TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 15, No. 2, August 201
5, pp. 259 ~
269
DOI: 10.115
9
1
/telkomni
ka.
v
15i2.809
9
259
Re
cei
v
ed Ma
y 13, 201
5; Revi
sed
Jun
e
28, 2015; Accepted July 1
4
,
2015
Aircraft Control System
Using Model Predictive
Controller
Laban
e
Chrif*
1
, Zemalache Meguen
n
i Kadd
a
2
1
Laba
ne chrif
Univers
i
t
y
of said
a, Dep
a
rtm
ent of electrote
c
hnic,Sa
ida
20
000, Alg
e
ri
a
2
LDEE Lab
orat
or
y
UST
O, MB, Departme
n
t of Automatic, Oran 31
00
0, Alge
ria
*Corres
p
o
ndi
n
g
author, em
ail
:
c_laba
ne@
ho
tmail.fr
A
b
st
r
a
ct
T
h
is pap
er co
ncerns the a
p
p
licat
i
on of mode
l-bas
ed pr
edictiv
e c
ontro
l to the long
itudi
nal a
n
d
latera
l
mo
de
o
f
an a
i
rcraft in
a terra
in fo
llo
w
i
ng task. T
h
e
pred
ictive c
o
n
t
rol ap
pro
a
ch
w
a
s base
d
o
n
a
qua
dratic c
o
st function
an
d
a li
near
state space pre
d
icti
on mo
de
l
w
i
th
inp
u
t an
d stat
e constra
i
nts.
T
h
e
opti
m
a
l
contro
l
w
a
s obtain
ed
as the
sol
u
tio
n
of a qua
dratic
progr
a
m
min
g
p
r
obl
em
defi
ned
over a rec
edi
n
g
hori
z
o
n
. Cl
ose
d
-lo
op si
mul
a
ti
ons w
e
re c
a
rri
ed o
u
t by
usin
g the
lin
ear
air
c
raft mo
del. T
h
is pr
oject th
e
s
is
provi
des a bri
e
f overview
of Mode
l
Predicti
v
e Contro
l (MPC).A brief his
t
ory of industri
a
l mod
e
l pre
d
i
c
tive
control tech
nol
ogy has b
e
e
n
prese
n
ted first follow
ed by
a s
o
me conc
epts l
i
ke the rece
din
g
hori
z
o
n
, mov
e
s
etc. w
h
ich for
m
th
e
basis
of
the MPC.
It follow
s
th
e Opt
i
mi
z
a
tio
n
prob
l
e
m w
h
ich
ulti
mate
ly l
e
a
d
s t
o
the
descri
p
tion
of the Dyn
a
m
ic M
a
trix Contro
l (D
MC).T
he MP
C prese
n
ted i
n
th
is report is b
a
s
ed on
DMC. After
this the
app
lic
ation s
u
mmary
and
t
he l
i
m
ita
t
ions of th
e ex
isting tec
h
n
o
lo
gy has
be
en
d
i
scusse
d a
nd t
h
e
n
e
x
t g
e
n
e
r
a
t
ion
MPC
,
wi
th
a
n
em
ph
a
s
i
s
o
n
p
o
t
e
n
t
i
a
l
bu
si
ne
ss a
n
d
re
se
a
r
ch
o
p
p
o
r
tu
n
i
ti
e
s
ha
s been
review
ed. F
i
nal
ly in
the
last
p
a
rt w
e
ge
ner
ate Matl
ab c
o
d
e
to i
m
pl
e
m
e
n
t b
a
sic
mo
de
l pr
e
d
ictive
contro
ll
e
r
and i
n
trod
uce
nois
e
into the
mo
de
l.
Ke
y
w
ord:
aircraft moti
on
, flight contro
l, lateral a
n
d
l
ong
itudi
na
l stability, mod
e
l pre
d
ictive
control
,
opti
m
i
z
at
ion
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Origin
ally de
veloped to
meet the
sp
ecia
li
ze
d co
ntrol ne
ed
s
of power pl
ants a
nd
petrole
um
ref
i
nerie
s, MP
C techn
o
logy
can no
w b
e
fo
und in
a
wid
e
variety of a
pplication a
r
e
a
s
inclu
d
ing ch
e
m
ical
s,
food
pro
c
e
ssi
ng, automot
ive, and aerospa
ce appli
c
atio
ns
it
s
rea
s
on
for
su
ccess i
s
m
any, like
it ha
ndle
s
multiva
r
iable
c
ontrol
probl
em
s n
a
turally. But th
e mo
st imp
o
rt
ant
rea
s
on fo
r its su
ccess is its ability to handle con
s
trai
nt
s. Model pred
ictive control (MPC) refe
rs
to
a class of
computer
cont
rol algorithm
s
that utilize an explicit
process model to
predi
ct the fut
u
re
respon
se
of a
plant. At each co
ntrol i
n
terval an MP
C algorithm
attempts
to
optimiz
e
future plant
behavio
ur by comp
uting a
seq
uen
ce of future
ma
nipul
ated variabl
e adju
s
tments.
The first inp
u
t
in the optim
a
l
seq
uen
ce i
s
then
se
nt i
n
to t
he pla
n
t, and the
enti
r
e
cal
c
ulation
is repe
ated
at
sub
s
e
que
nt control inte
rval
s. The
basi
c
MPC cont
roll
er can b
e
de
sign
ed
with p
r
ope
r rest
ricti
ons
on the predi
ction hori
z
o
n
and mo
del le
ngth. The p
r
edictio
n ho
rizon ha
s to be
kept sufficie
n
tly
large
r
tha
n
control ho
ri
zon. But after applying to
many othe
r appli
c
ation
s
we find
as the
compl
e
xity incre
a
ses then
we
ne
ed te
chni
que
s
oth
e
r th
an
DM
C li
ke
gen
eralize
d
p
r
edi
ct
ive
control (GPC) which
are
better. In mo
dern
proce
ssing pla
n
ts th
e MPC
cont
roller i
s
pa
rt
of a
multi-level hi
era
r
chy of co
ntrol fun
c
tion
s. It is o
ften
difficult to tra
n
slate th
e co
ntrol requi
re
ments
at this level i
n
to an ap
pro
p
riate
conve
n
tional
control stru
ctu
r
e. In the MPC
methodol
ogy
this
combi
nation
of blocks i
s
re
plac
ed by a si
ngle MPC
co
ntrolle
r.
1.1
.
Aircraft Con
t
rol
and Mov
e
ment
There are th
ree prim
ary ways for an ai
rcraft to cha
nge its o
r
ient
ation relative
to the
passin
g
ai
r.
Pitc
h
(m
ove
m
ent of the
nose up
or d
o
wn
),
Roll
(rotation a
r
ou
n
d
the lo
ngitu
dinal
axis, that is,
the axis
whi
c
h ru
ns
alon
g
the len
g
th o
f
the aircraft) and
Ya
w
(m
ovement of t
he
nose to left o
r
right
). Tu
rnin
g the
aircraft
(ch
ang
e of
h
eadin
g
)
req
u
i
r
es the
aircra
ft firstly to roll
to
achi
eve an a
ngle of ban
k; when the d
e
sired c
han
g
e
of heading
has bee
n a
c
compli
she
d
the
aircraft mu
st again b
e
rolle
d in the oppo
site dire
ct
ion
to redu
ce the
angle of ba
nk to zero [7].
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 259 –
269
260
Figure 1. Axis of aircraft movements
1.2
.
Flight Dy
namics
The scien
c
e
of air vehicle
orientation a
nd c
ont
rol there is three d
i
mensi
o
n
s
. The three
critical flight dynamics pa
rame
te
rs
are
the angle
s
of rotation in
three dime
n
s
ion
s
ab
out the
vehicle'
s ce
nter
of
ma
ss, known
a
s
pit
c
h
,
roll
and
yaw
(quite
different from the
i
r u
s
e
as Tait
-
Bryan a
ngle
s
). Ae
ro
sp
ace en
gine
ers develo
p
co
nt
rol
sy
st
em
s f
o
r a
vehi
cle'
s orie
ntation
(attitude) a
b
o
u
t its cente
r
of mass. The
control
sy
ste
m
s incl
ude a
c
tuators, which exert forces in
variou
s dire
ct
ions, an
d gen
erate rotation
al force
s
o
r
moment
s abo
ut the aerody
namic
cente
r
of
the aircraft, and thus rotate the aircraft in pitch, ro
ll, or
yaw. For example, a pit
c
hi
ng mom
e
nt is a
vertical force
applied at a distan
ce [1, 6], Ro
ll, p
i
tch and yaw refer to rot
a
tions ab
out
the
respective axes
starting from a defined
equilibrium
st
ate. The equil
i
brium
roll angle is
known
as
wing
s l
e
vel o
r
zero b
a
n
k
a
ngle, e
quival
ent to a
level
heeli
n
g
angl
e on
a
shi
p
.
Yaw i
s
kn
own a
s
"headin
g
". The equilib
rium
pitch an
gle in
subma
r
in
e a
nd airship p
a
rlance is kno
w
n as "trim".
1.3. Longitu
dinal Modes
Oscillating m
o
tions
can b
e
descri
bed by
two param
et
ers, the p
e
rio
d
of time req
u
ired fo
r
one
compl
e
te
oscillation, a
nd the time required to
da
mp to half-a
m
plitude, or t
he time to do
uble
the amplitud
e
for a dyn
a
mi
cally un
stabl
e
motion. Th
e
l
ongitudi
nal m
o
tion con
s
ist
s
of two di
stinct
oscillation
s, a long-pe
riod
oscill
ation called
a ph
ug
oid mode a
n
d
a sho
r
t-p
e
r
iod o
scill
atio
n
referred to a
s
the sho
r
t-pe
riod mode [7].
1.3.1. Phugoid Oscillations
The lo
nge
r
p
e
riod
mo
de,
called th
e "ph
u
goid m
ode"
is the o
n
e i
n
which
there i
s
a large-
amplitude va
riation of
air-sp
eed, pit
c
h
angle, a
n
d
altitude, bu
t almost n
o
angle
-
of-attack
variation. T
h
e ph
ugoi
d o
s
cillation
is re
ally a
slo
w
in
terch
a
ng
e of
kineti
c
e
nerg
y
(velo
c
ity) a
n
d
potential
e
n
e
r
gy
(h
eight) about some equilib
rium
e
ner
gy level
a
s
the ai
rcraft
attempts to
re-
establi
s
h
the
equilib
rium l
e
vel-flight con
d
ition fr
o
m
which
it ha
d b
e
en di
sturbed.
The m
o
tion i
s
so
slo
w
that the
effects of ine
r
tia force
s
a
n
d
dampi
n
g
forces a
r
e very lo
w. Although t
h
e dam
ping i
s
very we
ak, t
he pe
riod
is
so lo
ng that
the pilot u
s
u
a
lly co
rre
cts
for this
motio
n
witho
u
t bei
ng
aware that the oscillation even exists. Typi
cally the period is 2
0–60 seconds [3, 11].
1.3.2. Phugoid Short peri
od oscillation
With no
spe
c
i
a
l name, the
sho
r
ter
peri
o
d mode
i
s
cal
l
ed sim
p
ly the "sho
rt-pe
r
iod
mode".
The sho
r
t-pe
riod m
ode i
s
a usually h
eavily dam
pe
d oscillation
with a p
e
rio
d
of only a fe
w
se
con
d
s. Th
e
motion is
a rapid pit
c
hing
of the ai
rcraft about the
ce
nter of g
r
avity. The peri
o
d
is
so short that
the spe
ed do
es not have ti
me to ch
a
nge
, so the oscill
ation is e
s
se
ntially an ang
le
-
of-attack vari
ation. The time to damp the amplit
ud
e to one-h
alf of its value
is usually on
the
orde
r of
1
se
con
d
. Ability to qui
ckly
self
damp
when
the sti
ck i
s
briefly displa
ce
d is
one
of th
e
many crite
r
ia
for gene
ral ai
rcraft ce
rtifica
t
ion [4].
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Aircraft Control System
Usi
ng Model Pre
d
ictive Co
ntro
ller (L
aba
ne
Chrif
)
261
2.
Aircra
ft Lo
n
g
itudinal D
y
namics
Figure 2. Aerodynami
c
ref
e
ren
c
e
2.1. Equatio
n
s of Mov
e
ments
Equation
s
of the moveme
nt are gove
r
ne
d by the equa
tions of me
ch
anics
∑
∑
(1)
Φ
0
(2)
Longitu
dinal
equatio
ns
ca
n be re
written
as:
Θ
∆
X
Θ
∆
Γ
Γ
Γ
Θ
Γ
∆
(3)
With:
∆
∆
∆
Γ
Γ
(4)
Re
write in sta
t
e spa
c
e form
as:
Since
0
in this mode, then
0
and ca
n elimin
ate the X force equatio
n:
Θ
Γ
Γ
Θ
01
0
∆
∆
M
0
(5)
Θ
Θ
01
0
∆
∆
M
0
(6)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 259 –
269
262
The tran
sfer
function can
be rep
r
e
s
ent
ed in stat
e-space form an
d output
equ
ation as stat
e b
y
equatio
n:
0.3149
235.89280
0.0034
0.42820
0
1
0
5.5079
0.0021
0
(7)
001
+
0
Figure 3. Ope
n
loop Impul
se Re
spo
n
se (Pitch angle
)
,
,
,
and
rep
r
e
s
e
n
t flight path angle, with
,
The input (el
e
vator defle
ction angl
e,
) wi
ll be 0.2 rad
(11 de
gree
s),
and the outp
u
t is
the pitch an
gl
e (theta).
,
,
,
and
rep
r
e
s
e
n
t flight path angle, with
,
The input (el
e
vator defle
ction angl
e,
) wi
ll be 0.2 rad
(11 de
gree
s),
and the outp
u
t is
the pitch an
gl
e (theta).
There are th
ree type
s of
possibl
e lateral-di
re
ctional dynamic motion:
roll sub
s
i
den
ce
mode, Dut
c
h
roll mod
e
, an
d spiral mod
e
.
Figure 4. Defi
nition of force
s
, moment
s a
nd angl
es
3.
Aircra
ft La
te
ral D
y
namic
s
There are th
ree type
s of
possibl
e lateral-di
re
ctional dynamic motion:
roll sub
s
i
den
ce
mode, Dut
c
h
roll mod
e
, an
d spiral mod
e
.
I
m
pul
s
e
R
e
s
pons
e
Ti
m
e
(
s
e
c
)
pi
t
c
h ang
l
e
(
r
ad
)
0
2
4
6
8
10
12
14
16
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
open l
o
o
p
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Aircraft Control System
Usi
ng Model Pre
d
ictive Co
ntro
ller (L
aba
ne
Chrif
)
263
3.1, Roll Subsidence Mo
d
e
Roll sub
s
ide
n
ce m
ode i
s
simply the
dampin
g
of rolling motion
. There i
s
n
o
dire
ct
aero
d
ynami
c
moment cre
a
t
ed tending t
o
dire
ctly re
store wi
ng
s-le
vel, i.e. there is no retu
rni
ng
"spri
ng fo
rce
/
moment" p
r
oportio
nal to
roll
angle.
Ho
wever, th
ere i
s
a da
mping m
o
m
ent
(propo
rtional to
roll
rate
) created
by the slewi
n
g
-
abo
ut of l
ong wi
ngs. Thi
s
p
r
e
v
ents large roll
rates from bu
ilding up whe
n
roll-co
ntro
l i
nputs a
r
e ma
de or it damp
s
the roll
rate
(not the angl
e)
to zero wh
en
there a
r
e no
roll-cont
rol i
nputs.
Roll m
ode can b
e
improve
d
by addin
g
dihe
d
r
al
effects to the aircraft de
sig
n
, such as hi
gh win
g
s, dih
edral a
ngle
s
or swee
p ang
les.
Figure 5. Roll
sub
s
ide
n
ce mode
Usi
ng
a p
r
o
c
edure
simila
r to the l
ongit
udinal
mod
e
, we
can
dev
elop th
e eq
u
a
tion of
motion for the lateral dynamic
s
.
,
,
(8)
: state vector
: control ve
ctor
,
: aileron an
d rudde
r defle
ction
,
: sidesli
p and
roll angl
e
,
: roll and yaw rate
0
0
01
t
a
n
0
,
0
0
,
1
0
0
0
0
0
0
1
,
0
0
0
0
(9)
If we assume
that the measurable o
u
tpu
t
s are the
sid
e
slip a
ngle
and roll an
gle
, the
matrixes
,
and
are:
0.0558
0.9968
0.0802
0.
0415
0.
5980
0.1150
0.0318
0
3.0500
0.3880
0.4650
0
0
0
.0805
1.000
0
,
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 259 –
269
264
0
.
0729
4.7500
0
.
15300
0
0.000
0.00775
0.1430
0
,
1
0
0
0
0
0
0
1
(10)
Figure 6. Ope
n
loop Impul
se Re
spo
n
se (Sideslip a
ngl
e)
Figure 7. Ope
n
loop Impul
se Re
spo
n
se (Roll an
gle)
3.2. Spiral Mode
If a spirally unstabl
e aircraft,
through the acti
on of a gust or other
disturbance, gets a
small initial roll angle to the right, for example,
a gentle side
slip
to the right is pro
d
u
c
ed. The
side
slip
cau
s
es
a ya
wing
moment to
th
e ri
ght. If the
dihed
ral
stabi
lity is lo
w, a
n
d
yaw d
a
mpi
ng i
s
small, the di
rectional
stability keeps
turning
the aircraft while
the
continui
ng bank angle
maintain
s the
side
slip
and
the yaw a
ngl
e. This
sp
i
r
al gets contin
uo
usly
ste
epe
r and
tighte
r
u
n
til
finally, if the
motion i
s
not
che
c
ked
a
st
eep, hi
gh-s
p
e
ed, spiral div
e
results. T
h
e
motion
devel
ops
so g
r
ad
ually, however tha
t
it is usu
a
lly co
rre
cted
un
con
s
ciou
sly b
y
the pilot, who may not
be
aware that
spiral instabilit
y exis
ts. If th
e pilot cannot see the hori
zon, for i
n
stance
because of
clou
ds, he mi
ght not notice that he is slowly goin
g
into the spi
r
al
dive, which
can le
ad into
the
graveya
r
d
spi
r
al. To
be
spi
r
ally sta
b
le, a
n
airc
raft mu
st have
some
co
mbinatio
n
of a
sufficie
n
tly
large di
hed
ra
l, which in
creases roll st
ability,
and a sufficiently long vertical tail arm, whi
c
h
increases yaw dam
ping. I
n
creasing the vertical tail area th
en magnifies the
degree of
stabilit
y or
instability. The spiral dive should n
o
t be confu
s
e
d
with
a spin.
0
2
4
6
8
10
12
14
16
18
20
-5
-4
-3
-2
-1
0
1
2
3
4
5
I
m
p
u
l
s
e R
e
s
pons
e
Ti
m
e
(
s
e
c
)
s
i
de
s
l
i
p
angl
e (
r
a
d
)
o
pen l
o
op
0
2
4
6
8
10
12
14
16
18
20
0
0.
05
0.1
0.
15
0.2
0.
25
I
m
p
u
l
s
e R
e
s
p
on
s
e
Ti
m
e
(
s
e
c
)
r
o
l
l
angl
e (
r
ad)
op
en l
o
o
p
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Aircraft Control System
Usi
ng Model Pre
d
ictive Co
ntro
ller (L
aba
ne
Chrif
)
265
Figure 8. Spiral mode
3.3. Dutc
h Roll
The se
co
nd l
a
teral motio
n
is an oscillat
o
ry com
b
ine
d
roll and yaw motion calle
d Dutch
roll, perhap
s
becau
se of its simila
rity to an ic
e
-
skatin
g motion of the sam
e
name
made by Dut
c
h
skaters; the o
r
igin of the n
a
me
is u
n
cl
e
a
r. The
Dutch
roll may be
descri
bed a
s
a yaw and
rol
l
to
the right, follo
wed
by a recovery toward
s the e
qu
ilib
ri
um condition,
then an
oversho
o
ting of th
is
condition
and a yaw
and roll to
the left, then
back past the equ
ili
brium
attitude, and
so on.
The
perio
d is
usu
a
lly on the o
r
de
r of 3–1
5
se
con
d
s,
b
u
t it can vary f
r
om a fe
w seco
nd
s for li
ght
aircraft to a minute or m
o
re for ai
rline
r
s. Dam
p
i
ng i
s
increa
se
d b
y
large directi
onal sta
b
ility and
small
dihe
dra
l
and
de
crea
sed
by
small
dire
cti
onal
st
ability and
large di
hed
ral.
Although
usu
a
lly
stable in
a
norm
a
l ai
rcraft, the motion may be
so
sli
ghtly damped that
the effect i
s
v
e
ry
unpleasant and undesi
rabl
e. In swept-bac
k wi
ng ai
rcraft, the Dut
c
h roll i
s
solved by installi
ng
a
yaw d
a
mpe
r
,
in effe
ct a
speci
a
l-p
u
rp
ose autom
atic
pilot that da
mps
out a
n
y
yawing
oscill
ation
by applying rudde
r corre
c
tions. Some
swept-win
g
aircraft have
an
unstabl
e Du
tch roll. If the
Dutch roll i
s
very lightly damped
or un
stable,
the y
a
w da
mpe
r
b
e
com
e
s
a sa
fety require
m
ent,
rathe
r
than a
pilot and pa
sseng
er
conv
enien
ce. Dua
l
yaw damp
e
r
s a
r
e
req
u
ired and
a fail
ed
yaw damp
e
r
is cau
s
e fo
r limiting flight to low altitudes, and p
o
ssibly lowe
r match num
b
e
rs,
where the Dutch roll
stability is improved.
Figure 8. Dut
c
h roll m
ode
From exa
m
in
ing Figu
re 6,
it can b
e
se
en t
hat dyna
mical b
ehavi
o
r of an
aircraft is not
accepta
b
le
consi
deri
ng ov
ershoot, ri
se
time, settling
time and
st
eady-state e
r
ror val
u
e
s
, a
n
d
must be mo
di
fied usin
g fee
dba
ck
control
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 259 –
269
266
4.
Model Predictiv
e
Controller
4.1. The Rec
e
ding Hori
zo
n Idea
The figu
re sh
ows the ba
si
c idea
of pre
d
ictive
contro
l. In this pre
s
entation of th
e ba
sics,
we co
nfine o
u
rselves
to discussin
g
the
control
of
a si
ngle-i
nput, si
ngle-output (SISO)
pla
n
t.
We
assume
a di
screte
-time
set
t
ing, and that
the cu
rrent
time is la
bele
d
as time
step
k.at the curre
n
t
time the pla
n
t output i
s
, and th
at the figure sho
w
s th
e p
r
evi
ous histo
r
y of
the
outp
u
t
trajec
tory. Als
o
s
h
own is
a s
e
t point trajec
tory,
whi
c
h i
s
the traj
ectory th
at the output
sho
u
ld
follow, ideally
. The value of the set-p
o
in
t trajecto
ry at any time t is denoted by
.
Distin
ct from
the set-point t
r
aje
c
tory is th
e
referen
c
e trajecto
ry
.T
his starts
at the curre
n
t
output
, and
define
s
an ideal trajecto
ry along whi
c
h
the plant shou
ld return to the set-poi
nt
trajecto
ry, for instan
ce afte
r a di
sturb
a
n
c
e o
c
cu
rs. Th
e refe
ren
c
e trajecto
ry there
f
ore defin
es
an
importa
nt asp
e
ct of the clo
s
ed
-loo
p beh
aviour of
the
controlled
pla
n
t. It is not nece
s
sary to insist
that the pl
ant
sh
ould
be
dri
v
en ba
ck to t
he
set-p
o
in
t t
r
aje
c
tory
as fast a
s
po
ssi
bl
e, althou
gh th
at
choi
ce
rem
a
i
n
s o
pen. It is frequ
ently assume
d that
t
he refere
nce
trajecto
ry a
s
fast a
s
po
ssib
le,
althoug
h that
ch
oice
rem
a
ins op
en. It
is freq
uentl
y
assum
ed t
hat the
refe
rence traje
c
to
ry
approa
che
s
the set poi
nt expone
ntially, which we shall denote
Tref
, definin
g the spe
ed
of
respon
se. Th
at is the cu
rre
nt erro
r is
(11)
Then the referen
c
e traje
c
tory is
ch
ose
n
su
ch that the error
ste
p
s later, if the output
followe
d it exactly, would b
e
:
1
/
∗
(12)
∗
Whe
r
e
is the
sam
p
ling i
n
terval a
n
d
(no
t
e that 0<
<
1
).That is
, the reference
trajecto
ry is d
e
fined to be:
(13)
The notation
indicates that
the
refe
ren
c
e traj
ecto
ry d
epen
ds on
th
e conditio
n
s
at time
, in g
eneral. Altern
ative definitio
ns of the
re
fe
ren
c
e traje
c
tory are po
ssi
b
le-F
or e.g.,
a
straig
ht line from the cu
rre
nt out
put whi
c
h meet
s the
set point traje
c
tory after a specifie
d time.
Summari
zin
g
the main steps involved
in impleme
n
ting DM
C o
n
a pro
c
e
s
s are as
follows
:
1. Develop a
discrete
step
respon
se m
o
del with len
g
th N ba
sed o
n
sample time
t
2. Specify the predi
ction (P
) and control (M) ho
ri
zon
s
.
N
≥
P
≥
M
3. Spe
c
ify the weightin
g o
n
the
control
action
(w=0 if
no
wei
ghting
on
the
co
ntrol a
c
tion
(w=0 if no we
ighting).
4. All calculat
ions
assum
e
deviation vari
able
form, s
o
remember to c
o
nvert t
o
/from
physi
cal unit
s
.
The effect of all these tuni
ng paramete
r
s is no
w di
scussed for SIS
O
system
s.
Model
-length
and sam
p
le-tim
e
selectio
n are ind
epe
ndent. The m
odel length
should be
approximatel
y the ‘settling time’ of the
pro
c
e
ss, t
hat
is, the time re
quire
d to rea
c
h a ne
w ste
ady
state after a
step inp
u
t ch
ange. Fo
r mo
st system
s, the model le
n
g
th is rou
ghly
50 coeffici
en
ts
.
The
sam
p
le ti
me i
s
u
s
u
a
lly on th
e
ord
e
r of on
e te
nth
the do
minant
time
con
s
tan
t, so th
e mo
d
e
l
length is roug
hly the settling time of the pro
c
e
ss.
Predi
ction a
nd co
ntrol h
o
rizon
s
differ in length.
Usually, the predi
ction h
o
rizon i
s
sele
cted to b
e
much long
er than the
control ho
riz
on. This i
s
p
a
rticul
arly tru
e
if the cont
rol
weig
hting fact
or is sele
cted
to be zero. Usually,
if the
predi
ction h
o
rizon i
s
much longe
r than th
e
control ho
rizon, the co
ntrol syst
em i
s
less sensitive
to model e
r
ror. Often P=20 or
so, wh
ile
M=1-3.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Aircraft Control System
Usi
ng Model Pre
d
ictive Co
ntro
ller (L
aba
ne
Chrif
)
267
Control weig
hting
i
s
ofte
n
step
to
ze
ro if the
p
r
ed
iction
ho
rizo
n
is mu
ch
lon
ger the
control
hori
z
on. As the
control
ho
rizo
n is in
cr
ea
se
d, the
co
ntrol
move
s ten
d
to b
e
come
more
aggressive so large
r
weig
ht is need
ed to penali
z
e th
e control mov
e
s.
4.2. Objectiv
e Functio
n
s
Here, there
are
seve
ral
different
choi
ce
s for obje
c
tives functio
n
s
. The
first
one that
come
s to mi
nd is a
stan
dard
le
as
t-s
q
u
a
r
es
o
r
“qu
adrati
c
“obje
c
tive functio
n
.
The obje
c
ti
ve
function i
s
a
“su
m
of squ
a
r
es
“of the predicte
d
errors (differen
c
e
s
betwee
n
the
set points
a
n
d
model
-predi
cted outp
u
ts)
a
nd the
control
moves
(chan
ges i
n
control
action f
r
om
step to step
).
A
quad
ratic o
b
j
e
ctive functio
n
for a pre
d
i
c
tion ho
rizon
of 3 and a control ho
rizon of 2 can
be
written.
Φ
1
1
^
2
2
2
^
2
3
3
^
2
∆
^
2
∆
1^2
(14)
Whe
r
e
repre
s
ent
s the mo
del pre
d
icte
d
output, r is the set poi
nt,
∆
is
th
e
c
h
an
ge
in
manipul
ated i
nput from on
e sampl
e
to the next,
is a
weig
ht for the chan
ge
s in the manip
u
lat
e
d
input, and th
e su
bscri
p
ts indicate the
sampl
e
tim
e
(k i
s
the
curre
n
t sam
p
le time). Fo
r a
predi
ction ho
rizo
n
of
and a co
ntrol
hori
z
on
of
, the lea
s
t Squ
a
re
s obj
ectiv
e
function i
s
written.
Φ
∑
1
1
^
2
∑
∆
1
^
2
(15)
Another po
ssi
b
le o
b
je
ctive
function
is to
simp
ly tak
e
a s
u
m o
f
th
e
ab
s
o
lu
te
va
lues
o
f
th
e
predi
cted
errors an
d
control move
s. Fo
r a p
r
edi
ct
ion
hori
z
on
of 3
and a
control hori
z
o
n
of 2,
the
absolute valu
e obje
c
tive function i
s
:
Φ
|
1
1
|
|
2
2
|
|
3
3
|
|
∆
|
|
∆
1
|
(16)
Figure 9. Sideslip a
ngle af
ter applying
MPC to the aircraft
The controll
e
r
is able to t
r
ack the
side
slip
an
gle ref
e
ren
c
e a
s
lo
ng as it is feasi
b
le
referen
c
e. At 10 second
s the MPC
stabl
e the sid
e
sli
p
angle to track a reference, and remain t
h
e
rudd
er d
e
flect
i
on in the init
ial conditions, the result is shown in Figure 9.
The p
r
edi
ctive controll
er
h
a
s
alre
ady in
clud
ed
it into
the linea
r m
o
del which di
d
help th
e
system to
provide the n
e
c
e
s
sary
stabi
lity to per
form the rudde
r and
with red
u
ce
d variatio
ns in
the side
slip a
ngle.
0
5
10
15
20
25
0
0.
5
1
1.
5
S
i
des
l
i
p an
gl
e
(
r
a
d
)
ti
m
e
P
l
ant
o
u
t
put
(
s
ol
i
d
)
an
d
s
e
t
-
po
i
n
t
(
d
as
h
ed)
0
5
10
15
20
25
-2
0
2
4
x 1
0
-3
u
ti
m
e
I
n
put
re
f
e
r
MP
C
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 15, No. 2, August 2015 : 259 –
269
268
Figure 10. Ro
ll angle after
applying MP
C to the aircraft
The controlle
r is a
b
le to track the roll a
ngle re
fere
nce as lo
ng a
s
i
t
is feasibl
e
referen
c
e.
At 10 second
s the
MPC st
able th
e
side
slip
angle
to
track a
refere
nce,
and
re
m
a
in the
aile
ro
n
deflection in the initial
co
nd
itions, the re
sult is sho
w
n i
n
Figure 10.
The p
r
edi
ctive controll
er
h
a
s
alre
ady in
clud
ed
it into
the linea
r m
o
del which di
d
help th
e
system
to provide the necessary
stability to perfo
rm
the ail
e
ron
and
with
reduced variations
in
the roll angl
e.
Figure 11. Pitch an
gle after applying MP
C to the aircraft
Con
s
e
quently
, by tuning the value of
500
, the followin
g
values of m
a
trix
K
are
obtaine
d. If
is increa
sed
even highe
r, improvem
ent
to the respo
n
s
e sh
ould be
obtaine
d even
more. But for this
case, the values
of
500
is cho
s
en
beca
u
se it satisfie
d the
desig
n
requi
rem
ents while kee
p
as sm
all as po
ssi
ble.
The
controlle
r is
able to
track the
pitch a
ngle
ref
e
ren
c
e
a
s
lo
ng a
s
it is f
easi
b
le
referen
c
e. At 20 second
s the MPC
stabl
e the sid
e
sli
p
angle to track a reference, and remain t
h
e
elevator defle
ction in the ini
t
ial conditio
n
s,
the result is
sho
w
n in Fig
u
re 12.
The p
r
edi
ctive controll
er
h
a
s
alre
ady in
clud
ed
it into
the linea
r m
o
del which di
d
help th
e
system to
provide the ne
cessary
stability to perform
the elevator
and with
reduced variations i
n
the pitch an
gl
e.
We can see that the reference
tracking i
s
good, but
the controller is st
ill not suff
iciently
robu
st
i
n
cert
ain segm
ents.
The main
a
d
vantage
of
model
predi
ctive co
ntrolle
r
is d
epi
cted i
n
a
clo
s
e up
of bird pe
rspe
ctive of the aircraft trajec
to
ry in Figure 11,
whe
r
e
we ca
n clea
rly ob
serve
a p
r
elimin
ary
actio
n
of the
co
ntroll
er,
a
dire
ct con
s
e
q
uen
ce of the long predictiv
e hori
z
on an
d
the look a
hea
d function.
0
5
10
15
20
25
0
0.
5
1
1.
5
R
o
l
l
a
ngl
e
(rad
)
ti
me
P
l
an
t
out
p
u
t
(
s
ol
i
d
)
and s
e
t
-
po
i
n
t
(
d
as
he
d)
0
5
10
15
20
25
-2
0
2
4
x 1
0
-3
u
ti
me
I
n
put
re
f
e
r
MPC
0
5
10
15
20
25
0
0.
5
1
p
i
t
c
h a
ngl
e (rad)
ti
m
e
M
o
de
l
P
r
ed
i
c
t
i
v
e
C
o
nt
r
o
l
0
5
10
15
20
25
0
0.
2
0.
4
0.
6
0.
8
u
ti
m
e
I
npu
t
ref
e
r
MP
C
Evaluation Warning : The document was created with Spire.PDF for Python.