Вы находитесь на странице: 1из 10

Systematic design of weighting matrices for the H

1
mixed sensitivity problem
M.G. Ortega*, F.R. Rubio
Departamento Ingeniera de Sistemas y Automa tica, Escuela Superior de Ingenieros, Universidad de Sevilla, 41092 Sevilla, Spain
Received 4 December 2002; received in revised form 1 April 2003; accepted 12 May 2003
Abstract
A systematic design of weighting matrices for the H
1
mixed sensitivity problem is presented in this paper. Once a nominal model
has been chosen, an initial design of the weighting matrices based on the multiplicative output uncertainty is proposed. The nal
weighting matrices (which permit that an appropriate closed loop behavior is achieved)are obtained by tuning just one parameter
for each output of the system. A multivariable control of two temperatures of a pilot plant (which constitutes a typical example of
an industrial process) is included as an application where the validity of the proposed methodology has been tested.
#2003 Elsevier Ltd. All rights reserved.
Keywords: H
1
control; Mixed sensitivity problem; Weighting matrix; Process control
1. Introduction
The control systems community has been paying
increasing attention to the H
1
control theory throughout
the past decade because of the robustness characteristics
supplied by its controllers. This control theory is based
on a more realistic hypotheses with respect to the
restrictions imposed on the dierent signals which appear
in a control scheme. In this case, no special statistical
distribution is imposed, and it is only supposed that the
signal energy are bounded.
This article deals with the design of appropriate
weighting functions for the mixed sensitivity problem,
which takes place in the frame of the synthesis of H
1
controller. Many articles have been published during
the last few years in which a particular control applica-
tion has been solved by means of this approach (see [110]
as representative examples). However, despite the
importance of the design of the weighting functions, no
methodology appears in these papers regarding the
chosen functions. Furthermore, most of them conrm
that these functions are tuned through a trial and error
process.
On the other hand, there are some publications where
the main aim is to expose some considerations about the
design of these functions (see, for example, [1118]).
However, it is necessary to have a wide knowledge of
control theory to understand them, and above all to use
them.
There are many industrial applications where the
engineers in charge to control them do not possess an
advanced knowledge about control theory. This is the
reason why most of these applications are controlled by
means of PID control laws which regulate SISO process
loops. In these cases, it may be better to apply advanced
control laws that improve the behavior of the controlled
system. Nevertheless, being able to tune these con-
trollers in an easy and intuitive mode, through no more
than one or two parameters which were representative of
the closed loop system performance, could be interesting.
In this paper, a simplied design methodology of the
weighting functions is proposed. Thereby, after an
initial adjustment based on the estimated uncertainty,
the system performance will basically depend on the
value of adimensional parameters k
i
(i =1, . . ., no. out-
puts). These parameters have been normalized so that
they should adopt positive values for a usual behavior.
Values equal to zero are proposed for initial selection.
0959-1524/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0959-1524(03)00035-0
Journal of Process Control 14 (2004) 8998
www.elsevier.com/locate/jprocont
* Corresponding author. Tel.: +34-9-544-87356; fax: +34-9-544-
87540.
E-mail addresses: mortega@esi.us.es (M.G. Ortega), rubio@esi.
us.es (F.R. Rubio).
As these parameter values increase, faster (although
more oscillatory) responses are obtained. Therefore, the
nal selection of these parameter values would be a
compromised solution between quality and speed of the
response.
The remainder of this article is organized as follows:
the mixed sensitivity problem in the frame of the H
1
control is presented in the rst part. This is followed by
an exposition of how each weighting matrix has to be
calculated and a summary of the design methodology.
Finally, a case study is exposed in order to check the
validity of the proposed methodology.
2. The H
1
mixed sensitivity problem
The feedback controller design problem can be for-
mulated as an H
1
optimization problem, which can be
posed under the general conguration shown in Fig. 1.
In this gure, P(s) is the generalized plant, K(s) is the
controller, u are the control signals, v the measured
variables, o the exogenous signals and z are the so-
called error variables.
The optimal H
1
control problem with this con-
guration consists of computing a controller such that
the ratio , between the energy of the error vector z and
the energy of the exogenous signals vector o is mini-
mized. This optimal problem is not solved yet, but a
solution exists for the suboptimal problem (see [19] for
the continuous time domain case and [20] in case of
discrete time domain). Thereby, the value of the energy
ratio , is decreased as much as possible by means of an
iteration process. This is the synthesis process which has
been implemented in various well-known software
packages such as [21] or [22].
A conguration for building up the generalized plant
is the S,T mixed sensitivity problem (see, for example,
[17]), which is exposed in Fig. 2. In this case, the
expression of the resulting closed loop transfer function
T
zo
(s) is as follows:
T
zo
s
W
S
s S
o
s
W
T
s T
o
s
_ _
where S
o
(s) is the output sensitivity transfer matrix and
T
o
(s) is the output complementary sensitivity transfer
matrix:
S
o
s I G s K s
1
T
o
s G s K s I G s K s
1
The terms W
S
(s) and W
T
(s) constitute their respective
weighting matrices, which allow the range of frequencies
of main importance for the corresponding closed-loop
transfer matrix to be specied. As it is known, shaping
T
o
(s) is desirable for tracking problems, noise attenua-
tion and for robust stability with respect to multi-
plicative output uncertainties. On the other hand, since
S
o
(s) relates the error signals with references and dis-
turbances [23], shaping the sensitivity function will
allow the performance (in terms of command tracking
and disturbance attenuation) of the system to be con-
trolled.
Therefore, since the controller is obtained from the
generalized plant, the synthesis problem with this con-
guration is reduced to the design of a nominal model
G(s) and some appropriate weighting matrices which
will impose the control specications. Once these
designs have been carried out, the generalized plant can
be built up, and consequently the controller can be
computed by means of a synthesis algorithm imple-
mented in a computer.
3. Design of the weighting matrices
According to what has been stated in the introduc-
tion, this paper supplies some design rules for the
weighting matrices which automate the synthesis pro-
cess of H
1
controllers through the mixed sensitivity
problem. Steps to be followed for an appropriate design
of each term of the generalized plant are exposed in the
following section.
Fig. 1. General formulation of the control problem. Fig. 2. S,T mixed sensitivity conguration.
90 M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998
3.1. Selection of the nominal plant and estimating
uncertainties
The synthesis of a robust controller is based on the
knowledge of the system uncertainty. In this paper, the
unstructured multiplicative output description of the
uncertainty is used, where the choice of the nominal
model is the rst step to estimate it.
A low order model is proposed as the nominal plant.
This selection yields a simplied nominal system, which
benets the calculus of the controller. However, an
inconvenience to take into account is that the resulting
uncertainty region may be widespread.
Once the nominal model G(s) is chosen, the multi-
plicative output uncertainty can be estimated as follows:
E
o.i
s G

i
s G s
_ _
G s
1
. i 1. 2. . . .
where G

i
s stands for the dierent nonnominal systems
at each operating point where the controller is needed to
work properly. Thereby, this calculus must be repeated
as many times as the number of nonnominal models
G

i
s are being considered. By denition, the multi-
plicative uncertainty means the percentage of ignorance
of the plant at each frequency. Usually, this percentage
increases as the frequency does, and there always exists
a frequency value that starting from this one, the system
ignorance is complete, that is, the value of the multi-
plicative uncertainty is greater than one.
3.2. Design of the weighting matrix W
T
(s)
Since the class of uncertainty employed corresponds
to the multiplicative output one, its associated robust
stability condition is given by the following expression
[23]:
W
1
s T
o
s W
2
s
_
_
_
_
1
41
where T
o
(s) is the output complementary sensitivity
function, and W
1
(s) and W
2
(s) matrices are such that
they normalize the estimated uncertainty, that is:
E
o
s W
2
s D s W
1
s : D s
_
_
_
_
1
41
_ _
If the value of W
2
(s) is set to the identity and W
1
(s) is
recalled as W
T
(s), the resulting robust stability condi-
tion has the following expression:
W
T
s T
o
s
_
_
_
_
1
sup
o
o W
T
jo T
o
jo 41
It can be seen how the term W
T
(s)T
o
(s) is the same
that appears in the mixed sensitivity models where the
complementary sensitivity function is involved. There-
fore, to diminish the innity norm in the mixed sensi-
tivity problem implies making the innity norm of
W
T
(s)T
o
(s) small, and so, making the system robust
against this uncertainty class.
The proposed design of W
T
(s) consists of a square
diagonal matrix with all its diagonal elements with the
same transfer function:
W
T
s W
T
diag
s I
qq
such that the matrix dimension q is equal to the number
of system outputs. The transfer function W
T
diag
(s) must
be stable, minimum phase and with module greater than
the maximum singular value of the uncertainty pre-
viously calculated for each nonnominal model and fre-
quency, that is,
W
Tdiag
jo

5o E
o.i
jo
_ _
8o. 8i
Moreover, taking into account that W
T
(s) must
weight to the complementary sensitivity function, it is
desirable that the module of W
T
diag
(s) has a high value in
order to impose that T
o
(s) has a small gain at high fre-
quencies.
Notice that since W
T
(s) is diagonal, its singular values
for each frequency are equal to the module of the
transfer function W
T
diag
(s), that is,
o
i
W
T
jo W
T
diag
jo

i 1. . . . . q 8o
Consequently, this selection of W
T
(s) ensures the
closed loop robustness respect to any multiplicative
error in the set:
E
o
s W
Tdiag
s D s : D s
_
_
_
_
1
41
_ _
D

s : o D

jo
_ _
4 W
Tdiag
jo

. 8o
_ _
and obviously, all calculated uncertainties are elements
of this set, that is, E
o.i
s 2 E
o
s , 8i.
3.3. Design of the weighting matrix W
S
(s)
The weighting matrix W
S
(s) will be employed to
impose performance conditions to the system. As it is
known, the sensitivity function is especially interesting
from the control point of view as it has properties that
characterize both the quality and the speed of the closed
loop temporal response (see, for example, [17]).
Bearing in mind that the matrix W
S
(s) must weight
the sensitivity function, its design is proposed to be a
square diagonal matrix of transfer functions:
W
S
s diag W
S11
s . . . . . W
Sii
s . . . . . W
Sqq
s
_ _
where the matrix dimension q is again the number of the
systems outputs. Each diagonal element W
Sii
(s) must be
M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998 91
designed bearing in mind that its inverse should shape
to S
ii
jo

as an upper bound, where S


ii
(s) is the ith
element of the diagonal of S
o
(s). Thereby, these ele-
ments are proposed with the following expression:
W
Sii
s
o
i
s o
Si
s [
i
o
Si
Notice that since the computed controller tries to
decouple the outputs, the resulting matrix function S
o
(s)
will be almost diagonal.
Next it will be shown how to design each one of these
parameters:
o
i
is the function gain at high frequency. As it is
well known, a characteristic value for the max-
imum peak of the magnitude of S
ii
(s) is about 2 if
an acceptable system behavior is required.
Thereby, since W
Sii
jo

1
must impose an upper
bound on S
ii
jo

, it is desirable that the gain of


W
Sii
s
1
at high frequency, that is 1,o
i
, is about
2. Consequently, an appropriate value for each o
i
should be approximately about 0.5.
[
i
is the function gain at low frequency. This
value supposes an upper bound on the allowed
steady state error. From the control point of
view, it would be desirable to get null steady
state errors, which implies adopting values for [
i
equal to zero. However, this selection would
give numeric problems in the synthesis algo-
rithm because the augmented plant would have
null values on the imaginary axis. An appro-
priate small value for these parameter may be
between 10
6
and 10
4
depending on the
application.
o
Si
is the crossover frequency of the function.
These values indicate the minimum bandwidth of
the transfer functions which are weighted. As an
initial value a decade below the crossover fre-
quency of the function W
T
diag
(s) previously
designed is proposed. This frequency will be
called in the following as o
T
. According to the
proposed design for W
T
(s), this frequency must
be lower than the one at which the maximum
singular value of the uncertainty gets the unity
value. This initial selection yields slow responses
and rarely oscillatory ones. Dierent experiments
showed the convenience of varying the value of
o
Si
in a logarithmic way. In this sense, it is pro-
posed to express this value as the following
expression:
o
Si
10
k
i
1
o
T
where the parameter k
i
is employed to vary the
value of o
Si
once the value of o
T
has been
obtained. The initial value of o
Si
is obtained for
k
i
equal to zero, while a value of this parameter
equal to one shows that o
Si
is equal to o
T
.
Therefore, the nal selection of this frequency is
determined by an adimensional parameter where
the value must be higher as the desired response
speed increases.
4. Design methodology
Once it has been shown how to design each weighting
matrix, in following section the steps to follow to syn-
thesize the nal controller is summarized.
1. Choose a nominal model with a low order.
2. Estimate the multiplicative output uncertainty of
the system with respect to the chosen nominal
model.
3. Design the weighting matrix W
T
(s) as exposed in
Section 3.2.
4. Design an initial weighting matrix W
S
(s) as
exposed in Section 3.3 with k
i
=0 and with
standard values of both o
i
and [
i
.
5. Build up the augmented plant P(s) and synthesize
an initial controller K(s).
6. Design W
S
(s) again by adjusting the values of k
i
until obtaining the desired behavior of the tem-
poral response.
Finally, it is convenient to remember the importance
of the system scaling as a previous stage to the controller
design. Besides avoiding some numeric problems (for
example, decreasing the condition number of the sys-
tem), in the case of multivariable systems, scaling per-
mits that all signals are comparable in magnitude, and
particularly the errors at the dierent channels. Bearing
in mind that the H
1
control attenuates the energy of the
error vector z, all its components will have the same
relative weight if the system is scaled.
5. A case study
In this section the validity of the proposed design
methodology of H
1
controllers is tested in a pilot plant,
which constitutes a typical example of industrial pro-
cess.
5.1. System description
The pilot plant consists of industrial components,
such as a water tank, a resistor to heat the water, a heat
exchanger, pumps and dierent pneumatic valves. A
diagram and a photograph of the plant which shows its
main elements as well as the localization of the various
92 M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998
instruments are presented in Fig. 3. The plant is used
nowadays as a test bed of control strategies (see [24,25])
which can be implemented on an industrial SCADA
which is connected to it.
The aim of this application is to control some tem-
peratures, as is usual in any industrial process. In this
case, they are the tank temperature (TT
5
) and the out-
put exchange temperature (TT
4
) (see Fig. 3b). The con-
trol variables are the heat supplied to the tank water by
a resistor (R), and the water ow in the recirculated
circuit (FT
4
), which is regulated by the aperture of the
valve V
8
. A heat exchanger reduces the temperature of
the recirculation water driven by the pump, using a
constant ow of cold water (temperature TT
2
); the other
valves are closed. This constitutes a MIMO system with
two inputs and two outputs.
A robust controller is needed to keep a good perfor-
mance of the temperature values due to the dierent
system behaviors depending on the operating point. The
heat exchanger is the main cause of these dierent
behaviors, where the eciency may change drastically
from one operating point to another. This inserts a high
nonlinearity into the system, that will be dealt like an
unstructured uncertainty.
5.2. Controller design
Steps to synthesize the controller using a S,T mixed
sensitivity model are related in this section. Since the
system is multivariable, a scaling of the plant will be
made as a previous step to the design of the weighting
matrices.
5.2.1. Obtaining a linear model
Linear models of the system were obtained by identi-
cation at several operating points by supplying a
pseudo random binary sequence to both actuators, i.e.
the water ow FT
4
and the resistor R. In this plant the
working space is bounded due to physic conditions.
This working space is shown in Fig. 4, where the
minimum and maximum values of the allowed control
variables for a proper behavior of the plant can be
observed. Taking into account this working space, the
nominal operating point (NP) has been chosen with
values of R =60% and V
8
=80%. This selection yields
equilibrium temperatures of TT
4
and TT
5
equal to 36.7
and 56.3

C respectively. Another four points are
Fig. 3. Pilot plant
Fig. 4. Working space of the control variables.
M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998 93
marked in Fig. 4, which correspond to dierent working
points around the nominal one. These points will be
used for the estimation of the uncertainty with respect
to the nominal model.
Experimental results at the dierent operating points
were given to an identication algorithm ([26]) using a
multivariable ARX model with a sample time of one
second. The transfer functions obtained in the case of
the nominal operating point were the following:
TT
4
TT
5
_ _

G
11
z G
12
z
G
21
z G
22
z
_ _
FT
4
R
_ _
where:
G
11
z
1.6511z 1.6515
z
2
0.99861z
G
12
z
0.00017087z
2
0.0000017633z 0.00027567
z
2
0.99861z
G
21
z
0.10879z 0.10306
z
5
0.99856z
4
G
22
z
0.00068788z
2
0.0000039192z 0.00029806
z
5
0.99856z
4
5.2.2. Scaling the plants and selection of the nominal
model
To carry out the scaling, it is necessary to determine
the expected magnitudes of the maximum changes of
control signals over each input and the maximum
variations allowed of the outputs. In this application
the following values have been chosen as maximum
variations:
TT
4 max
2.5

C TT
5 max
5.0

C FT
4 max
2 l,m
R
max
10%
Taking into account these values, the system can be
scaled by means of the following expression ([17]):
G
^
z D
1
e
G z D
u
where the two scaling matrices D
e
and D
u
can be built as
follows:
D
e

TT
4 max
0
0 TT
5 max
_ _

2.5 0
0 5
_ _
D
u

FT
4 max
0
0 R
max
_ _

2 0
0 10
_ _
Once the systems are scaled, the nominal model is
chosen as the model identied at the nominal operating
point but removing the delays which appear in it.
This is carried out in order to reduce the controller
order.
5.2.3. Estimating uncertainties
The output multiplicative uncertainty is estimated by
means of the expression:
E
^
o
z G
^

z G
^
z
_ _
G
^
z
1
where G
^
z is the nominal model and G
^

z stand for
each nonnominal identied model. Maximum singular
values of these estimated uncertainties (calculated using
the transformation z =e
joT
with T =1 s) are shown in
Fig. 5. It can be observed how there is a model (dashed
line) where its maximum singular value of the uncer-
tainty is greater than one at low frequency. This means
that there is no knowledge at all of the system for this
operating point (P
3
in the working space of Fig. 4) with
respect to the chosen nominal model. Taking into
account that this working point is close to limit of the
allowed region, it will not be considered for the con-
troller design. Therefore, it will be expected that the
controller does not work properly at this operating
point.
5.2.4. Initial design of the weighting matrices
The continuous time domain is used to design the
weighting lters, which may be discretized by a bilinear
transformation in case a discrete synthesis algorithm is
applied.
In this application, the weighting matrix W
T
(s) has
been designed as follows:
W
T
s W
T
diag
s I
22
0.6310
100s 1
0.1s 1
I
22
The module of its diagonal elements is plotted in Fig. 5,
where it can be seen how it satises the constraints
imposed in Section 3.2.
The initial W
S
(s) is designed as a square diagonal
matrix where each of its diagonal elements are as follows:
W
Sii
s
o
i
s 10
k
i
1
o
T
s [
i
10
k
i
1
o
T
. i 1. 2
The parameters of these functions have been selec-
ted as the proposed values, that is, o
1
=o
2
=0.5 and
[
1
=[
2
=10
4
. In addition, a value of o
T
equal to
0.015 rad/s approximately is obtained from the design
of W
T
(s) carried out previously. Finally, values of k
1
and k
2
(associated to the response speed of TT
4
and
TT
5
respectively) equal to zero are selected for the
initial controller. As will be exposed afterwards, these
values must be adjusted to obtain an appropriate
performance.
94 M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998
5.2.5. Building up the generalized plant and obtaining
the controller
In this case, a continuous time H
1
synthesis algo-
rithm ([19]) has been employed to obtain the controller.
This was carried out by converting the nominal model
to the continuous time domain by a bilinear inverse
transformation as a previous step of building up the
generalized plant. The synthesized controller is then
converted to the discrete time domain again by a
bilinear transformation. The synthesis process in this
case is summarized in the following scheme:
Finally, at this point it is important to remember that
the controller has been synthesized from a scaled model.
To obtain the controller to be implemented in the real
application it will be necessary to carry out a scaling
inverse process. Thereby, the real controller has to be
calculated from the following expression:
K z D
u
K
^
z D
1
e
where D
e
and D
u
matrices are the same ones used in the
scaling process of the plant.
5.3. Experimental results
The step responses of the output temperatures will be
plotted in order to evaluate the performance of the
controllers. A step of 2.5

C is supplied as reference for
TT
4
while one of 5

C is supplied for TT
5
. These mag-
nitudes coincide with the maximum variation allowed
for each output in the plant scaling.
Fig. 6 shows the results obtained with the initial
selection for W
S
(s). It can be noted that, as expected, the
system response is too slow and without any overshoot.
Next, both k
1
and k
2
were increased until values were
equal to 0.5. The system response in this case is shown
in Fig. 7. The evolution of TT
4
is still too slow while the
one of TT
5
starts to be oscillatory. This means that the
value of k
1
could be increased again while it would be
convenient to decrease the value of k
2
.
Finally, in Fig. 8 are presented the step response of
the system corresponding to a controller computed with
k
1
and k
2
equal to 1.3 and 0.2, respectively. In this case,
both responses are faster than the ones obtained with
Fig. 5. W
T
(s) matrix as an upper bound of the singular values of the multiplicative output uncertainty.
M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998 95
Fig. 6. Experimental results with k
1
=k
2
=0.
Fig. 7. Experimental results with k
1
=k
2
=0.5.
96 M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998
the initial controller, with a reasonable rise time and
overshoot. In addition, it can be seen how a change in a
temperature reference hardly aects the other output.
6. Conclusions
A design methodology of weighting functions has
been proposed for the H
1
mixed sensitivity problem.
This method allows the application of this control tech-
nique without a wide theoretical knowledge of advanced
control, although a large amount of information of the
real system is recommended. Despite the simplicity of
the method, which is practically reduced to an easy
tuning of a parameter for each system output, the H
1
controller designed is able to cope with changing oper-
ating conditions and disturbances acting on the systems.
The validity and characteristics of the H
1
controllers
have been tested experimentally with a real plant.
Acknowledgements
The authors wish to acknowledge CICYT for funding
this work under grants DPI-2001-2424-C02-01 and
DPI2000-1218-C04-01. They also acknowledge the
many helpful suggestions by reviewers and editors that
have been incorporated into the manuscript.
References
[1] V. Athans, S. Agarwal, Designing of a robust controller for a
supersonic aircraft using 1 approach, Control Eng. Practice 6
(1994) 10511061.
[2] Lopez, M.J., Rubio, F.R. H
1
Multivariable design methodology
for a ship, in: Proc. of 3rd European Control Conference, Rome,
Italy, 1995, pp. 607612.
[3] R. Banavar, P. Dominic, An LQG,H
1
Controller for a exible
manipulator, IEEE Trans. on Control Systems Technol. 3 (4)
(1995) 409416.
[4] B. Bergeon, P. Martinez, P. Coustal, J.P. Granier, Design of an
active suspension system for a micro-gravity experiment, Control
Eng. Practice 4 (11) (1996) 14911502.
[5] J. Yan, S.E. Salcudean, Teleoperation controller design using
H
1
-optimization with application to motion-scaling, IEEE
Trans. on Control Systems Technol. 4 (3) (1996) 244258.
[6] H. Lee, H. Hwang, Design of two-degree-of-freedom robust
controllers for a seeker scan loop system, International Journal of
Control 66 (4) (1997) 517537.
[7] H. Imanari, Y. Morimatsu, K. Sekiguchi, H. Ezure, R. Matuoka,
A. Tokuda, H. Otobe, Looper 1 control for hot-strip mills,
IEEE Trans. Indust. App. 33 (3) (1997) 790796.
[8] H. Kitada, O. Kondo, H. Kusachi, K. Sasame, H
1
control of
molten steel level in continous caster, IEEE Trans. Control Syst.
Technol. 6 (2) (1998) 200207.
Fig. 8. Experimental results with k
1
=1.3 and k
2
=0.2.
M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998 97
[9] J.L. Lin, B.C. Tho, Analysis and m-based controller design for an
electromagnetic suspension system, IEEE Trans. Educ. 41 (2)
(1998) 116129.
[10] A. Bittar, R. Moura-Sales, H
2
and H
1
control for maglev vehi-
cles, IEEE Control Syst. 18 (4) (1998) 1825.
[11] P. Lundstrom, S. Skogestad, Z.Q. Wang, Performance weight
selection for H
1
and m-control methods, Trans. Inst. MC 13 (5)
(1991) 241252.
[12] E.J.M. Geddes, I. Postlethwaite, An H
1
based loop shaping
method and m-synthesis, Proceedings of the 30th Conference on
Decision and Control 1 (1991) 533538.
[13] Y.S. Hung, Pokrud, An H
1
approach to feedback design with
two objective functions, IEEE Trans. on Automatic Control 37
(6) (1992) 820824.
[14] H.A.H.A. Anwar, A Survey of Weighting Functions for 1 Con-
trol, Control Systems Centre, Manchester, 1993 (Report No.
782).
[15] M.J. Grimble, Robust Industrial Control. Optimal Design
Approach for Polynomial Systems, Prentice-Hall International
(UK), 1994.
[16] G. Meinsma, Unstable and nonproper weights in 1 control,
Automatica 31 (11) (1995) 16551658.
[17] S. Skogestad, I. Postlethwaite, Multivariable Feedback Control.
Analysis and Design, John Wiley & Sons, New York, 1996.
[18] N. Sebe, A design of controllers for simultaneous 1 control
problem, International Journal of Systems Science 30 (1) (1999)
2531.
[19] J.C. Doyle, K. Glover, P. Khargonekar, B. Francis, State-space
solutions to standard H
2
and H
1
control problems, IEEE Trans.
on Automatic Control 34 (8) (1989) 831847.
[20] P. Iglesias, K. Glover, State-space approach to discrete-time H
1
control, International Journal of Control 54 (5) (1991) 1031
1073.
[21] G. Balas, J. Doyle, K. Glover, A. Packard, R. Smith, m-Analysis
and Synthesis Toolbox Users Guide, 2nd Edition, The Math
Works, Natick, MA, 1995.
[22] R. Chiang, M. Safonov, Robust Control Toolbox Users Guide,
2nd Edition, The MathWorks, Natick, 1998.
[23] K. Zhou, J.C. Doyle, K. Glover, Robust and Optimal Control,
Prentice-Hall, NJ, 1996.
[24] F. Castan o, F.R. Rubio, Application of a LQG/LTR controller
to a pilot plant, in: Proceedings of CONTROLO98, 1998.
[25] D.R. Ram rez, D. Limo n-Marruedo, J. Go mez-Ortega, E.F.
Camach, Model based predictive control using genetic algo-
rithms. Application to a pilot plant, in: Proceedings of the Eur-
opean Control Conference, ECC99, 1989, pp. 8185.
[26] L. Ljung, System Identication Toolbox, The Math Works,
1986.
98 M.G. Ortega, F.R. Rubio / Journal of Process Control 14 (2004) 8998

Вам также может понравиться