P.L.D. Peres J.C. Geromel - H2 Control For Discrete-Time Systems Optimality and Robustness
P.L.D. Peres J.C. Geromel - H2 Control For Discrete-Time Systems Optimality and Robustness
225-228, 1993
0005-1098/93$6.00+ 0.00
~) 1992PergamonPress Ltd
Printedin GreatBritain.
Brief Paper
2. Preliminaries
(1)
z k = Cxt + DUk,
1. Introduction
(2)
(3)
(4)
A'ILoA d - L o + C'lCct = 0.
(5)
(6)
(P1)
(7)
225
(8)
226
Brief Paper
ActLcA'ct- L,: = - B t B ~.
(9)
A'PA-P-A'PB2(B~PBz+D'D)-~B~PA+C'C=O.
(P2)
(10)
J* = x;,Pxo.
(20)
x 0 given.
xk+ 1 = A x k + B2Uk,
(19)
A d W A " I - W + B I B ~~ O.
W
=[KW
WK'
K W K ' ] >'0'
(21)
(12)
and we conclude from (5) that P = L o and IIHIbZ=J *
provided we choose B ~ = x o. Note that the necessary
optimality conditions for (P1) are given by equations (4)-(5)
and
[(B~LoBz + D ' D ) K - B~LoA]L~ = 0,
3. Main results
(22)
+ x'[Bz(W3 - W~W;'Wz)B~]x.
Vv .~P q=O:G'v = 0
(23)
~V'.>-O~ W~ >" W ~ W ~ l W z
o:[0,]
.4)
o0o]
.,,
(16)
w~
~'=[wW'~ W3],
(24)
The
last
(17)
- A W z B ~ + B2W~B ~ + BIB;]x
(13)
Q]v
(a3)
(25)
Non convex
1, K* = W ~ W { l ~ ,
w=w/( I
o')= t KW KWX" ]
r
being so a
Convex
Brief Paper
227
(26)
DK)W~(C - DK)'}
( C W ~ C ' + DW:~W(IW2 D'}
(27)
(P4)
L c K ' ] <g2,
KLeK' J
(29)
(32)
I
--1
B2iW~W
I )WI(A
I
--1 t
B2iW~W,
)
W,
B,B~ <-0,
Vi=l.-.M.
(33)
#:->llHll2z, V B 2 ~ z
i = 1 . . . M,
and
(34)
the proof is
[]
(35)
(30)
Vi = 1 . - . M,
(28)
where K = ( B ~ L o B 2 + D ' D ) - t B ~ L o A ,
being P = L o the
definite positive solution of the discrete-time Riccati
equation (9) and Lc being the solution of the controllability
Gramian (eguation (4)). Now, it is simple to see that
K* = W ~ W ~ ~ provides the optimal gain and
(31)
(A
OW*= [ Lc
KL~
(36)
(37)
:go
given,
C=
1 ,
B2=
D=
[0.6135
0.2749
0.8807
0.6538]
0.4899[,
0.7741 ]
0,
0
0
and B I = I . Since the eigenvalues of matrix A are
(0.3827, -0.4919, 1.6133), it is open-loop unstable. Using the
discrete-time Riccati equation, we get
KRiccati=[0.2968 0.3758
L0.2302 0.4953
0.3114]
0.4997J'
228
Brief Paper
LQR w i t h folLure a t k : 6
2I
1.4582
-0.3930
-0.0926
0.2597
0.0918q
1-0.3930
1.6316
-0.1254
0.4629
0.6612 /
o.e
~t~'*= / - 0 . 0 9 2 6
-0.1254
1.1972
0.3013
0.51511,
0.6
1.0
0.2597
0.4629
0.3013
0.3489
0.4411 /
0.4
0.0918
0.6612
0.5151
0.4411
0.60841
0.2
0.2749 0.48991,
0.8807 0.7741 /
B23 =
B22 =
0.4899 | .
0.7741_1
Kr=L0.4823
s'J"
-o.s
IO
20
30
40
50
60
/r
"!
o.6
0.4 ii
0 :
"~
:'
-0.2
i(~
-0,4
.'"'""
"
I
/"
-0,4
L !1
0.2749
0.8807
-0.2
0.3803 0.3148]
0.4989 0.5002 J'
.,'
..'
]i?.
:'...~,~.
"'.
.'
.-
-0.6
-0.8
-i.o
0.6356 0.7405]
0.6592 0.4925J'
IO
20
30
40
50
60
5. Conclusion
In this paper we have proved that the optimal state
feedback solution of the discrete-time LQP can be
determined by means of a convex problem. This is an
important result, mainly due to two facts. First, for a given
TABLE 1. CLOSED-LOOP EIGENVALUES
KRiccat i
gf
Both actuators
-0.4826
0.2085
0.4385
0.4153
-0.5326
-0.4069
-0.4861
0.3347
1.0959
0.4047
0.5973
-0.5068
Contingencies
References
Anderson, B. D. O. and J. B. Moore (1971). Linear Optimal
Control. Prentice Hall, Englewood Cliffs, NJ.
Bernussou, J., P. L. D. Peres and J. C. Geromel (1989). A
linear programming oriented procedure for quadratic
stabilization of uncertain systems. Systems and Control
Letters, 13, 65-72.
Dorato, P. and A. H. Levis (1971). Optimal linear
regulators: the discrete time case. IEEE Trans. on Aut.
Control, 16, 613-620.
Geromel, J. C., P. L. D. Peres and J. Bernussou (1991). On
a convex parameter space method for linear control design
of uncertain systems. SIAM J. on Control and Optimiz.,
29, 381-402.
Kwakernaak, H. and R. Sivan (1972). Linear Optimal
Control Systems. John Wiley, New York.
Lucnberger, D. G. (1973). Introduction to Linear Programruing. Addison-Wesley, Reading, MA.