state feedback
DESCRIPTION
State Feedback. State Feedback. State Feedback. Closed loop matrices. R=0 the system is called regulator. State Feedback. The CL state matrix is a function of K. => Faster/stable system. By appropriate changing K we can change eig( A CL ). This method is called pole placement. - PowerPoint PPT PresentationTRANSCRIPT
State FeedbackState Feedback
UG(s)
Y UG(s)
YGc(s)
R
UG(s)
Y
Gc(s)
RGc(s)
R1 UG(s)
Y
K
RK
R1
UG(s)
Y
K
RK1
R1
State FeedbackState Feedback
BU dt C
DX X Y
A
BU dt C
DX X Y
A
K
K1
R
BU dt CDX X Y
A
ControllerR
X
State FeedbackState Feedback
BU dt CDX X Y
A
ControllerR
X
)()()(
)()()( 1
ttt
ttt
BUAXX
KXRKU )()()()( 1 tttt KXRKBAXX
)()()()( 1 ttt RBKXBKAX
)()()()()()( ttttt RDKXDKCDUCXY 1
)()()(
)()()(
ttt
ttt
CLCL
CLCL
UDXCY
RBXAX
1DKD
DKCC
BKB
BKAA
CL
CL
CL
CL
1
)( Closed loop matrices
R=0 the system is called regulator.
State FeedbackState Feedback
BU dt CDX X Y
A
ControllerR
X
1DKD
DKCC
BKB
BKAA
CL
CL
CL
CL
1
)(
The CL state matrix is a function of K
By appropriate changing K we can change eig(ACL) => Faster/stable system
This method is called pole placement.
WE MUST CHECK IF THE SYSTEM IS CONTROLLABLE
State FeedbackState Feedback
10,3 xuxx
kxu
xkx 3
Eigenvalues of CL system: 3-k
CL eigenvalues at -10
k=13
State FeedbackState Feedback
)(01)(
0
1)(
43
21)(
tty
utt
X
XX If the system is unstable create a controller that will stabilise the system.
>> eig(A) >> rank(ctrb(A,B))
ans = -0.3723 5.3723
ans = 2
43
21
0
1
43
21 2121
kkkkACL
43
21 21 kkeig 0
43
21 21
s
kks
0143
2110
4103
2110 2121 kkkk
036141540231114 2121 kkkk
0153
2120
4113
2111 2121 kkkk
036151800231215 2121 kkkk
26174315
1483141
21
21
kkk
kk
722 k
State FeedbackState Feedback
)(01)(
0
1)(
43
21)(
tty
utt
X
XX -10 and -11 261 k
722 k 210
1
43
21kkACL
Matlab
If the system is n x n then you need to solve an n x n system of equations!
K=place(A,B, [-10 -11])
010
001
125022520
A T001B
C=[0 0 1],
D=0.
P=[-10 -20 -30];
Matlab
State Feedback - LQRState Feedback - LQR
)(10
01)(
)(43
21)(
tty
tt
X
XXK =[26 72]T
P=[-10 -11]
Matlab: State feedbackResponse of the systemReference signalStudy the signal u=-KxPlace the poles at [-100 -110] Response of the systemStudy the signal u=-Kx
State Feedback - LQRState Feedback - LQR
Compromise between speed and energy that we use. Similar problem/dilemma if we had an input.
Solution: Linear Quadratic Regulator (optimum controller)
BUAXX
KXU
0
dtJ TT RUUQXX
Q and R are positive definite matrices
Square symmetric matrices, positive eigenvalues 0QXXT for all nonzero X
Q: Importance of the error, R: Importance of the energy that we use
XBKAX
(assume that BKA is stable)
State Feedback - LQRState Feedback - LQR
BUAXX
KXU
0
dtJ TT RUUQXX XBKAX
(assume that BKA is stable)
0
dtJ TT KXRKXQXX
0
dtJ T XRKKQX T
PXXXRKKQX TT
dt
dT P is positive definite
XBKAX
TTT
BKAXX
X=1
BKAPPBKARKKQ T T
XPXPXX
PXX TT
T
dt
d
dt
d
dt
d
PBRK T1
01 QPBPBRPAPA TTReduced Riccati Equation
State Feedback - LQRState Feedback - LQR
Steps to design an LQR controller:
1.To find the optimum P solve: 01 QPBPBRPAPA TT
2.Find the optimum gain PBRK T1
[K, P, E]=lqr(sys, Q, R) E are the eigenvalues of A-BK
1
0,
10
10BA
Find K,
The eigenvalues of A-BK
The response of the system
For R=1 and Q=eye(2) and Q=2*eye(2) (X(0)=[1 1]).
Matlab
CAD Exercise
Estimating techniquesEstimating techniques
Magic trick
BU dt CDX X Y
A
ControllerR
X
BU dt CDX X Y
A
Estimating techniquesEstimating techniques
A
U BU +
+dt
X X
AX
CB
Plant
Y
A
U BU +
+dt
X X
AX
CB
Plant
Y
A
BU +
+dt CB
Estimator
X~ X
~
XA~
Y~
XCY
BUXAX~~
~~
Example: A=[1 2;3 4]; B=[1 0]'; C=[1 0]; D=0;
Estimating techniquesEstimating techniques
The error between the estimated and real state is
ttt XXE~
ttt XXE
~ ttttt BUXABUAX~
tt AEE
Homogeneous ODE
00 E
A is unstable
tE
A is slow
A=[1 2;3 4]; B=[1 0]'; C=[1 0]; D=0; U=1, 9.09.00~,110 XX
Estimating techniquesEstimating techniques
A
U BU +
+dt
X X
AX
CB
Plant
Y
A
BU +
+dt CB
Estimator
X~ X
~
XA~
Y~
A
U BU +
+dt
X X
AX
CB
Plant
Y
A
BU +
+dt
X~
X~
XA~
CB Y~
Estimator
-
+-
YY~
Y~
A
U BU +
+dt
X X
AX
CB
Plant
Y
A
BU +
+dt
X~
X~
XA~
CB Y~
Estimator
-
+-
G
YY~
Estimating techniquesEstimating techniques
A
U BU +
+dt
X X
AX
CY
B
F
BU
+
+dt
X~
X~
XF~
CY~
B
GYG+
Estimator
PlantY
GYBUXFX
ttt~~
tEGCA
tXBKAX
G=place(A’, C’, P)
But the system must be observable
A=[1 2;3 4]; B=[1 0]'; C=[1 0]; D=0; U=1,
9.09.00~,110 XX
Where to place these eigenvalues???Where to place these eigenvalues???
YGBUXAX
tttt~~
YYGBUXAX~~~
tttt
XCYGBUXAX~~~
tttt
GYBUXGCAX
tttt~~
ttt XGCAXGCAE~
GCXBUXGCAX
tttt~~
ttt BUXAX