 2: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/2.jpg)
Computer algebra & Applications
• Objective of this work:
Use computer algebra (algebraic manipulations) to simplify systemscoming from mathematical physics, applied mathematics, engineering
sciences or control theory
• Interest:
Simplify the equations of the system
⇒ simplify the study of its structural properties
Pre-conditioner to numerical analysis methods
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 2 / 31
![Page 3: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/3.jpg)
Outline of the talk
1 Existing works
2 Goal of the talk
3 Methodology and statement of the problem
4 Result
5 Examples
6 Implementation / Conclusion / Perspectives
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 3 / 31
![Page 4: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/4.jpg)
I
Existing works
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 4 / 31
![Page 5: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/5.jpg)
Bloc decomposition of a linear differential system
• Consider an ordinary differential system of the first order:
∂ y = E (t) y , E (t) ∈ Q(t)n×n
• Does-it exist an invertible change of variables
y = P(t) z ,
such that∂ y = E (t) y ⇔ ∂ z = F (t) z ,
where F = P−1 (E P − ∂P) has the following form:
F =
(? 00 ?
)?
• Many algorithms exist: Beke, Jacobson, Schwarz, Singer, Bronstein,Tsarev, van Hoeij, Barkatou, Pflugel, Giesbrecht, ...Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 5 / 31
![Page 6: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/6.jpg)
The eigenring method (Jacobson, Singer, Barkatou,...)
• Consider the following system :
∂ y = E (t) y , E (t) =
t (2 t + 1) −2 t3 − 2 t2 + 1
2 t −t (2 t + 1)
• Eigenring of the system: E = {P ∈ Q(t)2×2 ; ∂P = E P − P E}
• We compute E = {P =
a1 − a2 (t + 1) a2 t (t + 1)
−a2 a2 t + a1
; a1, a2 ∈ Q}
• Jordan form of P: J =V−1 P V=
a1 0
0 a1 − a2
• Let y =V z : ∂ y = E (t) y ⇔ ∂ z =
−t 0
0 t
z
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 6 / 31
![Page 7: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/7.jpg)
Smith canonical form
• D = k[s], k field (e.g., Q, R, C): euclidian ring.
• Theorem. ∀ R ∈ Dq×p, ∃ V ∈ GLq(D), U ∈ GLp(D):
R = V R U =
α1 0 . . . . . . 0 . . . 0
0 α2. . .
......
.... . .
. . .. . .
......
0 . . . 0 αr 0 . . . 00 . . . . . . . . . 0 . . . 0...
.... . .
...0 . . . . . . . . . 0 . . . 0
,
ou α1 |α2 | . . . |αr 6= 0 et αi ∈ D, i = 1, . . . , r .
• Applications in control theory: polynomial approach (Rosenbrock,Kailath,...)• Generalization: Jacobson form: D = K
[ddt
], K differential field
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 7 / 31
![Page 8: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/8.jpg)
II
Goal of the talk
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 8 / 31
![Page 9: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/9.jpg)
Generalization of the previous methods?
• Observation: many systems coming from mathematical physics, appliedmathematics, engineering sciences or control theory can not be handled bythe previous methods: partial differential equations, differential time-delayequations, . . .
• Question: can we generalize these methods to handle more linearfunctional systems?
• Approach: constructive homological algebra
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 9 / 31
![Page 10: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/10.jpg)
Example 1
• Model of a one-dimensional tank containing a fluid subjected to anhorizontal move (Petit-Rouchon, IEEE TAC 02):{
y1(t)− y2(t − 2 h) + α y3(t − h) = 0,
y1(t − 2 h)− y2(t) + α y3(t − h) = 0,α ∈ R, h ∈ R+.
• Consider D = R[
ddt , δ
], δ(y(t)) = y(t − h), and the matrix
R =
d
dt− d
dtδ2 α
d2
dt2δ
d
dtδ2 − d
dtα
d2
dt2δ
∈ D2×3.
• Question: ∃ U ∈ GL3(D), V ∈ GL2(D) such that:
V R U =
(α1 0 0
0 α2 α3
), α1, α2, α3 ∈ D ?
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 10 / 31
![Page 11: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/11.jpg)
Example 2
• Consider the 4 matrices:
γ1 =
0 0 0 −i0 0 −i 00 i 0 0i 0 0 0
, γ2 =
0 0 0 −10 0 1 00 1 0 0−1 0 0 0
γ3 =
0 0 −i 00 0 0 ii 0 0 00 −i 0 0
, γ4 =
1 0 0 00 1 0 00 0 −1 00 0 0 −1
• In electromagnetism, the Dirac equations can be written :
4∑i=1
γi ∂ ψ(x)
∂ xi= 0,
where ψ = (ψ1, ψ2, ψ3, ψ4)T and x = (x1, x2, x3, x4) contains the space
coordinates and the time.Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 11 / 31
![Page 12: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/12.jpg)
Example 2
• We can then write the Dirac equations as follows:d4 ψ1 − i d3 ψ3 − (i d1 + d2)ψ4 = 0,
d4 ψ2 − (i d1 − d2)ψ3 + i d3 ψ4 = 0,
i d3 ψ1 + (i d1 + d2)ψ2 − d4 ψ3 = 0,
(i d1 − d2)ψ1 − i d3 ψ2 − d4 ψ4 = 0,
di = ∂/∂xi .
• Consider D = Q(i)[d1, d2, d3, d4] and the matrix
R =
d4 0 −i d3 −(i d1 + d2)0 d4 −i d1 + d2 i d3
i d3 i d1 + d2 −d4 0i d1 − d2 −i d3 0 −d4
.
• Question: ∃ U ∈ GL4(D), V ∈ GL4(D) such that:
V R U =
? ? 0 0? ? 0 00 0 ? ?0 0 ? ?
?
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 12 / 31
![Page 13: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/13.jpg)
III
Methodology and statement of theproblem
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 13 / 31
![Page 14: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/14.jpg)
Methodology
1 A linear system is defined by a matrix R with coefficients in a ring Dof functional operators:
R y = 0. (?)
2 To (?) we associate a left D-module M (finitely presented).
3 There exists a dictionary between the properties of (?) and M.
4 Homological algebra allows to check the properties of M.
5 Effective algebra (non-commutative Grobner/Janet bases) givesalgorithms.
6 Implementation (Maple, Singular/Plural, Cocoa. . . ).
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 14 / 31
![Page 15: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/15.jpg)
D: Ore algebra of functional operators
• Differential operators: A = Q, Q[x1, . . . , xn], Q(x1, . . . , xn),
D = A[∂1, . . . , ∂n], ∂i = ∂∂xi,
P =∑
0≤|µ|≤m aµ(x) ∂µ ∈ D, ∂µ = ∂µ11 . . . ∂µn
n , aµ ∈ A.
• Shift operators:
D = A[σ], A = Q, Q[n], Q(n),
P =∑m
i=0 ai (n)σi ∈ D, σ(a(n)) = a(n + 1).
• Differential time-delay operators:
D = A[ ddt , δ], A = Q, Q[t], Q(t),
P =∑
0≤i+j≤m aij(t)d i
dt iδj ∈ D, δ(a(t)) = a(t − h).
• Theorem. For every monomial order, there exists a Grobner basis whichcan be computed by Buchberger algorithm.Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 15 / 31
![Page 16: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/16.jpg)
The left D-module M
• Let D be an Ore algebra, R ∈ Dq×p and a left D-module F .
• Consider kerF (R.) = {η ∈ Fp | R η = 0}.
• As in number theory or algebraic geometry, to kerF (R.) we associate thefinitely presented left D-module:
M = D1×p/(D1×q R).
Theorem (Malgrange)
kerF (R.) ∼= homD(M,F) = {f : M → F , f is left D-linear}.
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 16 / 31
![Page 17: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/17.jpg)
Statement of the problem
• Let D be a Ore algebra of functional operators
• Let R ∈ Dq×p be a matrix.
• Question:
∃ W ∈ GLp(D), V ∈ GLq(D) s.t. V R W =
(? 0
0 ?
)?
• Remark: with M = D1×p/(D1×q R), this is equivalent to
∃ M1, M2 : M = M1 ⊕M2?
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 17 / 31
![Page 18: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/18.jpg)
IV
Result
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 18 / 31
![Page 19: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/19.jpg)
Endomorphisms of M (Cf. eigenring method)
• Let D be a Ore algebra of functional operators.
• Let R ∈ Dq×p be a matrix.
• We have the following exact commutative diagram:
D1×q .R−→ D1×p π−→ M −→ 0↓ .Q ↓ .P ↓ f
D1×q .R−→ D1×p π−→ M −→ 0.
Theorem
f ∈ endD(M) is defined by P ∈ Dp×p et Q ∈ Dq×q satisfying the relation:R P = Q R.
• Algorithms for computing P and Q implemented in the Maple packageOreMorphisms based on the library OreModulesThomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 19 / 31
![Page 20: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/20.jpg)
Bloc decomposition theorem
Theorem
Let R ∈ Dq×p, M = D1×p/(D1×q R) and f ∈ endD(M) defined by P andQ satisfying
P2 = P, Q2 = Q (idempotent matrices) ⇒ f 2 = f .
If the left D-modules
kerD(.P), imD(.P), kerD(.Q), imD(.Q)
are free, then there exist U ∈ GLp(D), V ∈ GLq(D) such that
R = V R U−1 =
(? 00 ?
)∈ Dq×p.
• U and V can be obtained by computing bases of free left D-modules
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 20 / 31
![Page 21: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/21.jpg)
V
Examples
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 21 / 31
![Page 22: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/22.jpg)
Example 1: tank model (Petit-Rouchon, IEEE TAC 02)
• Consider D = Q(α)[
ddt , δ
], the matrix of the system
R =
(ddt − d
dt δ2 α d2
dt2 δ
ddt δ
2 − ddt α d2
dt2 δ
)∈ D2×3.
• The matrices P = 12
1 −1 0−1 1 00 0 2
et Q = 12
(1 11 1
)satisfy:
R P = Q R, P2 = P, Q2 = Q.
• Using linear algebra, we get:
U =
1 1 0
1 −1 0
0 0 1
∈ GL3(D), V =
(1 −1
1 1
)∈ GL2(D),
⇒ R = V R U−1 =
(ddt (1− δ) (1 + δ) 0 0
0 ddt (δ2 + 1) 2α d2
dt2 δ
).
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 22 / 31
![Page 23: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/23.jpg)
Example 2: Dirac equations
R =
d4 0 −i d3 −(i d1 + d2)0 d4 −i d1 + d2 i d3
i d3 i d1 + d2 −d4 0i d1 − d2 −i d3 0 −d4
.
P =1
2
1 0 −1 00 1 0 −1−1 0 1 00 −1 0 1
, Q =1
2
1 0 1 00 1 0 11 0 1 00 1 0 1
,
⇒ U =
1 0 1 00 −1 0 −1−1 0 1 00 1 0 −1
, V =
1 0 −1 00 1 0 −11 0 1 00 −1 0 −1
,
⇒ V R U−1 =
i d3 − d4 −i d1 − d2 0 0i d1 − d2 i d3 + d4 0 0
0 0 i d3 + d4 i d1 + d2
0 0 i d1 − d2 −i d3 + d4
.
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 23 / 31
![Page 24: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/24.jpg)
Example 3: string model(Mounier-Rudolph-Fliess-Rouchon,COCV 98)
• Consider the model of a vibrating string with interior mass:φ1(t) + ψ1(t)− φ2(t)− ψ2(t) = 0,
φ1(t) + ψ1(t) + η1 φ1(t)− η1 ψ1(t)− η2 φ2(t) + η2 ψ2(t) = 0,
φ1(t − 2 h1) + ψ1(t)− u(t − h1) = 0,
φ2(t) + ψ2(t − 2 h2)− v(t − h2) = 0,
where h1 and h2 ∈ R+ satisfy dimQ(Q h1 + Q h2) = 2.
• Consider D = Q(η1, η2)[
ddt , σ1, σ2
], M = D1×6/(D1×4 R),
R =
1 1 −1 −1 0 0
d
dt+ η1
d
dt− η1 −η2 η2 0 0
σ21 1 0 0 −σ1 0
0 0 1 σ22 0 −σ2
∈ D4×6.
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 24 / 31
![Page 25: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/25.jpg)
Example 3: string model
• The following matrices satisfy R P = Q R, P2 = P and Q2 = Q :
P =
1 0 0 0 0 0−σ2
1 0 0 0 σ1 00 0 0 −σ2
2 0 σ2
0 0 0 1 0 00 0 0 0 1 00 0 0 0 0 1
,Q =
1 0 −1 1
0 1 − d
dt+ η1 η2
0 0 0 0
0 0 0 0
.
• The modules kerD(.P), imD(.P), kerD(.P), imD(.P) are free:
U =
σ2
1 1 0 0 −σ1 00 0 1 σ2
2 0 −σ2
1 0 0 0 0 00 0 0 1 0 00 0 0 0 1 00 0 0 0 0 1
,V =
0 0 1 00 0 0 11 0 −1 1
0 −1d
dt− η1 −η2
.
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 25 / 31
![Page 26: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/26.jpg)
Example 3: string model
• R is then equivalent to the bloc-diagonal matrix:
R = V R U−1 =
0BBBBBB@
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1− σ21 σ2
2 − 1 σ1 −σ2
0 0 σ21
„d
dt− η1
«−„
d
dt+ η1
«−η2 (σ2
2 + 1) −σ1
„d
dt+ η1
«η2 σ2
1CCCCCCA .
• Consider the second diagonal bloc
S =
0B@ 1− σ21 σ2
2 − 1 σ1 −σ2
σ21
„d
dt− η1
«−„
d
dt+ η1
«−η2 (σ2
2 + 1) −σ1
„d
dt+ η1
«η2 σ2
1CA ,
and the D-module N = D1×4/(D1×2 S).
• An idempotent g ∈ endD(N) is defined by the matrices
P′ =
0BB@1 0 0 0a 0 b 00 0 1 00 0 0 1
1CCA , Q′ =1
2
σ2
2 + 1 (σ22 − 1)/η2
−η2 (σ22 + 1) −σ2
2 + 1
!,
{a = (σ2
1
(ddt − (η1 + η2)
)− d
dt + (η2 − η1))/(2 η2),
b = −σ1 ( ddt − (η1 + η2))/(2 η2).
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 26 / 31
![Page 27: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/27.jpg)
Example 3: string model
• The modules kerD(.P), im (.P), kerD(.Q) et im (.Q) are free.
U′ =
0BBBBB@σ2
1
“ddt− η1 − η2
”−“
ddt
+ η1 − η2
”−2 η2 −σ1
“ddt− η1 − η2
”0
1 0 0 0
−σ1 0 1 0
σ21 σ2 (d − η1 − η2)− σ2 (d + η1 − η2) 0 −σ1 σ2 (d − η1 − η2) −2 η2
1CCCCCA ,
V ′ =
(η2 1
η2 (σ22 + 1) σ2
2 − 1
).
⇒ S = V ′ S U ′−1 =
1 0 0 0
0d
dt+ η1 + η2 σ1
(d
dt+ η2 − η1
)σ2
.
• Let U ′′ = diag(I2,U′), V ′′ = diag(I2,V
′). We then have:
R = (V ′′ V ) R (U ′′ U)−1 = diag(I2,S).
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 27 / 31
![Page 28: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/28.jpg)
Example 3: string model
• Note α = η1 + η2 et β = η2 − η1. We have obtained:φ1(t) + ψ1(t)− φ2(t)− ψ2(t) = 0,
φ1(t) + ψ1(t) + η1 φ1(t)− η1 ψ1(t)− η2 φ2(t) + η2 ψ2(t) = 0,
φ1(t − 2 h1) + ψ1(t)− u(t − h1) = 0,
φ2(t) + ψ2(t − 2 h2)− v(t − h2) = 0,
⇔
z1(t) + α z1(t) + z2(t − h1) + β z2(t − h1) + z3(t − h2) = 0.
⇒ We can then easily compute a parametrization of the string.
⇒ The system is σ2-free and σ1-free. . .
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 28 / 31
![Page 29: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/29.jpg)
VI
Implementations / Conclusions /Perspectives
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 29 / 31
![Page 30: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/30.jpg)
The OreMorphisms package
• Algorithms are implemented in a Maple package called OreMorphismsbased on the library OreModules developed by Q. et Robertz:
http://wwwb.math.rwth-aachen.de/OreModules
• List of functions:
Morphisms, MorphismsConstCoeff, MorphismsRat,
Idempotents, IdempotentsConstCoeff, IdempotentsRat
IdempotentsMat, IdempotentsMatConstCoeffs, IdempotentsMatRat
KerMorphism(Rat), ImMorphism(Rat), CokerMorphism(Rat),CoimMorphism(Rat),
TestSurj(Rat), TestInj(Rat), TestIso(Rat).
• It is freely available with a library of examples at:
http://www.ensil.unilim.fr/˜cluzeau/OreMorphisms
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 30 / 31
![Page 31: On algebraic simplifications of linear functional systems · 2009. 8. 10. · On algebraic simplifications of linear functional systems Thomas Cluzeau XLIM (UMR CNRS 6172), University](https://reader035.fdocuments.us/reader035/viewer/2022071420/6119946d6317663ffb34fbc9/html5/thumbnails/31.jpg)
Conclusion / Perspectives
• We have used algebraic manipulations and computer algebra to simplifysystems coming from mathematical physics, applied mathematics,engineering sciences or control theory.
⇒ Algorithms & Implementation & Open Questions.
• Future works:
Use these techniques in the study of generalized Smith forms
Study the links with the quadratic conservation laws studied inengineering sciences, the integrability of Hamiltonian systems. . .
Study the algebraic structure of endD(M) by means ofnon-commutatives Grobner bases (regular elements, idempotents,. . . )
Thomas Cluzeau (XLIM-DMI-ENSIL) ACA 2008 - AADIOS 31 / 31