Recent De V Elopments In Gain Scheduling Control

Recent
Chemical
De
Engineering,
v
elopments
N-7034
Norwegian
Kjetil
T
rondheim,
in
University
b
1
Ha
y
Gain
vre
Norway
of
Scheduling
Science
and
T
echnology
Control
,

Summary
Linear
Systems
Parameter
Mathematical
Classiﬁcations
Introduction,
parameter
.
with
dependent
deﬁnition
linear
descriptions
of
varying
gain-scheduling
fractional
systems
and
systems
of
motivation.
linear
dependence
and
Outline
control
with
2
control.
time
induced
varying
techniques.
in
the
quadratic
systems.
parameters
performance.
(LFT
-systems).

£
Example:
tion
operation
A
steady-state
Originally
¥
)
gain
is
is
tabulated
scheduled
scheduling
the
T
a
ime
process
control
parameter
What
as
varying
(determined)
controller
a
function
gain
scheme
is
PID
is
gain
.
measured
is
control,
of
a
to
according
scheduling
operating
parameterized
counteract
3
where
(available)
to
conditions.
¡
the
control?
nonlinear
¢
(may
parameter
set
and
of
also
the
linear
variations
.
controller
include
controllers.
£
in
¤
in
and
ac-
the
In

¦
¨
Gain
scheduling
4
control.
R4
R3
R2
R1
¦
§

Steps
4)
3)
2)
1)
Implement
Design
Design
Select
in
the
a
the
the
set
design
the
scheduling
controllers
of
parameter-scheduled
stationary
of
a
algorithm.
for
gain
5
each
operating
scheduling
operating
points.
controller
point.
controller
.
.

2)
1)
Application
Gain
-
-
Altitude
Mach
scheduling
number
of
(height
gain
control
(static
scheduling
above
Motivating
air
of
sea
airplane.
pressure
level).
6
control
examples
Scheduling
and
in
chemical
velocity).
variables:
process
control.

T
ennessee
Eastman
7
FI
FI
8
FI
1
CWS
9
A
Purge
JI
Compressor
TI
Sp
PI
XA
plant
FI
CWR
LC
LI
2
SC
Condenser
XB
D
A
XC
N
7
A
XD
L
FI
TI
Y
LI
PI
XE
3
5
Z
E
FI
E
CWS
XF
10
R
PI
XG
Sp
XH
XA
6
TI
Vap/Liq
A
TC
XB
Separator
N
A
CWR
Stripper
XC
L
FI
12
XD
TI
Y
A
TI
XD
Z
N
XE
E
A
XE
R
Reactor
FI
L
XF
LI
Y
Stm
XF
Z
XG
E
Cond
R
XH
FI
FI
4
C
Product
11
LC Sp

Adaptive
©
linearly
8
controller?

In
Adam
this

The
Extended
Gain-scheduling
Linear
presentation
Lagerberg
Classiﬁcation
D-method.
Parameter
linearization
(1996):
only
with
V
arying
the
of
Linear
and
gain-scheduling
LFT
(LPV)
linearization
and
Fractional
approach.
the
10
LPV
T
families.
ransformations
approaches
control
techniques
will
(LFT).
be
considered.

Note,
2)
1)

L
L
Linear
Linear
Linear
TV
TI-system
a
-system
LPV
T
Parameter
T
-system
ime
ime
Three
for
for
V
Invariant
"
arying

"

becomes
types
V
"
const.
arying
(L
!%$
(L
#
TV)

(along

TI)
of

and
a:

systems

(LPV)

systems

“linear

a

time-varying
systems
1

!
"
!
"
1
!%$
!%$

”
#
#
#
#
system

!%$
!
"
!%$
!
"

#
#
#
#

trajectory).

descriptions

For
The
Parameter-Dependent

The
The
control:
there
takes
is
parameter
time-varying,
parameter
time
are
values
variations
known
"
:
in
values
a
i.e.
bounds
compact
'
&

Parameter-Dependent

©
(

"
linear
are
)

!%$
"
#
!%$
not
,
on
plants:
#
'
&
are
known

set

"
1
0
!
"
(which
!
!
"
"
measured
3
#
,
#
#
and
in

12
advance.
may
1
0
0
0
0
!
"
!
!
"
"
#
,
#
#
in
or
real-time

may
1
0
1
Systems
1
1
!
"
!
!
"
"
#
(
not
#
#
)
with
be
'
&
2

(

)
exploited)
sensors.

Optimizing
-

Parameter-Dependent
"
b
Parameter
nonlinearly
and
closed-loop
dependent
.
7
performance
I
controller
C
Control
6
6
@
@BA
with
¡
D
G
4
P
@
E
which
5
@
H
of
respect
Parameter-Dependent
13-1
D
processes
C
to
parameter
F
698
variations.
Systems

Distinctions
Consider

The
the
T
The
Linear
Linear
o
which
analysis.
dif
allowable
two
ferent
Fractional
Parameter
:
extent
types
techniques
dependence
of
information
T
Parameter-Dependent
V
Parameter-Dependent
ransformation
arying
used
(LPV)
that
about
to
the
,
analyze
(LFT)
and
the
13
state-space
parameter
models.
the
systems
systems.
Systems
data
’
s
variation
has
on
are
the
parameters.
exploited
in

Natural

Choosing
Parameterization
Stabilizability
Stability
to
look
.
¡
Obvious
into:
to
and
optimize
of
Detectability
all
14
questions
stabilizing
performance.
.
controllers.

LFT
T
wo

LPV
LFT
L
When
Stability
and
TV
facts:
Parameter-Dependent
-systems:
-systems:
-system.
LPV
evaluating
of
systems,
L
TV
Parameter
Structured
-system
along
two
dif
a
is
allowable
dependent
Small-Gain
ferent
well
characterized.
Systems:
approaches
15
trajectory
L
theorems
yapunov
Stability
,
to
the
.
functions
stability
P-D
system
Deﬁnitions
tests:
,
becomes
an

is
W
Detectability
is
R
Stabilizability
stable
!
"
stable
!
"
#
,
#

such
!%$
#
P-D
for
for
,
such
that:
all
all
of
Systems:
of
possible
possible
!
that:

!
!
"
!
"
#
Q

#
Q

parameters
!
"
parameters
!
"

Stabilizability
#
#
:
#
#
!
"
If
!
"
:
#
there
#
If

!%$
there
!%$
16
#TS
#
S
"VU
"VU
exists

W
exists
!
"
!
"
3
3
#
#
and

a
R
!
"
P-D
!
"
a
#
#
P-D

Detectability
!%$
output
!%$
#
#
state-feedback
to
state-feedback
2
!%$

#

Output
Extension

Usual
2)
1)
if
There
Given
with
controlled
the
and
The
The
feedback
two
their
observer
only
exists
an
to
pair
pair
types
Y
open-loop
corresponding
by
oula
P-D
if

!
!

a
stabilization:
their
of
!
"
!
"
ﬁnite-dimensional
parameterization
structure.
Systems:
#
Q
#
Q
P-D

corresponding
!
"
!
"
P-D
systems
#
#
#
#

is
is

©
system
notations

detectable.
stabilizable.
All

“stabilizing”
(W
P-D
17
P-D
!
"
!
"
on
#
#
ei
controller
P-D
controllers
Min

!
"
!
"
stability
Lu)
#
#

for
controllers
that
2

.
stabilizes
the
system

The
Given:

where

System
Real-parameters
LFT
!%$
!%$
#
#

-system
Ya`
allowable
g
matrix

bdc

!%$
'
&

#

e
f
g
!%$
1
!
$
0
!%$

0
0

(
`
#
#
#
h
piecewise
is
)

,
i
X

b
partitioned
X
¤
described
0Vp
!
1
which
0
X
1
1
q
!
$
'
&
r

Qts
S

#
s
#
Q
s

1
0
0
0
0
Q
may
continues

X
m
X
¤
by

according
u
p
1
0
LFT
!%$

q
be

1
0
#
v
w

1
1
1
m
n

repeated,

-systems
x

!%$
U
X
18
o
Q
¤

#
k
h
i

1
1
p
!
(

1
l
!%$
)

to:
however
y
#
!
trajectories

'
&

Q

give

!%$
!%$
!%$

(
y
#
#
#

!
$
)
q
,
#
!%$
Q
rise

#
no
#
,
q
Q
#
with
restrictions

to:
satisfy:
0
!%$

c
d

state-space

#

0
!%

0
Q
!%$

#
1
on
Qts

s
1
!%$
s
e

f
#
j
X
equations
Q
.

u
!
$
!%$
#
#
#

v
where

!
#
given
v

p
7

!%$
!%$
v
!%
#
#
q
#

!
in

is

#
terms
the
6
6
6

2-input,

!

of
!%$
Q

1
#
1
1

D
!%$
LFT

is

2-output
#
#

1

t
-systems
2

S
!%$
!%$

v
v
#
#
p
6
6sr
8

1
0

0
0

transfer
!%
!%
19

1
#
#
0
0

v
v
graphically

!
p
1
0
1
1
function
7

!
!
#
#

q

0

0
6
6

#
such

0

d

!%v
that
!
$
0
Q
#
u
1
D

#

0

t
e
6
8

Robust
If
The
2)
1)
so,
satisfying
x
there

LFT
small-gain
0
p
7
0
stability
e
f
g
is
-system
exists
Hurwitz,
q
h

i

6
6
6
x
!%$
0

argument
Q
a
#
!
condition
x
constant

1
Qts
and
Q
!%$
s
i

s
#

#
Q
0
is
D
0
S
x
verify
u
w
stable
matrix
Q

t
for
~
~
~
0
x

exponential
with
if
p
6
6sr
8
LFT

r
x
v

!%$
#
1
20
p
!
-systems
1
x
x
xzy
U

¤

r
{
|
7
~
~
~
y

stability
B
q

}

of
q
q
!%$
Q
}
the
with
o
#
6
6
#
0
of
such
form

linear

!%$
w
0
that
#
u

D
!%$
time
#
x
t

varying
!
$
8
6

#

!%$
w
#
x

With
2)
1)

is
above)
There
satisfying
v
fulﬁlled.
is
1
Scaled
x

1
stable
exist
such

e
f
g

!
h
and
p
solutions
bounded
i
that
x

0
Q
there
x
#
1
Qts
0
s

s
exists
S

Q

'
&
real

x
y

u
w
,
Q

y
a
y
then
lemma
S

~
~
~
constant
x

with
{

|

r
v

the
1

21
and
1
following
(Gahinet
x
x
U

diagonal
¤

x

r
x

~
~
~
y
B
q
}
Q

y
y
x
two
y
x

matrix
o
and
(
0
such

)
statements
{
|
|
Apkarian,
that
x

(with
xzy
x
the

are
|
{
!%$
structure

#
equivalent:
1994)

!
$
#
x
give

The
Interconnection
Deﬁne
7
time
q
I
time
varying
6
6
6
invariant
`
¡
of
parameters

b

2
!
!
X
X
#
and
#
rational
D
D
'
&

©
(

LFT
©
,
)

i.e.
F
t
X
transfer
closed-loop
enter
2

'
&

68

1
0
0
0
0
¡
both
22
b
function

!
X

1
0
1
1
1
#
in
©

`

!

system

1
0
#
(
b

matrix
!
)
X
#
'

&
'
&
and

2
(

)
1

0
in

!
#

¡
1

0

!
X

#

!
1

0
0
0
0
0

1

0
1
1
1
1
(
)

7
q
q
I
F

6
6
6

6
6

`

`

b

!
!

X
X
#
#

D
D

LFT

I
F
t
t

closed-loop
68
23
7
system
q

q
I

F

6
6
6

6
6
`

`

!
b

X

#
!
X

#

D
D

I
F
t
t

68

Closed-loop
Deﬁne
Closed-loop

block
performance
The

LFT
£

uncertainty

!
!

#

Q
transfer
transfer
gain-scheduling
£
¡
Q
!
`

control
Q
¡
b
Q
!
function
`
function
X
®
¯
Q
#
`

b
°
problem
q

t
t
F
8

!
X

±³²

#
Q
`
LFT
!
´
X
in
is:
¯
control
#

¡

#
terms

¥

¥

¥

¥
¦
!
µ
X
²
±
with
§
closed-loop
#

#
´
¶
¥
¥

¨
¨
¨
ª
.
problem
the
¬
of
«
©
©
©
©

24

¥
¥

¨
¨
¨
!
!
nominal
¢
£

ª

«
ª
ª
©
ª

!

!
#

!
:

¥
¥
can
¨
¨
¨
Q
`
#
Q
ª
«
©
system
«
«
«
¡
b
¥
¥
¥
¥
plant
be

!
!
¦

X
§
#
#

#
Q
treated
¤

#
Q
`

q

q
t
I
7

!
,

¡

and
Q
!
`
X
as
#

`
with
a
!
X
classical
b
#
#
!
repeated
X
#

robust
real

Induced
from
Induced
Given
i.e.
take

to
a
W
the
W
time

1
.
1
Induced
performance
performance
“worst-case”
y
varying
¹
y

¸
!
º

!%$
vector
·
¨
#

#
!%º
¸
performance
bound
metric
c
#
(over
1
allowable

º
¼
g
½

n
!%$

all
#
¾
,
¾
¹

1
then
25
!
parameter
¼
½
g
W
y
¥
¸
¹
1

-gain
!%º
y
the

¸
#

in
¸
¸
!%º
W
1

#

1
#
¸
º
LFT

trajectories)
1
-norm
¸
!
º
f
1

#
r
-systems

º
of
»
¿

£
!%$
#
ÀÁ|
induced
is
deﬁned
W
1
by
gain

Find
Given:
2)
1)
2)
1)
controller
V
Finite
plant’
V
Finite
ector
(if
ector
¨
possible):
±³Â
s

dimensional
´ÄÃ
dimensional
of
parameter
of
’
s
integers
integers

Å
parameter
È
È
È
ª
É
É
Ç
É
©
Æ
structure.
ª
Ê
linear
Ç
©
Ç
É
É
Ç
Ç
Ç
É
linear

,
structure.
describing
describing
Æ
ª
Ç
©
LFT
©
É
É
É
¡
©
©
©

.
Æ
ª
ª
Ç
ª
ª
©
ª
synthesis

the
the

26
problem
7
q
q
I
F

6
6
6

6
6

`

`

b

!
!

X
X
#
#

D
D

I
F
t
t

68

Deﬁne
then
Let
Let
Let

!%$
U
Ë
#

x
U

Ï
Ì
¤
`
Q
denote
U
Î
Ï
and
b
y

!
Ï
Ï
q

X
}

0
Ë
#

x
.
q

Í
}
.
e
f
g
,
d
the

y
d
Î
be
U

1
¤
h

set
i
bases

y
y

0
Q
of
q

0
}
1

e
Q
matrices
LFT
q
1
Qts
for
}

s
y
s
and
0

Ï
the
1
Q

0
|
synthesis

deﬁne
null
u
w
such
e

27

spaces

0
and
and

that
setup
and
for
d
Î
Ï

of

U
1
x

e
f
g
Ï

0
x

0
1
h

i
Î
so
0
Q

Î
1
that
0
1

0
Qts
|

s
s
x
Q

e

Î

0
0
0
w
u

x

such
bound
The

that
¹
LFT
¾
,

y
if
gain-scheduling
U
¹
there
Ë
Ë
Ì
yÑÐ
Í
exist
Ð
'

&

'

&

y
Q

Î
¹
matrices:

S
y

Ï
Ï
¹

Î
S

y
Ï
0
Ï
0
p

¹

0
0
Î

control
¹
¹
¹
Î

¹
LFT

¹
Î
y
y
p
Î
¹
U

¹

problem
synthesis
ÀÁ|
yÑÐ
¾
¾
|
Î

|
Ð

Ï

¹

y
Ï
¹
Ï
Î

28
and
Ï
Q
0
Ï

y
0
Ï
0
|
|
0
0
0
p
p
is

Ï

solvable

solution

p

Ï

y
Ï
¾
¾
U
|
|

p

y

Ï

y
Î

Ï
Ï
Ï
Î

Ï

Ï
0
Ï
0
with
0

|
0
0
x
0
|
p

ÀÁ|
p
(

closed-loop
(

)

and
)

Ë
Ë

Ì

Í
|
Î
Ï
|

performance
Î
y
Ï
U
x

(3)
(2)
(1)

Some
has
Suf

ﬁcient
Robust
The
¾
satisfying
if
performance
-suboptimal
and
comments:
linear
only
condition
numerical
matrix
AMI
if

there
controller
level
0
for

!
inequalities
e
f
g

methods
LFT
Q
are
existence
h
¾

¾
i
#

matrices
:
of
|
synthesis
¹
Q
order
Q

for
0
Qts
of
are
AMI
s
solving
s
Ó
Ô

LFT
Ò
Q

ﬁnite
p
!
if
1

!
-controller
Î
solution
29
u
w
Q

y
Q

the
dimensional.
¾

¹
#
Î
|
{
|
Î
af
¹

#
n
ﬁne
and
such
summary
Ò
and
e
f
g
linear
Î
h

i
that
Î
¹
Q

matrix
Î
y
p

Î
the
0
¤
Qts
s
|
{
s
p
Î
closed
Q
inequalities
¤

Î
with
u
w
ÀÁ|
Q
structure
loop
system
exist.

Parameter

The
Implementation:
solutions
The
parameter
Ê

control
F
L

TI
,
dependent
i.e.
q
part

to
structure.
action
the
6
6
the
of
`
the
controller
af

Õ
ﬁne
2
controller:
LFT
!
is
X
#
matrix
linear
D
controller
’
s
parameter
in
t

inequalities.
LFT
©
,
I
6
nonlinear
¡
-controller
,
30
is
structure
reconstructed
in
X
.
With
is
¡
the

!
X
X

2

#
given
from
same

¡

!
(measured):
as
¡
!
and
Q
X
`
the
#
©

Î
plant’
!
,
X
#
i.e.
#
s
the

the
The
by
An
rameters
Helmersson
parameters
The
But
Results
-
-
-
-
Braatz
rate
improved
Controller
The
Pierre
Andy
note,
results
Small-Gain
synthesis
variations
on
Packard,
(upper
similar
and
Apkarian
LFT
may
X
upper
has
.
structure
Morari
synthesis
also
bound
LFT
ideas
generalized
conditions
LFT
in
Greg
bound
and
the
be
(1997).
test
have
is
extensions
on
Pascal
Becker
parameters
conservative
still
may
derived
on
Ö
been
are
a
for
Ö
the
be
LFT
for
Gahinet.
repeated
and
still
31
ideas
conservative
proposed
independently
repeated
of
AMI’
Fen
X
and
.
since
a
ﬁxed
to
s
W
real
remarks
in
exploit
u,
they
by
real
block
L
and
parameters).
TI-system.
Lu
due
do
parameters
by
the
and
diagonal
to
not
(Packard,
realness
the
Doyle
take
realness
form.
(1992,1995).
into
is
1997):
also
of
account
the
given
of
pa-
the

where
Parameter

"
tion
for
!%$
U
.
each
#
a
allowable
3
set
"
,
¤
for
,
and
3

all
$
trajectory
'
&

all

©
(
y

and
)
u
$

,
Ù
,
LPV
¤
!
"
'
&

"

-system
!%$
1
0
!Ø×
#
!
"
!
!
"
"
LPV
#
#
n
satisfy:
#
#
#

"

-systems
¤

!
$
32
1
0
0
P
#
0
0
n
!
"
!
!
"
"
on
#
Ù
#
#
Ú
¤
state-space
!
"

|

!%$
0
1
1
#
!
"
!
"
#
,
#
(
i.e.
#
)
bounded
'
&
2

(

form:
)
rate
of
varia-

The

L
Induced
gain-scheduling
yapunov
W
LPV
function
1
performance
y
¹

-systems:

y
2
Û

output
Ü

!
º
!

#

Ý
Q
!%º

feedback

bound

#
Û

Û
º
!
"
control
!
"
!
"
n
#
Q
Q

#

"
"
¾
¾
#
#
1
33

controller
y
Ý

¹

problem

Û
Û
y
!
"

!
"
!
"
Q
Q
#

!%º
"
"
Ý
#
#
#

¡

!
º
!
"

formulation
#
#

º
Û
is
©
!%$
given
#
¿

£
ÀÁ|
by:

Î
and
There

'

&

LPV
Þ
S
The
y
!
Î
a
y

Û

S

performance
exists

0
-systems:
S

controller

{
|
0
S
S
S

Þ
Þ

0
and

a
1

Û

Û
gain-scheduling
1

Û

a
1

1
S

!
Û
0
y
P-D
1
#
bound
Basic

Û
Q
!%ß
1

quadruple
#

Û
Q

¾
characterization
Q

,
Î
Û
S
whenever
!
Q

output-feedback
Î

0

S
0

!
Þ
Û
Î
Î
and

S
p
S
#

Û
can
Q
1
ß
Þ

34
p
there
Û
Î
0
Û

!
Þ
be
1
Þ
1
Q
Þ

Þ

Û
Û
reconstructed
1
Û
|
S
{
0
Q
y
Û
exist
#
Q
Þ
(Apkarian

controller
!%ß
Û
Q
Û
#
Þ
P-D

#
,
Û

Q
such
0

matrices
0
S

Û
enforcing
from:
#
that
and

ß
ß
0
¾
p
1

Adams,1998)

Û

internal
1
0

y
ß
ß
ß
{
¾
|
p
stability
(

)

and
|

LPV
where
Î
and
There

-systems:
Î
a
y
Ë
performance
exists
Ì
'
&
{
|
'
&
|
|
and
Ë
Ë
,
Í
Ì
a
such
|
|
gain-scheduling
Ë
Projected
p
p
(
Í
(
y
)
that
are
y
)
bound

bases
'

&
'

&

Î
S
¾
S
solvability
,
for
whenever

Î

y

output-feedback

y

the
0
y
0
S
0
Î
0

S

p
nullspaces

y

p
Î
there
conditions
35
Î

Î

{
|

y
¾

exist
y
p
0
¾
p
0
0
of
0
0
0
controller

d

y
P-D

y

¾
(Apkarian
p
0
0
1
¾
p
0
0
0
0
(
(
matrices
)

)

enforcing
1
'
&
0
'
&
|
|
e
Ë
and
Ë
Ì
Ì
|
|

and
p

p
d

(
y
(
internal

)
1

)
Adams,1998)

y
|
|

y
0
{
|
1
stability
e
.
and

LPV

The
posed
Existence
-systems:
controller
through
conditions
Projected
can
L
yapunov
be
Ü
!

constructed
Ý
become
Q

solvability
function:
!
"
#

#
necessary
y

from
Ý

conditions
!
"
36

#

and
Ý
and
Q

Î
with
suf
along
ﬁcient
(Apkarian

Ý
the

if
lines
quadratic

Û
and
of

(Gahinet,
Adams,1998)
stability
1994).
is
im-

The
The
Solve
Pick
The
This
ities
Induced
number
number
a
suggested
yields
(AMI).
the
basis
W
AMI’
1
a
However
performance
of
of
for
inﬁnite
s
LPV
inequalities
inequalities
at
the
solution
the
parameter
-systems:
dimensional
,
these
grid
is
can
to
grow
grow
points.
are
grid
be
dependent
dependent
tested
linearly
exponentially
Summary
the
convex
37
parameter
in
with
optimization
terms
solutions
on
and
number
the
with
of
set
two
parameters
remarks
the
3

of
.
af
!
"
problem.
number
ﬁne
grid
#
and
linear
points.
"
Î
of
.
!
"
matrix
parameters.
#
to
the
inequal-
AMI’
s.

pendence
The
ﬁnite
Then
the
If
the
parameters
controller
dimensional
it
state-space
is
suf
must
ﬁcient
becomes
"
LPV
be

,
AMI’
i.e.
matrices
removed,
to
!
"
!
"
-systems:
#
#
test
s
dependent
become

the
!
"
!
"

!
#
#
for

corner
!
"
further
#
Q
Summary
ﬁnite

on

!
"
38
points
¹
¹

"
dimensional.
#
Q
details
.

In
!
"
¹
¹
order
(Apkarian

#
Q
S
and

see
!
"
¤
à
u
to
#
0
(Apkarian
#
"
remarks
depend
be
¤
et

practically

al.
¤
¤
,
1995).

in
and
¤
¤

af
ﬁne
Adams,
valid,
Then
manner
this
1998).
the
de-
in-
on

control.
Several
Number
Applications
proaches.
Focuses
Some
which
powerful
Renewed
Summary
nonlinear
can
parallels
of
on
techniques
interest
be
papers
analysis
using
applied
on
control
between
in
and
the
“recent
gain-scheduling
and
and
to
LFT
the
problems
LMI’
theoretical
computational
gain-scheduling
focus
and
work”
s.
LPV
from
39
can
in
development
techniques
control
academia
be
gain-scheduling
schemes
solved.
control
and
is
start
LPV
rather
(interior
increasing.
and
-systems,
to
emerge.
than
model
point
control.
ad.
due
predictive
methods)
hoc.
to
new
ap-

Balas,
multipliers,
Balakrishnan,
ãtäå
ãtäå
ãtäå
ãtäå
¡
the
Apkarian,
systems:
Apkarian,
T
Apkarian,
T
Apkarian,
33
Apkarian,
å
õ
r
r

ansactions
ansactions
¥
ì
(4):
ü
33rd
õ
©
æ
ç
æ
ç
æ
ç
æ
ç
G.
è
é
è
é
è
é
è
é
÷
655–661.

Conf
J.,
¢
!
êëì
êëì
êëì
êëì
a
å
P
P
P
P
P
Proc.
ü
ý
.,
design
.,
.
.
.
ì
ì
Fialho,
ì
ì
and
and
(1997).
V
í
í
erence
Gahinet,
Gahinet,
on
on
í
í
ý
ú
ø
ù
ù
ú
ø
ù
ô
.,
§
§
î
î
A
Control
Gahinet,
Adams,
of
Huang,
ï
ð
ñ
ï
utomatic

ö
ï
example,
I.,
the
ì
ó
ì

ó

ý
ý
ò
on
On
é
ò
Packard,
P
P
Amer
æ
ù
õ
æ
ù
.
.
ù
õ
æ
Decision
Systems
û
Y
ù
ø
æ
ô
õ
ö
¢
and
and
the
R.
ý
.,
ù
P
é
é
Control
ç
.
Packard,
þ
J.
ican
ø
÷
A
(1995).
discretization
ø
÷øùú
Biannic,
Becker
ê
ú
ø
ù

ç

utomatica
ç

(1998).
A.,
ô
ø
ú
Control
ø
ì
ê

and
T
ø
ì
ê

û
40
echnology
Renfrow
õ
æ
û
,
§
ø
§
ç
ûüýþ
A.
î
G.
(5):
A
î
Control
J.-M.
û
"
ý
convex
Advanced
ï
ö
ø
ï
ð
ø
and
ý
æ
ù
ï
ù
31
(1995).
ì
¥
Conf
ý
ì
853–864.
ö
ï
of
æ
ù
ò
ú
(9):
ä
å
þ
,
ò
å
ä
(1994).
J.
Doyle,
#
ê

LMI-synthesized
ö
,
ú
6
erence
ö
Lake
ç
ø
and
þ
1251–1261.
characterization
(1):
ô
õ

¢
ú
ç
ì
ÿ
Self-scheduled
References
ø
û
ý

gain-scheduling
õ
æ
¢
ü
ý
21–32.
ù
ö
Mullaney
J.
ø
ÿ
ö
ç
õ
Buena
Self-scheduled
÷
,
ú
40
Baltimore,
(1994).
û
ý

ö
é

ÿ
ø

ù
æ
§
ñ
þ
ê
õ
ì
ø
û
ö
û
ö
ñ
ù
ç
æ
V
û
æ
ý

,
ô

ista,
ü
æ
ñ
linear
C.
Linear
ú
þ
©
ö
õ
ä
æ
ç
ê
ú

of
ä
ö
á
(1997).
USA,
ü
ù
¡
pp.
å
gain-scheduled
techniques
â
ý

å
¡
á
ñ
ø
¡
ö
parameter-varying
ö
æ
ç
ö
matrix
ù
3312–3317.
â
control
ì
ú
pp.
ü

¢
ø
ç
ç
ï
ü
control
On
¡
õ
û
1228–1232.
ö
û
ý
ü
þ

inequalities
ù
the
÷
of
ä
ê
for
ø
ý
linear
ö
of
design

ö
æ
å
ç
¢
ù
uncertain
õ
å
å
æ
a
á
é
missile
â
ø
parameter-varying
ê

ê
ý
þ
ì
û
controllers,
controllers,
of
æ
ö
in
ç
û
þ
ø
§
ì
LPV
analysis
ö
systems,
£
¤
via
¡
ÿ
î
¦¥
controllers
¡
§
æ
LMIs,
ì
ô
õ
ò
¨§
ñ
ð
IEEE
A
with
©
©
§
IEEE
utomatica

æ
Proc.
ô
ö

¡
ù
¡
for
ð
©
of
ì
ÿõ
ç

ù
õ
þ
õ
ö
ù
$
þ
õ
å
ý

Amer
Cardello,
the
Breedijk,
Inter
Braatz,
Control
Bequette,
thesis,
Becker
pp.
A
Becker
parametrically-dependent
Becker
T
Becker
multiple
Banerjee,
New
the
r
single
iennial
Amer
2795–2799.
F-14
national
ican
Mexico,
,
University
,
,
,
,
R.
pp.
G.
G.,
G.
G.
quadratic
linear
ican
R.
T
W
aircraft
Control
.,
D.
B.
A.,
S.
and
or
(1996).
Packard,
and
Edgar
1–28.
Jour
W
Control
and
ld
(1993).
Arkun,
pp.
models,
.
Packard,
Cong
(1997).
San,
lateral-directional
nal
Conf
,
Morari,
of
123–127.
lyapunov
T
Additional
California
.
of
ress
Y
A.,
K.-Y
Conf
F
Quadr
.,
erence
.
Proc.
Rob
and
Pearson,
Gain-scheduled
Philbrick,
A.
,
M.
.
linear
erence
San
(1987).
ust
T
(1997).
atic
approach,
(1994).
of
,
rachtenberg,
at
results
pp.
and
Francisco,
3rd
Berkeley
Stability
feedback,
,
R.
682–686.
Baltimore,
D.
axis
Application
Nonlinear
European
Robust
and
On
and
on
Proc.
during
the
parameter-dependent
process
and
Ogunnaike,
.
Balas,
USA,
Systems
I.
performance
stability
USA,
(1994).
of
P
Control
Control
powered
erf
of
the
pp.
G.
control:
or
gain
pp.
mance
Amer
(1993).
&
351–356.
of
Model-based
B.
7
Control
Conf
41
scheduling
887–891.
:
systems
approach,
(1995).
105–1
ican
of
A
of
erence
review
Control
linear
Linear
Control
Letters
12.
controllers
with
%
parametrically
,
to
,
control
Rome,
&
Proc.
NA
of
the
mixed
P
control
ar
Conf
23
T
parametrically-dependent
O
ameter
:
of
control
205–215.
ASI
erence
for
Italy
of
the
time-varying
of
rapid
LPV
Nonlinear
,
Amer
nonlinear
Dependent
pp.
of
varying
,
batch
San
systems,
thermal
2671–2676.
ican
Francisco,
parameters,
bioreactors,
Model
systems
processes
Control
processes,
Systems
Proc.
Based
linear
Conf
using
USA,
of
,
IF
using
Proc.
PhD
erence
A
Proc.
Process
systems:
C
13th
of
,
of

Goodwin,
parametric
Gahinet,
Rob
Gahinet,
(¨)
a
Fossen,
parameter-dependent
Feron,
Amer
Fei,
Control
Engell,
Conf
El-Zobaidi,
Driankov
Decision
Doyle,
T
Chilali,
2
b
r
ansactions
4
0
ust
1¨2
S.
erence
0
ican
3
and
3
4
E.,
and
J.,
Conf
S.
M.
T
P
P
,
G.
5
and
.
.,
.
0
6
7
Apkarian,
Control
and
D.,
Packard,
and
uncertainty
and
and
Huo,
H.
Apkarian,
Nonlinear
erence
on
C.,
Palm,
Control
on
8
9
@
Klatt,
M.
Gahinet,
Grøvlen,
Decision
Apkarian,
Graebe,
A
W
8
A
B
H.
Conf
utomatic
.
7
,
(1995).
P
San
K.
and
R.
A.
C
,
,
P
Control
lyapunov
.
Brighton,
and
IEEE
and
erence
and
U.
.
D
E
P
and
and
Jaimoukha,
S.
Francisco,
.
P
F
(1993).
.
D
E
Zhou,
(1996).
Gahinet,
Rehfuess,
Control
F
(1994).
Robust
.
T
Chilali,
A
G
Control
,
and
r
1
A
H
ansactions
Seattle,
:
functions,
England,
421–448.
B
K.
I
Nonlinear
Levine,
41
%
%
(1991).
M.
A
P
Q
P
USA,
,
7
I.
.
Kobe,
(3):
&
&
U.
linear
P
M.
(1996).
USA,
(1996).
design
8
E
control
(1996).
pp.
358–367.
7
W
on
IEEE
(1996).
pp.
C
7
9
Japan,
Review
.
control
A
matrix
pp.
1227–1232.
S.
H
2941–2945.
utomatic
with
Af
G
Analysis
T
for
(1993).
r
2389–2390.
A
ﬁne
R
ansactions
B
Robust
inequality
8
nonlinear
pp.
T
pole
of
akagi-Sugeno
of
42
parameter-dependent
@
S
a
LFT
Control
B
3982–3983.
and
non-minimum-phase
placement
Internal
P
9
normalised
9
s,
7
R
synthesis
LMIs,
on
systems
approach
F
@
41
model
T
A
E
@
7
utomatic
(3):
fuzzy
and
constraints
A
H
8
LPV
436–443.
with
of
'
control
to
T
U
gain-scheduler
,
@
robust
Proc.
%
Control
time-varying
gain
lyapunov
F
&
T
CSTR,
D
of
control,
@
scheduling,
:
T
control
of
An
linear
8
41
the
VW
LMI
A
(7):
Proc.
functions
7
E
30th
Inter
P
systems
systems
uncertainty
,
approach,
IEEE
@
B
1041–1047.
C
7
of
Proc.
Conf
national
9
the
pp.
I
and
A
with
T
erence
via
Amer
of
W
C
,
1053–1059.
IEEE
A
Proc.
the
real
Jour
saturating
B
C
ican
I
35th
on
nal
P
Q
of
E
T
of
A
the
F
@
F
8
V
XY
`
`
`

implementation
Kaminer
of
Kamen,
scheduled
Johansen,
T
Hyde,
Control
Helmersson,
Control
Helmersson,
of
Helmersson,
3rd
H
mixing
H
antiwindup
Graebe,
actuators,
r
ansactions
agglund,
¨
aggblom,
¨
parameters,
Electrical
European
R.
,
tank,
,
Conf
E.
Kobe,
S.
I.,
A.
W
control
T
T
Proc.
K.
controller
F
Pascoal,
.
and
erence
Engineering.
.
Ind.
.
.
A.,
on
A.
A.
A.
and
and
Control
E.
and
IEEE
Japan,
of
Hunt,
A
(1996).
(1995b).
(1995a).
Glover
utomatic
T
Eng.
(1993).
of
gain-scheduled
with
Ahlen,
Khargonekar
engvall,
,
European
A.
T
Rome,
design,
r
ansactions
application
K.
pp.
Conf
,
Chem.
M.,
K.
Application
J.,
'
Methods
Experimental
A.
1666–1671.
Control
(1993).
erence
Khargonekar
Italy
synthesis
A.
Gawthrop,
L.
IEEE
Control
(1995).
Res
B.
,
,
P
pp.
(1996).
controllers,
on
.
,
to
.
P
38
f
The
or
Rome,
T
of
pp.
r
.
vehicle
A
(7):
153–158.
ansactions
(1984).
and
Conf
Rob
utomatic
real-
An
P
2653–2661.
comparison
application
,
.
1021–1039.
P
J.
automatic
Dynamic
LFT
ust
Italy
erence
.
'
P
speed
and
gain
.
A
On
Gain
and
gain
,
utomatica
Control
pp.
Fritz,
on
,
the
scheduling,
pp.
Coleman,
control,
of
Scheduling
scheduling
2450–2455.
of
Control
transfer
tuning
control
scheduled
H.
conventional
1072–1077.
43
AC-29
(1998).
31
procedure
unkno
Systems
among
(8):
of
E.
(1):
Proc.
,
linear
with
PhD
1
E.
185–1
wn
Of
%
25–33.
(1995).
&
and
alternative
f-equilibrium
mixed
.
of
thesis,
T
systems
controllers
for
echnology
the
191.
nonlinear
unsymmetrical
35th
uncertainties,
A
Linkøping
velocity
controllers
whose
Conf
linearisation
to
4
model-based
(1):
a
erence
VST
algorithm
coef
92–99.
University
Proc.
processes,
and
OL
ﬁcients
on
and
aircraft,
its
Decision
of
for
control
,
relation
3rd
are
design
Sweden,
the
Proc.
functions
IEEE
European
of
and
to
of
a
of
Dep.
gain
the

Ling,
uncertainty
Lind,
Science
Lin,
A
Lawrence,
varying
Lall,
Chem.
Lagerberg,
School
Lagerberg,
control,
Klatt,
stirred
Klatt,
regulable
Khargonekar
European
Khalil,
utomatica
J.-Y
S.
C.
R.,
K.-U.
K.-U.
tank
H.
and
.
systems,
and
Eng.
of
Ind.
Balas,
48
and
Electrical
realizations
K.
Control
,
(18):
D.
Proc.
31
Glover
A.
A.
and
reactors,
and
Edgar
T
Eng.
(1993).
,
Y
(3):
A.
echnol.
and
(1996).
P
u,
.
G.
3159–3171.
Engell,
Engell,
and
P
C.-C.
Proc.
of
381–390.
,
J.
,
Chem.
.
Breitholtz,
Conf
and
T
K.
and
.
the
and
Rugh,
Robustness
Proc.
F
(1995).
20
of
Gain
.
(1993).
of
S.
S.
Sontag,
erence
(1992).
Amer
:
Computer
linear
Packard,
the
435–444.
Res
(1996).
(1995).
of
W
scheduling
ican
.
C.
Amer
.
3rd
,
J.
Robust
systems
pp.
pp.
Automatic
E.
(1997).
A
(1995).
issues
European
new
Control
Eng.
A.
351
Nonlinear
Gain
D.
ican
58–62.
(1982).
(1995).
performance
1–3518.
control
fuzzy
Control
A
Chalmers
scheduling
over
in
Gain
Conf
tuning
study
output
Control
control
rings,
gain
On
Optimal
erence
scheduling
and
Conf
of
and
the
feedback
scheduling
University
IEEE
gain
its
trajectory
and
of
Conf
erence
gain
relation
,
application
44
scaled
neutralization
Seattle,
adaptation
scheduling
erence
T
r
scheduling
dynamic
ansactions
,
control
Seattle,
of
control
between
algorithm
%
USA,
&
T
,
ech,
Rome,
to
full
of
linear
using
control
a
USA,
processes
of
pp.
information
for
Sweden.
feedback
chemical
on
stable
neutralization
for
Italy
3463–3467.
pH
A
controllers
receding
utomatic
process
pp.
applied
control,
,
matrix
pp.
2384–2388.
reactor
linearizable
by
2665–2670.
synthesis
horizon
to
gain-scheduling
control,
Control
fraction
Chemical
for
an
processes
model,
a
exothermic
nonlinear
%
Proc.
AC-27
systems,
with
factorizations
&
Engineer
T
control
echnical
real
in
of
(3):
continuous
trajectory
plant,
the
CSTR,
parametric
Proc.
627–638.
of
ing
repor
time
and
of
t
,

T
Schei,
and
Russell,
Rugh,
heat
Puruchuri,
and
Packard,
Packard,
example,
Nichols,
of
Muske,
W
Megretski,
T
Lu,
Amer
Lu,
Amer
r
r
iennial
or
ansactions
a
process
Control
crude
ld
W
W
exchanger
.-M.
ican
.-M.
ican
T
W
Cong
.
K.
W
E.
.
R.
S.
J.
A.
A.
IEEE
or
V
oil
R.,
and
Control
and
Control
L.
A.
(1993).
,
A.,
(1991).
.
ress
and
(1994).
ld
constraints,
Kobe,
P
and
distillation
on
Logue,
(1996).
.
Reichert,
Doyle,
Doyle,
Cong
,
and
T
,
Proc.
Kantner
r
A
ansactions
San
utomatic
Braatz,
Conf
Conf
Japan,
Automatic
ress
Analytical
Rhinehart,
Gain
D.
Francisco,
J.
J.
c
of
erence
erence
,
unit,
d
C.
C.
R.
A.,
,
Presented
the
Sydney
R.
M.
bibo
scheduling
pp.
Control
(1995).
(1994).
T
Keaton,
D.
Amer
(1996).
on
.
unpub
and
,
,
tuning
framework
3938–3941.
output
Baltimore,
pp.
R.
(1996).
Control
USA,
,
ican
R.
Rugh,
Australia,
lished
40
2284–2290.
%
%
at
M.
(1994).
Gain
of
via
feedback
(9):
&
&
the
Control
pp.
simple
Analysis
Systems
M.
control
control
linear
W
1668–1675.
1996
for
pp.
USA,
scheduling
.
and
435–440.
J.
pp.
gain
Experimental
1–12.
(1993).
decouplers
Conf
fractional
Hayward,
stabilization
of
AIChE
pp.
of
65–71.
of
T
scheduling,
echnology
nonlinear
nonlinear
large
2098–2102.
erence
the
45
meeting,
Gain
transformations,
scale
R.
LPV
in
,
demonstration
with
Baltimore,
scheduling
A.
systems:
systems:
multivariable
1
IEEE
way
(2):
systems
(1991).
Chicago
saturated
,
69–79.
Control
Proc.
A
A
USA,
Gain
for
with
convex
convex
,
USA
of
of
Systems
control,
control
H-inﬁnity
System
multiple
pp.
nonlinear
the
scheduled
,
pp.
characterization,
characterization,
3533–3537.
35th
systems,
Proc.
&
1–16.
Magazine
controllers:
time
Conf
Control
model-based-control
model
of
delay
erence
IF
Proc.
Letters
A
pp.
predictive
C
uncertainties
A
13th
IEEE
Proc.
on
of
79–84.
ﬂight
IF
Decision
22
T
A
:
r
C
iennial
of
79–92.
control
control
the
12th
of
a

Sugie,
thermal
Stuber
control,
Shue,
varying
Shamma,
autopilot
Shamma,
Control
Shamma,
parameter-varying
Shamma,
T
Shamma,
scheduling
Shahruz,
anti-windup:
Seron,
IEEE
Scherer
r
ansactions
T
,
S.
T
r
M.
.
processes,
J.
Jour
transformations,
Systems
ansactions
,
and
P
design,
C.,
S.
D.,
.,
J.
J.
J.
J.
J.
M.,
technique,
nal
Sawan,
S.
S.
S.
S.
S.
M.
Gahinet,
Kawanishi,
T
on
A
Graebe,
rachtenberg,
and
and
and
and
and
uniﬁed
and
of
A
Proc.
Magazine
utomatic
Guidance
plants,
on
Proc.
Cloutier
Cloutier
Athans,
Athans,
Athans,
Behtash,
M.
P
Jour
review
A
of
S.
.
E.
utomatic
and
Jour
M.
of
Amer
F
and
A
Control
nal
.
,
,
pp.
utomatica
(1995).
the
I.,
,
M.
M.
M.
and
J.
J.
Control
nal
,
Mahmoud,
S.
Proc.
Edgar
Rokhsaz,
of
R.
ican
R.
101–107.
(1992).
(1991).
(1990).
33rd
(1992).
of
Goodwin,
Mathematical
Control
(1993).
(1992).
Guidance
35
Control
of
'
,
Conf
and
(8):
T
27
analysis/synthesi
.
the
Gain
Guaranteed
Analysis
C.
F
(3):
K.
Design
42
.,
898–907.
erence
Dynamics
Gain-scheduled
A
Amer
G.
Elliott,
(7):
(1997).
(1997).
Conf
linear
559–564.
,
scheduling:
C.
Control,
Analysis
896–91
erence
of
of
ican
(1994).
on
J.
parameter
gain
controllers
Multiobjective
Mixed
Decision
K.
20
properties
Control
46
(4):
and
1.
and
,
s
scheduled
and
pp.
Potential
All
based
699–706.
%
missile
Dynamics
Breedijk,
stabilizing
d
1317–1321.
Applications
Conf
and
/
varying
for
%
of
on
&
gain
linear
erence
output-feedback
Control
autopilot
hazards
control
exact
method
T
16
approach
.
scheduled
controllers,
(1994).
parameter-varying
(2):
,
expression
,
168
Baltimore,
for
Lake
suitable
design
and
256–263.
:
nonlinear
195–217.
Model-based
to
possible
Buena
control
gain
feedback
control
using
for
USA,
of
physical
V
gain
scheduled
plants,
ista,
linear
remedies,
for
via
systems
pp.
scheduled
linear
linearization
pp.
control
LMI
1685–1689.
parameter
IEEE
parameter
79–85.
missile
optimization,
IEEE
by
of
the
rapid
aircraft
and
gain

Control
Y
bounded
W
parameter
W
Conf
W
Conf
W
Proc.
W
W
pp.
Whatley
A
W
Mech
W
T
variations,
ang,
utomatic
etley
u,
u,
u,
u,
u,
u,
ang,
alker
75–78.
F
F
erence
F
erence
F
F
F
.,
.,
.,
.
X.
of
.
.
Eng
,
and
and
(1995).
J.
,
B.
Conf
Y
Y
Packard,
,
D.
H.,
ang,
ang,
the
parameter
M.
and
rs
and
variation
Control
J.
Proc.
,
Packard,
Packard,
erence
on
W
Amer
J.
Seattle,
21
(1997).
X.
X.
Rugh,
Ulf
u,
and
Decision
Control
1
:
H.,
H.,
F
A.
263–268.
f,
of
ican
.
AC-32
A.
,
Pott,
and
variation
Packard,
rates,
Packard,
and
3rd
Seattle,
USA,
A.
A.
W
Gain-scheduled
(1989).
Control
.
of
Packard,
(1995b).
(1995a).
J.
European
&
Balas,
D.
(10):
Linear
Proc.
(1987).
Control
pp.
C.
USA,
rates,
A.
A.
G.
4440–4444.
Conf
(1984).
935–940.
Gain
and
of
and
P
A.
LQG
Optimal
,
(1995).
ar
Control
Feedback
pp.
the
New
erence
scheduling
Submitted
(1995).
ameter
Becker
Becker
ﬂight
3371–3372.
Amer
control
Adaptive
Orleans,
LQG
LPV
,
Conf
Seattle,
,
Adaptive
ican
,
V
control
G.
G.
ar
linearization
design
to
control
performance
for
erence
ying
(1996).
(1995).
gain
Int.
Control
USA,
process
via
USA,
Systems
Jour
improves
for
,
control
two-degree-of-freedom
47
design
Rome,
pp.
Induced
Induced
Conf
LPV
nal
pp.
families
control,
188–193.
of
,
of
of
erence
systems,
4435–4439.
for
PhD
Italy
linear
reactor
full
Nonlinear
c
c
pitch-axis
for
,
d
d
thesis,
C
information
pp.
-norm
-norm
,
uncertain
&
nonlinear
Seattle,
control,
Proc.
I
159–164.
pp.
and
University
control
control
missile
83–85.
of
Rob
USA,
Hydrocarbon
%
problem,
systems
systems,
the
&
for
for
ust
autopilots,
optimization,
Amer
pp.
of
LPV
LPV
Control
California
2379–2383.
using
Proc.
ican
IEEE
systems
system
Processing
.
Proc.
Control
state-feedback,
of
T
r
ansactions
Proc
the
at
with
with
of
Berkeley
Amer
Instn
bounded
the
ican
34th
on
.