Académique Documents
Professionnel Documents
Culture Documents
x
1
= x
2
+ l
1
(x
1
x
1
)
.
.
.
x
n1
= x
n
+ l
n1
(x
1
x
1
)
x
n
= x
n+1
+ l
n
(x
1
x
1
) + bu
x
n+1
= l
n+1
(x
1
x
1
) + h( x, w)
(3)
where x = [ x
1
, x
2
, , x
n+1
]
T
R
n+1
, and l
i
, i =
1, 2, , n+1, are the observer gain parameters to be chosen.
In particular, let us consider a special case where the gains
are chosen as
[l
1
, l
2
, , l
n+1
] =
_
1
,
2
o
2
, ,
n+1
o
n+1
(4)
with
o
> 0. Here
i
, i = 1, 2, , n + 1, are selected
such that the characteristic polynomial s
n+1
+
1
s
n
+ +
n
s +
n+1
is Hurwitz. For simplicity,let s
n+1
+
1
s
n
+
+
n
s +
n+1
= (s + 1)
n+1
where
i
=
(n+1)!
i!(n+1i)!
, i =
1, 2, , n + 1. Then the characteristic polynomial of (3)
is
o
(s) = (s +
o
)
n
. (5)
and
o
, the observer bandwidth, becomes the only tuning
parameter of the observer.
Let x
i
= x
i
x
i
, i = 1, 2, , n + 1. From (2) and (3),
the observer estimation error can be shown as
x
1
= x
2
o
1
x
1
.
.
.
x
n1
= x
n
n1
o
n1
x
1
x
n
= x
n+1
n
o
n
x
1
x
n+1
= h(x, w) h( x, w)
n+1
o
n+1
x
1
.
(6)
Now let
i
=
xi
i1
o
, i = 1, 2, , n + 1, then (6) can be
rewritten as
=
o
A + B
h(x, w) h( x, w)
n
o
(7)
where A =
_
1
1 0 0
2
0 1 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
n
0 0 1
n+1
0 0 0
_
_
, B =
[0 0 0 1]
T
. Here A is Hurwitz for the
i
, i =
1, 2, , n + 1, chosen above.
Theorem 1: Assuming h(x, w) is globally Lipschitz with
respect to x, there exists a constant
o
> 0, such that
lim
t
x
i
(t) = 0, i = 1, 2, , n + 1.
Proof. Since A is Hurwitz, there exists a unique positive
denite matrix P such that A
T
P + PA = I. Choose the
Lyapunov function as V () =
T
P. Hence
V () =
o
2
+ 2
T
PB
h(x, w) h( x, w)
n
o
. (8)
46th IEEE CDC, New Orleans, USA, Dec. 12-14, 2007 ThB07.6
3502
Since the function h(x, w) is globally Lipschitz with re-
spect to x, that is, there exists a constant c
such that
|h(x, w) h( x, w)| c
n
o
2
T
PBc
x x
n
o
. (9)
When
o
1, one has
x x
n
o
=
x
n
o
=
2
1
+
2
2
2
o
+
2
3
4
o
++
2
n+1
2n
o
n
o
. Therefore, we obtain
2
T
PB
|h(x, w) h( x, w)|
n
o
c
2
(10)
where c = 1 + PBc
2
. From (8) and (10), one has
V () (
o
c)
2
. (11)
That is,
V () < 0 if
o
> c. Therefore, lim
t
x
i
(t) = 0, i =
1, 2, , n + 1, for
o
> c. Q.E.D.
B. Convergence of the LESO with Plant Dynamics Largely
Unknown
In many real world scenarios, the plant dynamics repre-
sented by f is mostly unknown. In this case, the LESO in
(3) now takes the form of
x
1
= x
2
+ l
1
(x
1
x
1
)
.
.
.
x
n1
= x
n
+ l
n1
(x
1
x
1
)
x
n
= x
n+1
+ l
n
(x
1
x
1
) + bu
x
n+1
= l
n+1
(x
1
x
1
) .
(12)
Consequently, the observer estimation error in (6) becomes
x
1
= x
2
o
1
x
1
.
.
.
x
n1
= x
n
n1
o
n1
x
1
x
n
= x
n+1
n
o
n
x
1
x
n+1
= h(x, w)
n+1
o
n+1
x
1
(13)
and Equation (7) is now
=
o
A + B
h(x, w)
n
o
. (14)
Theorem 2: Assuming h(x, w) is bounded, there exist a
constant
i
> 0 and a nite T
1
> 0 such that | x
i
(t)|
i
, i = 1, 2, , n + 1, t T
1
> 0 and
o
> 0.
Furthermore,
i
= O
_
1
k
o
_
, for some positive integer k.
Proof. Solving (14), it follows that
(t) = e
oAt
(0) +
_
t
0
e
oA(t)
B
h(x() , w)
n
o
d. (15)
Let
p (t) =
_
t
0
e
oA(t)
B
h(x(), w)
n
o
d, (16)
since h(x(), w) is bounded, that is, |h(x(), w)| ,
where is a positive constant, for i = 1, 2, , n + 1, we
have
|p
i
(t)|
n+1
o
_
_
A
1
B
_
i
_
A
1
e
oAt
B
_
i
.
(17)
For A and B dened in (7), A
1
=
_
_
0 0 0
1
n+1
1 0 0
1
n+1
0 1 0
2
n+1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 1
n
n+1
_
_
, and
_
A
1
B
_
i
(18)
where = max
i=2, ,n+1
_
1
n+1
,
i1
n+1
_
. Since A is Hurwitz,
there exists a nite time T
1
> 0 such that
_
e
oAt
ij
n+1
o
(19)
for all t T
1
, i, j = 1, 2, , n + 1. Hence
_
e
oAt
B
n+1
o
(20)
for all t T
1
, i = 1, 2, , n + 1. Note that T
1
depends on
o
A. Let A
1
=
_
_
s
11
. . . s
1,n+1
.
.
.
.
.
.
.
.
.
s
n+1,1
s
n+1,n+1
_
_ and e
oAt
=
_
_
d
11
. . . d
1,n+1
.
.
.
.
.
.
.
.
.
d
n+1,1
d
n+1,n+1
_
_. One has
_
A
1
e
oAt
B
_
i
n+1
o
(21)
for all t T
1
, i = 1, 2, , n + 1, where =
max
i=2, ,n+1
_
1
n+1
, 1 +
i1
n+1
_
. From (17), (18), and (21), we
obtain
|p
i
(t)|
n+1
o
+
2n+2
o
(22)
for all t T
1
, i = 1, 2, , n+1. Let
sum
(0) = |
1
(0)| +
|
2
(0)| + + |
n+1
(0)|. It follows that
_
e
oAt
(0)
sum
(0)
n+1
o
(23)
for all t T
1
, i = 1, 2, , n + 1. From (15), one has
|
i
(t)|
_
e
oAt
(0)
+ |p
i
(t)| . (24)
Let x
sum
(0) = | x
1
(0)| + | x
2
(0)| + + | x
n+1
(0)|. Ac-
cording to
i
=
xi
i1
o
and Equations (22)-(24), we have
| x
i
(t)|
x
sum
(0)
n+1
o
ni+2
o
+
2ni+3
o
=
i
(25)
for all t T
1
, i = 1, 2, , n + 1. Q.E.D.
46th IEEE CDC, New Orleans, USA, Dec. 12-14, 2007 ThB07.6
3503
In summary, it has been proven that 1) when the plant
model is given, the dynamic system describing the estimation
error of the LESO (3) is asymptotically stable; and 2) in the
absence of such model, the estimation error of the LESO (12)
is bounded and its upper bound monotonously decreases with
the observer bandwidth, as shown in (25). The stability of
LADRC, where LESO is employed, is analyzed next.
III. STABILITY ANALYSIS OF LADRC
Assume that the control design objective is to make the
output of the plant in (1) follow a given, bounded, reference
signal r, whose derivatives, r, r, , r
(n)
, are also bounded.
Let [r
1
, r
2
, , r
n
, r
n+1
]
T
= [r, r
1
, , r
n1
, r
n
]
T
. Em-
ploying the LESO of (2) in the form of (3) or (12), the
ADRC control law is given as
u = [k
1
(r
1
x
1
) + k
2
(r
2
x
2
) +
+k
n
(r
n
x
n
) x
n+1
+ r
n+1
] /b (26)
where k
i
, i = 1, 2, , n, are the controller gain parameters
selected to make s
n
+k
n
s
n1
+ +k
1
Hurwitz. The closed-
loop system becomes
y
(n)
(t) = (f x
n+1
) + k
1
(r
1
x
1
) + k
2
(r
2
x
2
)
+ + k
n
(r
n
x
n
) + r
n+1
. (27)
Note that with a well-designed ESO, the rst term in the
right hand side (RHS) of (27) is negligible and the rest of
the terms in the RHS of (27) constitutes a generalized PD
controller with a feedforward term. It generally works very
well in applications but the issues to be addressed are: 1) the
stability of the closed-loop system (27); and 2) the bound of
the tracking error. Note that the separation principal does not
apply here because of the rst term in the RHS of (27).
A. Convergence of the LADRC with the Given Model of the
Plant
Consider
(t) = N (t) + g (t) , (28)
where (t) = [
1
(t) ,
2
(t) , ,
n
(t)]
T
R
n
, g (t) =
[g
1
(t) , g
2
(t) , , g
n
(t)]
T
R
n
, and N is an nn matrix.
Lemma 1: If N is Hurwitz and lim
t
g (t) = 0, then
lim
t
(t) = 0.
Proof. In (28), since lim
t
g (t) = 0, then for any > 0,
there is a nite time T
2
> 0 such that g (t) for all
t T
2
. The response of (28) can be written as
(t) = e
Nt
(0) +
t
_
0
e
N(t)
g () d. (29)
When t T
2
, we have
(t)
_
_
e
Nt
(0)
_
_
+
_
_
e
Nt
_
_
_
_
_
_
_
_
T2
_
0
e
N
g () d
_
_
_
_
_
_
+
t
_
T2
_
_
_e
N(t)
_
_
_d. (30)
Now consider the third term of right hand side of (30). For
N, there is nonsingular matrix J and block diagonal matrix
= block diag {
1
, ,
m
} such that
N = JJ
1
(31)
and each
i
has a single eigenvalue
i
with its algebraic
multiplicities being q
i
. Suppose
1
2
m
. Let
q = max {q
1
, q
2
, , q
m
}. Let us choose
1
or
for
the matrix norm. It follows that
_
_
_e
(t)
_
_
_ e
1(t)
q1
k=0
c
k
(t )
k
, t , (32)
where c
k
are positive constants. Note that
_
_
_e
N(t)
_
_
_ J
_
_
_e
(t)
_
_
_
_
_
J
1
_
_
. (33)
Hence we have
(t)
_
_
e
Nt
(0)
_
_
+
_
_
e
Nt
_
_
_
_
_
_
_
_
T2
_
0
e
N
g () d
_
_
_
_
_
_
+J
_
_
J
1
_
_
q1
k=0
c
k
t
_
T2
e
1(t)
(t )
k
d
=
_
_
e
Nt
(0)
_
_
+
_
_
e
Nt
_
_
_
_
_
_
_
_
T2
_
0
e
N
g () d
_
_
_
_
_
_
+e
1(tT2)
J
_
_
J
1
_
_
q1
k=0
c
k
1
k+1
1
j=0
(1)
j
k!
(k j)!
[
1
(t T
2
)]
kj
+J
_
_
J
1
_
_
q1
k=0
c
k
(1)
k
k!
k+1
1
. (34)
From (34), it can be seen that
lim
t
_
_
e
Nt
(0)
_
_
= 0
lim
t
_
_
e
Nt
_
_
_
_
_
_
_
T2
_
0
e
N
g () d
_
_
_
_
_
= 0
lim
t
_
e
1(tT2)
J
_
_
J
1
_
_
q1
k=0
c
k
1
k+1
1
j=0
(1)
j
k!
(kj)!
[
1
(t T
2
)]
kj
_
= 0.
(35)
Therefore there exists T
3
> T
2
such that
_
_
e
Nt
(0)
_
_
, t T
3
,
_
_
e
Nt
_
_
_
_
_
_
_
T2
_
0
e
N
g () d
_
_
_
_
_
, t T
3
,
e
1(tT2)
J
_
_
J
1
_
_
_
q1
k=0
c
k
1
k+1
1
j=0
(1)
j
k!
(kj)!
[
1
(t T
2
)]
kj
_
, t T
3
.
(36)
46th IEEE CDC, New Orleans, USA, Dec. 12-14, 2007 ThB07.6
3504
Let c
= J
_
_
J
1
_
_
q1
k=0
c
k
(1)
k
k!
k+1
1
. Then we have
(t) (c
+ 3) , t T
3
. (37)
Since can be arbitrarily small, it can be concluded that
lim
t
(t) = 0.
Theorem 3: Assuming h(x, w) is globally Lipschitz with
respect to x, there exist constants
o
> 0 and
c
> 0, such
that the closed-loop system (27) is asymptotically stable.
Proof. Dene e
i
= r
i
x
i
, i = 1, 2, , n. From (26),
one has
u = [k
1
(e
1
+ x
1
) + + k
n
(e
n
+ x
n
)
(x
n+1
x
n+1
) + r
n+1
} /b. (38)
It follows that
e
1
= r
1
x
1
= r
2
x
2
= e
2
,
.
.
.
e
n1
= r
n1
x
n1
= r
n
x
n
= e
n
,
e
n
= r
n
x
n
= r
n+1
(x
n+1
+ bu)
= k
1
(e
1
+ x
1
)
k
n
(e
n
+ x
n
) x
n+1
.
(39)
Let e = [e
1
, e
2
, , e
n
]
T
R
n
, x = [ x
1
, x
2
, , x
n+1
]
T
R
n+1
, then
e (t) = A
e
e (t) + A
x
x(t) (40)
where A
e
=
_
_
0 1 0 0
0 0 1 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 1
k
1
k
2
k
n1
k
n
_
_
and
A
x
=
_
_
0 0 0 0
0 0 0 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 0
k
1
k
2
k
n
1
_
_
.
Since k
i
, i = 1, 2, , n, are selected such that the
characteristic polynomial s
n
+ k
n
s
n1
+ + k
1
is Hur-
witz, A
e
is Hurwitz. For tuning simplicity, we just let
s
n
+ k
n
s
n1
+ + k
1
= (s +
c
)
n
where
c
> 0 and
k
i
=
n!
(i1)!(n+1i)!
n+1i
c
, i = 1, 2, , n. This makes
c
,
which is the controller bandwidth, the only tuning parameter
to be adjusted for the controller.
From Theorem 1, lim
t
A
x
x(t) = 0 if h(x, w) is
globally Lipschitz with respect to x. Since A
e
is Hurwitz,
according to Theorem 1 and Lemma 1, it can be concluded
that: assuming h(x, w) is globally Lipschitz with respect
to x, there exist constants
o
> 0 and
c
> 0, such that
lim
t
e
i
(t) = 0, i = 1, 2, , n. Q.E.D.
B. Convergence of the LADRC with Plant Dynamics Largely
Unknown
Now we consider the case where the plant dynamics is
unknown and the LESO in the form of (12) is used instead.
Theorem 4: Assuming h(x, w) is bounded, there exist a
constant
i
> 0 and a nite time T
5
> 0 such that |e
i
(t)|
i
, i = 1, 2, , n, t T
5
> 0,
o
> 0, and
c
> 0.
Furthermore,
i
= O
_
1
j
c
_
for some positive integer j.
Proof. Solving (40), we have
e (t) = e
Aet
e (0) +
_
t
0
e
Ae(t)
A
x
x() d. (41)
According to (40) and Theorem 2, one has
[A
x
x()]
i=1, , n1
= 0
|[A
x
x()]
n
| k
s
i
= , t T
1
(42)
where k
s
= 1 +
n
i=1
k
i
. Similar to Theorem 3, choose
k
i
=
n!
(i1)!(n+1i)!
n+1i
c
, i = 1, 2, , n, such that
A
e
is Hurwitz. Dene = [0 0 0 ]
T
. Let (t) =
_
t
0
e
Ae(t)
A
x
x() d. It follows that
|
i
(t)|
_
A
1
e
_
i
_
A
1
e
e
Aet
_
i
. (43)
and
_
A
1
e
_
1
=
k1
=
n
c
_
A
1
e
_
i
i=2, , n
= 0
(44)
Since A
e
is Hurwitz, there exists a nite time T
4
> 0 such
that
_
e
Aet
ij
n+1
c
(45)
for all t T
4
, i, j = 1, 2, , n. Note that T
4
depends on
A
e
. Let T
5
= max {T
1
, T
4
}. It follows that
_
e
Aet
_
i
n+1
c
(46)
for all t T
5
, i = 1, 2, , n, and
_
A
1
e
e
Aet
_
i
_
1 +
n
i=2
k
i
n
c
n+1
c
i=1
n+1
c
i=2, ,n
(47)
for all t T
5
. From (43), (44), and (47), we obtain
|
i
(t)|
_
n
c
+
1 +
n
i=2
k
i
n
c
n+1
c
i=1
n+1
c
i=2, ,n
(48)
for all t T
5
. Let e
Aet
=
_
_
o
11
. . . o
1n
.
.
.
.
.
.
.
.
.
o
n1
o
nn
_
_ and e
s
(0) =
|e
1
(0)| + |e
2
(0)| + + |e
n
(0)|. It follows that
_
e
Aet
e (0)
e
s
(0)
n+1
c
(49)
46th IEEE CDC, New Orleans, USA, Dec. 12-14, 2007 ThB07.6
3505
for all t T
5
, i = 1, 2, , n. From (41), one has
|e
i
(t)|
_
e
Aet
e (0)
+ |
i
(t)| . (50)
According to (42), (48)-(50), we have
|e
i
(t)|
_
_
e
s
(0)
n+1
c
+
k
s
n
c
+
_
1 +
n
i=2
k
i
_
k
s
2n+1
c
i=1
e
s
(0) + k
s
n+1
c
i=2, ,n
(51)
for all t T
5
, i = 1, 2, , n, where
i
=
max
_
_
_
es(0)
n+1
c
+
ksi
n
c
+
(1+
n
i=2
ki)ksi
2n+1
c
,
es(0)+ksi
n+1
c
_
_
_
. Q.E.D.
In summary, it has been shown that 1) with the given
model of the plant, the closed-loop system (27) is asymptot-
ically stable; and 2) with plant dynamics largely unknown,
the tracking error and its up to (n 1)
th
order derivatives of
LADRC are bounded and their upper bounds monotonously
decrease with the controller bandwidth, as shown in (51).
IV. CONCLUDING REMARKS
Existing estimators and their characteristics are rst sum-
marized in this paper, concerning unknown disturbance and
plant dynamics. The main result in this paper is the analysis
of the stability and tracking characteristics of a particular
class of such observers, LESO, and the associated feedback
control system, LADRC. Both design scenarios, with and
without a detailed mathematical model of the plant, are
considered. It is shown that the asymptotic stability is assured
in the former and boundedness of the estimation error and
the closed-loop tracking error in the later. Furthermore, it is
shown that the tracking error monotonously decreases with
the control loop bandwidth.
Acknowledgment
The authors would like to thank Prof. Sally Shao at
Cleveland State University for her insightful and valuable
suggestions.
REFERENCES
[1] R. Brockett, New Issues in the Mathematics of Control, Mathematics
Unlimited - 2001 and Beyond, B. Engquist and W. Schimid Ed., pp.
189-220, Springer, 2001.
[2] Z. Gao, Active disturbance rejection control: a paradigm shift in
feedback control system design, Proc. of the American Control
Conference, pp. 2399-2405, 2006.
[3] G. Basile and G. Marro, On the observability of linear, time-invariant
systems with unknown inputs, Journal of Optimization Theory and
Applications, vol. 2, no. 6, pp. 410-415, 1969.
[4] C. Johnson, Accommodation of external disturbances in linear regula-
tor and servomechanism problems, IEEE Transactions on Automatic
Control, vol. AC-16, no. 6, pp. 635-644, 1971.
[5] G. Hostetter and J. Meditch, On the generalization of observers to
systems with unmeasurable, unknown inputs, Automatica, vol. 9, pp.
721-724, 1973.
[6] C. Johnson, Theory of disturbance-accommodating controllers,,
Control and Dynamic Systems, vol. 12, pp. 387-489, 1976.
[7] V. Gourishangkar, P. Kudva, and K. Ramar, Reduced-order observers
for multivariable systems with inaccessible disturbance inputs, Inter-
national Journal of Control, vol.25, pp.311-319, 1977.
[8] P. C. Muller, Indirect measurements of nonlinear effects by state ob-
servers, Proc. IUTAB Symposium Nonlinear Dynamics in Engineering
Systems, pp. 205-215, 1990.
[9] J. Profeta, W. Vogt, and M. Mickle, Disturbance estimation and
compensation in linear systems, IEEE Transaction on Aerospace and
Electronic Systems, vol. 26, no. 2, pp. 225-231, 1990.
[10] J. Chen, R. J. Patton, and H. Zhang, Design of unknown input
observers and robust fault detection lters, International Journal of
Control, vol. 63, no. 1, pp. 85-105, 1995.
[11] T. Umeno, and Y. Hori, Robust speed control of DC servo motors
using modern two degrees-of-freedom controller design, IEEE Trans-
actions on Industrial Electronics, vol. 38, pp. 363-368, 1991.
[12] Y. Hori, K. Shimura, and M. Tomizuka, Position/force control of
multi-axis manipulator based on the TDOF robust servo controller for
each joint, Proc. of the American Control Conference ACC/WM9, vol.
1, pp. 753-757, 1992.
[13] H. Lee and M. Tomizuka, Robust motion controller design for
high-accuracy positioning systems, IEEE Transactions on Industrial
Electronics, vol. 43, pp. 48-55, 1996.
[14] T. Mita, M. Hirata, K. Murata, and H. Zhang, H control versus
disturbance-observer-based control, IEEE Transactions on Industrial
Electronics, vol. 45, no. 3, pp. 488-495, 1998.
[15] R. Bickel and M. Tomizuka, Passivity-based versus disturbance
observer based robot control: equivalence and stability, ASME Journal
of Dynamics Systems, Measurement, and Control, vol. 121, pp. 41-47,
1999.
[16] E. Schrijver and J. van Dijk,Disturbance observers for rigid mechan-
ical systems: Equivalence, stability, and design, ASME Journal of
Dynamics Systems, Measurement, and Control, vol. 124, no. 4, pp.
539-548, 2002.
[17] Y. Choi, K. Yang, W. K. Chung, H. R. Kim, and I. H. Suh, On the
robustness and performance of disturbance observers for second-order
systems, IEEE Transactions on Automatic Control, vol. 48, no. 2, pp.
315-320, 2003.
[18] K. Yang, Y. Choi, and W. Chung, On the tracking performance
improvement of optical disk drive servo systems using error-based
disturbance observer, IEEE Transactions on Industrial Electronics,
vol. 52, no. 1, pp. 270-279, 2005.
[19] S. Kwon and W. K. Chung, Robust performance of the multiloop
perturbation compensator, IEEE/ASME Transaction on Mechatron.,
vol. 7, no. 2, pp. 190-200, 2002.
[20] S. Kwon and W. K. Chung, A discrete-time design and analysis of
perturbation observer for motion control applications, IEEE Trans-
actions on Control Systems Technology, vol. 11, no. 3, pp. 399-407,
2003.
[21] S. Kwon and W. K. Chung, Combined synthesis of state estimator
and perturbation observer, ASME Journal of Dynamic Systems, Mea-
surement, and Control, vol. 125, pp. 19-26, 2003.
[22] S. Kwon, Robust kalman ltering with perturbation estimation
process, Proc. of the American Control Conference, pp. 997-1002,
2006.
[23] Z. Gao, Y. Huang, and J. Han, An alternative paradigm for control
system design, Proc. of IEEE conference on Decision and Control,
pp. 4578-4585, 2001.
[24] Z. Gao, Scaling and parameterization based controller tuning, Proc.
of the American Control Conference, pp. 4989-4996, 2003.
[25] B. Yao, M. Al-Majed, and M. Tomizuka, High-performance robust
motion control of machine tools: An adaptive robust control approach
and comparative experiments, IEEE/ASME Transaction on Mecha-
tron., vol. 2, pp. 63-76, 1997.
[26] B.-K. Choi, C.-H. Choi, and H. Lim, Model-based disturbance atten-
uation for CNC machining centers in cutting process, IEEE/ASME
Transaction on Mechatron., vol. 4, pp. 157-168, 1999.
[27] A. Tornambe and P. Valigi, A Decentralized controller for the robust
stabilization of a class of MIMO dynamical systems, Journal of
Dynamic Systems Measurement and Control, vol. 116, pp. 293-304,
1994.
[28] Q. Zheng and Z. Gao, Motion control opmitimization: problem and
solutions, Interntional Journal of Intelligent control and systems, Vol.
10, No. 4, pp. 269-276, 2006.
46th IEEE CDC, New Orleans, USA, Dec. 12-14, 2007 ThB07.6
3506