Upload
anonymous-n3lpax
View
9
Download
0
Tags:
Embed Size (px)
DESCRIPTION
dfgru
Citation preview
Convex Optimization and SystemTheory
Lecturer : Prof.dr. Anton A. Stoorvogel
E-mail: [email protected]
1/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a system:
x = f(x,w)z = g(x,w)
(∗)
The system with supply function s is said to be dissipative if
there exists a function V such that:
V(x(t0))+t1∫
t0
s(w(t), z(t))dt � V(x(t1))
for all t0 � t1 and all (w,x, z) satisfying (∗).
2/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a system:
x = f(x,w)z = g(x,w)
(∗)
The system with supply function s is said to be conservative if
there exists a function V such that:
V(x(t0))+t1∫
t0
s(w(t), z(t))dt = V(x(t1))
for all t0 � t1 and all (w,x, z) satisfying (∗).
3/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
First law of thermodynamics
E(x(t0))+t1∫
t0
Q(t)+W(t)dt = E(x(t1))
E Internal energy
S Entropy
Q Rate of heating
T Temperature
W Rate of work
4/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Second law of thermodynamics
S(x(t0))+t1∫
t0
−Q(t)T(t)
dt = S(x(t1))
E Internal energy
S Entropy
Q Rate of heating
T Temperature
W Rate of work
5/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a system:
x = f(x,w)z = g(x,w)
(∗)
The system with supply function s is said to be strictly
dissipative if there exists a function V and an ε > 0 such that:
V(x(t0))+t1∫
t0
s(w(t), z(t))dt − ε2
t1∫
t0
‖w(t)‖2 dt � V(x(t1))
for all t0 � t1 and all (w,x, z) satisfying (∗).
6/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
x∗ ∈ X is a fixed reference point in the state space. We consider
normalized storage functions which satisfy
V(x(t0))+t1∫
t0
s(w(t), z(t))dt � V(x(t1))
and additionally
V(x∗) = 0
We assume that we can reach all points from x∗ via an
appropriate choice for w and conversely.
7/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Available storage
Vav(x0) := supw,t1
−t1∫
0
s(w(t), z(t))dt
such that (w,x, z) satisfy (∗) while t1 � 0, x(0) = x0 and
x(t1) = x∗.
8/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Required storage
Vreq(x0) := infw,t−1
0∫
t−1
s(w(t), z(t))dt
such that (w,x, z) satisfy (∗) while t1 � 0, x(t−1) = x∗ and
x(0) = x0.
9/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
The system is dissipative if and only if the available storage is
well-defined and finite for all initial conditions
The system is dissipative if and only if the required storage is
well-defined and finite for all end conditions
10/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Linear systems with quadratic supply rate
Consider a system:
x = Ax + Ewz = Cx +Dw
(∗)
and
s(w, z) =⎛⎜⎝wz
⎞⎟⎠
T⎛⎜⎝Q S
ST R
⎞⎟⎠⎛⎜⎝wz
⎞⎟⎠
11/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
A dissipative system with a quadratic supply rate has a
quadratic storage function.
V(x) = xTPx
Moreover available and required storage are also both quadratic.
12/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a linear system with a quadratic supply rate
We define V by:
V(x) = xTPx
The system is dissipative if and only if there exists a symmetric
matrix P such that:
⎛⎜⎝ A
TP + PA+ CTRC PE +DTRC + SCETP + CTST + CTRD Q+ SD +DTST +DTRD
⎞⎟⎠ � 0
In that case V is a storage function.
13/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a linear system with a quadratic supply rate
We define V by:
V(x) = xTPx
The system is strictly dissipative if and only if there exists a
symmetric matrix P such that:
⎛⎜⎝ A
TP + PA+ CTRC PE +DTRC + SCETP + CTST + CTRD Q+ SD +DTST +DTRD
⎞⎟⎠ ≺ 0
In that case V is a storage function.
14/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Positive real lemma
A system is dissipative with respect to the quadratic supply rate
s(w,z) = zTw if and only if there exists a symmetric matrix Psuch that: ⎛
⎜⎝ATP + PA PE − CT
ETP − CT −D −DT
⎞⎟⎠ � 0
15/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Bounded real lemma
A system is dissipative with respect to the quadratic supply rate
s(w,z) = wTw − zTz if and only if there exists a symmetric
matrix P such that:
⎛⎜⎝A
TP + PA+ CTC PE + CTD
ETP +DTC DTD − I
⎞⎟⎠ � 0
16/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Stability
Consider a nonlinear system:
x = f(x)
x0 is called an equilibrium point if f(x0) = 0
17/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a nonlinear system:
x = f(x), x(0) = x0
The equilibrium point x∗ is called stable if for all ε > 0 there
exists δ > 0 such that
‖x∗ − x0‖ < δ �⇒ ‖x(t)− x∗‖ � ε for all t � 0
18/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a nonlinear system:
x = f(x), x(0) = x0
The equilibrium point x∗ is called asymptotically stable if
• the equilibrium point is stable,
• there exists δ > 0 such that
‖x∗ − x0‖ < δ �⇒ limt→∞
x(t) = 0
19/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Lyapunov stability
Consider a nonlinear system:
x = f(x), x(0) = x0
If there exists a positive definite, continuously differentiable
function V with V(x∗) = 0 and
ddtV(x(t)) � 0
for all initial conditions x0 and t > 0 then the equilibrium x∗ is
stable.
20/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Lyapunov stability
Consider a nonlinear system:
x = f(x), x(0) = x0
If there exists a positive definite, continuously differentiable
function V with V(x∗) = 0 and
ddtV(x(t)) < 0
for all initial conditions x0 and t > 0 then the equilibrium x∗ is
asymptotically stable.
21/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a linear system
x = Ax
If there exists a matrix P 0 such that
ATP + PA � 0
then the equilibrium 0 is stable.
22/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
Consider a linear system
x = Ax
If there exists a matrix P 0 such that
ATP + PA ≺ 0
then the equilibrium 0 is asymptotically stable.
23/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
H∞ performance
Consider a system:
x = Ax + Ew, x(0) = 0
z = Cx +Dw(∗)
supw∈L2w≠0
‖z‖‖w‖
24/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
The H∞ performance is less than γ if and only if there exists P∞such that
⎛⎜⎝A
TP∞ + P∞A+ CTC P∞E + CTD
ETP∞ +DTC DTD − γ2I
⎞⎟⎠ � 0
25/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
H2 performance
Consider a system:
x = Ax + Ew, x(0) = 0
z = Cx +Dw(∗)
p∑i=1
‖zi‖2
where zi is the output of the system with input wi = δ(t)ej .Here w(t) ∈ �p and e1, . . . ep is a basis of �p while δ(t) is the
Dirac impulse.
26/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS
The H2 performance is equal to
traceEP2ET
where
ATP2 + P2A+ CTC = 0
27/27 Electrical Engineering, Mathematics and Computing ScienceEEMCS