2.2
See RS I
Im
0.5
L(j180 )
1
GM
PM
Re
= +
0.5
L(jc )
0.5
L(j)
2-5
MS
GM
MS 1
1
1
P M 2 arcsin
[rad.]
2MS
MS
kwP Sk < 1
wP (s) =
s/M + B
,
s + B A
where,
kwT T k < 1
wT (s) =
s
BT
1
,
MT
where,
ln |S()| d =
Np
X
i=1
|S()| 1/R,
|S()| 2.1/|G
for all
R where
|Gd ()| > 1
Neglected
unmodelled
dynamics
d ()|, and
|G()| > |G
dynamics,
1, for allusually
Rat high frequencies,
ofd ()|
missing
Multiplicative uncertainty:
! wI
"
Gp
! I
!"
!+ ! G
+
Additive uncertainty:
! wA ! A
"
Gp :
(a)
wiA
!+ "
!+!
!+ "
!+ !
kA k 1.
wI |G
I G()|
!p ()
|wA ()| !max
for all R
"
Gp
!+ "
!!
for all + R
(b)
! wO ! O
wiI
!+ "
!+
for all R
iA
(d)
iI
(e)
wiO
n1 sn1 + + 0
sn + n1 sn1 + + 0
x(t) = eAt x0 +
eA(t ) B u( ) d.
0
1
G(s) = C(sI A)
n1
n2
..
.
G(s) =
B +D
..
..
.
n1
n2
..
.
Controllability matrix:
C =
Observability matrix:
O =
B
C
AB
A B
n1
B
T
(CA)
2 T
(CA )
n1 T
(CA
H -norm:
kG(s)k = sup max (G())
R
= max kGdk2
kdk2 =1
H2 -norm:
kG(s)k2 =
=
1
2
1
2
|G()|2 d
1/2
(SISO case)
trace(G() G() d
1/2
(MIMO case)
Multivariable control
Push-through rule:
G1 (I G2 G1 )1 = (I G1 G2 )14.2
G1
Condition number:
(M ) =
max (M )
min (M )
Classes of uncertainty
1. Parametric uncertainty. He
the model (including the order)
some of the parameters are unc
2. Neglected and unmodelled
6.4.8
more
general
control structure
Loopshape, L(s) is stabilizable and detectable. Robustness:
det(I a +
L(s))
= # unstable
poles source
of L(s)of uncertainty.
including
model
uncertainty
[3.8.8]
Generalized plant:
! wI
"
"
u
!
!
!
v
"
"
z = Fu (N, ) w
=
N12 w
!
!
z
!
6-42
u
"
Figure 76:IfM-structure
for robust stability
analysis
(M ()())
< 1, for
all
!1
max
!2
then closed-loop
Outputs is stable.
!1
z
!2
System
Ifwith
(M ()()) =
!
If we Actuators,
also pullSensors
out the controller K, we get3!the
4
max
|
(M
()())|
generalized
plant iP , as ishown in Figure 74. <
For1,
and Controller
>
analysis ofthen
the uncertain
system,
we use the
closed-loop
isw stable.
N
N -structure
in!Figure 75.
!
3
for all R,
z
(a)
!z
! I
(b)
:= { | = diag(1 , . . . , n ) }
Block structured :
F (s)
f (nT )
F (z)
1
s
1(nT )
1
s2
z
z1
nT
1
s+a
eanT
a
s(s + a)
1 eanT
a
+ a)
s2 (s
Tz
(z 1)2
z
z eaT
z(1 eaT )
(z 1)(z eaT )
zi = esi T
Nyquist frequency:
GZOH (z) = (1 z 1 )Z
Discrete Approximations
Bilinear
s=
2 (z 1)
T (z + 1)
Prewarped Tustin
s=
0
(z 1)
tan(0 T /2) (z + 1)
T
G(s)
s