Int
ern
at
i
onal
Journ
al of
R
obot
ic
s
and
Autom
ati
on (I
JRA)
Vo
l.
8
,
No.
3
,
S
eptem
ber
201
9
,
pp.
1
89
~
1
9
3
IS
S
N:
20
89
-
4856
,
DOI: 10
.11
591/
i
jra
.
v
8
i
3
.
pp
1
89
-
1
9
3
189
Journ
al h
om
e
page
:
http:
//
ia
escore.c
om/j
ourn
als/i
ndex.
ph
p/IJRA
Artifici
al p
ancreas te
chniq
ues based on
ro
bu
st mod
el predictiv
e
co
nt
rolle
r
Waleed
Kh
alid Al
-
Az
z
awi
Facul
t
y
of
Infor
m
at
ion
T
e
chnol
o
g
y
,
Arab
Open U
nive
rsit
y
B
ahr
ai
n,
Ira
q
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
y
3
, 2
01
9
Re
vised
Ju
l
5
,
201
9
Accepte
d
Aug 1
, 2
01
9
Diabe
t
es
is
kno
wn
as
the
m
aj
or
ca
use
of
d
eath
in
th
e
world
l
ea
d
in
g
to
kidn
e
y
,
ret
inop
at
h
y
and
c
ard
iova
scu
la
r
dise
ase
s
as
we
ll
.
In
thi
s
paper,
a
Robust
Model
Predi
ct
i
ve
Contro
ll
er
(
RMP
C)
is
int
r
oduce
d
to
d
esign
artificia
l
panc
re
as
that
sol
ved
the
m
odel
u
nce
rt
ai
nt
y
and
k
ee
p
th
e
blood
gl
ucose
le
v
el
in
th
e
norm
al
r
ange
b
y
r
egul
a
ting
the
siz
e
of
i
nsulin
infusion
from
pum
p
base
d
on
RMP
C.
Th
e
sim
ula
ti
on
result
s
will
pre
s
ent
a
good
per
fo
rm
anc
e
of
the
propos
ed
cont
roller
to
a
void
disturb
ance
and
robustn
ess
aga
i
nst
unce
rt
ai
nt
ie
s.
Ke
yw
or
d
s
:
Ar
ti
fici
al
p
a
ncrea
s
MPC
techn
i
ques
Robust m
od
el
pr
e
dicti
ve
con
t
ro
ll
er
(RM
PC)
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Walee
d K
halid
A
l
-
Azza
wi
,
Faculty
of In
form
ation
Tec
hn
ology,
Ar
a
b Open
Un
iversity
Bahra
in, Ira
q
.
Em
a
il
:
waleedk
i@y
ah
oo.c
om
1.
INTROD
U
CTION
Ty
pe
1
d
ia
bete
s
is
an
inve
te
ra
te
disease
in
w
hich
the
pa
ncrea
s
do
e
s
not
pr
oduce
s
uffici
ent
insu
li
n
to
con
t
ro
l
glu
c
ose
le
vels
in
the
blood.
In
the
s
hortage
of
in
sul
in,
the
adm
it
tan
ce
of
gl
uco
s
e
into
s
kelet
al
,
c
ard
ia
c,
sm
oo
th
m
us
cl
e
and
ot
her
ti
ss
ues
is
re
duced
.
W
he
n
i
ns
ulin
is
m
issi
ng
f
or
a
longer
pe
rio
d
of
ti
m
e,
the
m
us
cle
and
ti
ssu
e
cel
ls
will
sta
rt
us
in
g
fat
a
s
energy
s
ource
,
in
place
of
gl
ucose
fro
m
the
blood
stream
.
The
dist
urba
nc
e
of
m
eal
eating
prese
nts
grea
t
chall
eng
es
to
autom
at
ic
blo
od
gl
uco
se
c
ontrol.
T
h
er
e
are
m
any
factors
m
us
t
be
m
easur
ed
t
o
regulat
e
bloo
d
glu
c
os
e
s
uch
a
s.
Fats
an
d
pr
ot
ei
ns
cause
delay
s
in
abs
orption
of
glu
c
os
e
f
r
om
c
arbo
hydr
at
es
e
at
en
at
the
sam
e
tim
e,
ph
ysi
cal
exer
ci
se
af
fec
ts
the
bloo
d
gl
ucose
re
gu
la
ti
on
by
reducin
g
the
r
equ
i
rem
ent
f
or
insu
li
n,
a
nd
f
ur
t
her
m
or
e,
str
ess
an
d
physi
cal
il
lness
are
involve
d
in
t
he
blood
glu
c
os
e
re
gula
ti
on
process.
T
hu
s
,
de
sig
ning
accurate
pr
e
dic
ti
ve
m
od
el
s
ar
e
hard
to
ac
hie
ve.
I
n
2011,
D
i
m
i
tr
i
Boirou
x
et
al,
app
li
ed
a ro
bus
t feed
forw
a
r
d
-
f
eedb
ac
k
c
ontr
ol
str
at
egy to
pe
op
le
with ty
pe 1
diabetes.
The
f
eed
forw
a
r
d
c
on
t
rol
le
r
co
ns
ist
s
of
a
bolus
cal
c
ulator
w
hich
c
om
pen
sat
es
th
e
distu
r
ban
ce
com
ing
f
ro
m
m
eal
s.
The feed
bac
k
c
on
t
ro
ll
er is
b
as
ed on a li
nea
rized
descr
i
ptio
n of t
he
m
od
el
de
scribin
g
t
he p
at
ie
nt [
1].
In
2008,
Gia
nni
Ma
rc
hetti
et
al
con
ce
r
ned
with
t
he
im
pr
ovem
ent
of
ne
w
fe
ed
f
orwa
rd
-
fee
dback
con
t
ro
l
te
ch
ni
qu
e
s
f
or
act
ua
l
glu
cose
c
ontrol
an
d
ty
pe
1d
ia
betes.
E
nhanced
post
-
m
e
al
respon
se
s
c
an
be
achieve
d
by
a
pr
e
-
pra
nd
ia
l
sna
ck
or
bolus
,
or
by
re
du
ci
ng
t
he
gl
uco
se
set
-
po
i
nt
pr
i
or
to
t
he
m
eal
[2
]
.
In
20
17,
Mojga
n
E
sn
a
-
Ash
a
ri
et
al
pro
posed
a
te
chn
i
qu
e
for
bl
ood
gl
ucose
l
evel
regulat
io
n
i
n
ty
pe
I
di
abetes.
The
c
ontr
ol
ap
proac
h
is
base
d
on
no
n
-
li
nea
r
m
od
el
pr
e
dicti
ve
c
on
t
ro
l.
T
he
pur
pose
of
the
pro
posed
co
nt
ro
ll
er
op
ti
m
iz
ed
with
gen
et
ic
al
gori
thm
s
is
to
m
easur
e
bloo
d
gl
uc
os
e
le
vel
e
very
tim
e
and
est
im
at
e
it
fo
r
the
nex
t
tim
e
interval
[3
]
.
I
n
2017,
Lu
kas
Or
tm
ann
et
al
,
i
ntrodu
ce
d
a
M
odel
Pr
e
dicti
ve
Con
tr
oller
that
ta
ke
s
the
per
i
od
ic
I
nsuli
n
sen
sit
ivit
y
into
acc
ount,
s
o
as
t
o
im
pr
o
ve
bloo
d
gl
ucose
c
on
tr
ol.
T
he
upc
om
ing
ef
f
ect
of
the
I
nsuli
n
se
nsi
ti
vity
is
est
i
m
at
ed
by
a
m
achine
le
a
rn
i
ng
te
chn
iq
ue,
na
m
el
y,
a
custo
m
iz
ed
Gaussi
an
P
rocess
(G
P
),
base
d
on
histo
rical
trai
ni
ng
data
[
4].
I
n
2017,
C
hiara
Toffa
nin
,
pro
pose
d
an
ada
ptive
Mo
del
P
redi
ct
ive
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
, No
.
3
,
Septem
ber
201
9
:
1
89
–
1
9
3
190
Con
tr
ol
a
pproach
to
ac
co
unt
for
it
an
d
te
st
it
in
sil
ic
o.
R
un
-
to
-
Ru
n
m
eth
od
a
da
pts
the
subcuta
neous
basal
insu
li
n
delive
r
y
throu
ghout
the
ni
gh
t
a
nd
the
car
bohy
dr
at
e
-
to
-
i
nsuli
n
rat
io
thr
ough
ou
t
t
he
day,
ac
co
rdi
ng
to
so
m
e p
erf
orm
a
nce indic
es m
e
asur
e
d
f
r
om
su
bcu
ta
neous
c
onti
nuous g
l
uc
ose
sen
s
or
data [5
]
. In 201
7,
B
erno J
.
E.
Mi
sg
el
d
et
al
pr
ese
nted
a
novel
m
et
ho
d
to
wards
real
-
tim
e
est
i
m
at
i
on
of
ins
ulin
sensiti
vity
in
ty
pe
-
1
diabetes.
T
he
pr
e
dicti
ve
of
i
ns
uli
n
se
ns
it
ivi
ty
is
inv
e
sti
gated
w
hile
a
ca
r
bo
hydrat
e
upta
ke
distu
rb
a
nce
ta
kes
place
sim
ultan
eousl
y
[6
]
.
I
n
const
ru
ct
,
i
n
th
is
pap
e
r,
th
e
Robust
Mo
del
P
red
ic
ti
ve
C
on
tr
oller
(RMPC
)
desi
g
n
appr
oach
will
be
em
plo
ye
d
for
the
regul
at
ion
of
c
on
st
raine
d
syst
em
s.
RM
PC
al
gorithm
s
are
able
to
coor
din
at
e
hard
c
o
nst
raints
on
sta
te
ou
t
pu
ts
an
d
in
puts.
Stabil
it
y
and
rob
us
tness
are
co
ns
ide
red
pr
op
e
rtie
s
o
f
RM
PC
con
t
ro
l
le
rs,
as
t
hey
ar
e
essenti
al
to
hard
war
e
im
ple
m
entat
ion
.
A
whole
the
or
y
on
r
obus
t
sta
bili
ty
of
RM
PC h
as
bee
n
a
pp
li
ed
for l
inear
and
nonli
near sy
ste
m
s.
2.
PROBLE
M
DE
SC
R
IPTIO
N
The
co
ntr
ollers
are
buil
t
in
bo
th
tra
diti
onal
a
nd
RM
PC.
Howev
e
r,
it
holds
only
for
li
near
sc
hem
e
of
the
syst
em
.
Wh
il
e
deali
ng
with
the
no
n
-
li
nea
r
syst
e
m
op
erati
on
s
,
the
syst
em
sh
oul
d
be
li
ne
arised.
This
li
near
ise
d
form
of
the
non
-
li
ne
ar
syst
e
m
cou
ld
be
pr
op
e
rly
con
t
ro
l
us
in
g
tra
diti
onal
con
tr
ollers
s
uch
as
(i.e.
Mo
del
Pre
dicti
ve
C
on
t
r
oller).
In
a
n
Ar
ti
fici
al
Pa
nc
reas
re
gu
la
ti
on
,
c
on
ti
nuous
s
ub
c
utane
ous
insu
li
n
inf
us
io
n
c
omm
on
with
sub
cutane
ous
gluc
os
e
sensi
ng
has
bee
n
sho
wn
to
be
e
ff
e
c
ti
ve
f
or
fe
ed
f
orwar
d
channel
“o
ut
-
of
m
eal
”
per
i
ods,
pr
e
ser
vatio
n
bl
ood
glu
c
os
e
in
the
regular
le
vel
in
the
post
a
bsor
ptive
case.
Howe
ver,
the
feedbac
k
c
hannel
has
not
be
en
s
o
s
uccess
f
ul
in
treat
in
g
i
ns
uli
n
de
sires
at
m
eal
s
sta
te
[7
,
8].
In
s
pite
of
de
vel
opm
ents
in
sen
sor
a
nd
pum
p
eq
uip
m
e
nt,
t
he
Ar
ti
fic
ia
l
Panc
reas
m
us
t
coo
r
din
a
te
wit
h
the
delay
s
a
nd
inaccu
racies
i
n
gl
ucose
se
ns
i
ng.
T
his
is
ve
ry
hard
w
he
n
a
sy
stem
disturba
nc
e,
s
uc
h
as
,
a
m
eal,
happe
ns
a
nd
trig
ger
s
a
fast
glu
c
os
e
inc
rea
se
that
is
co
nsi
der
a
bly
faster
than
t
he
ti
m
e
require
d
for
i
ns
uli
n
abs
orption.
T
he
sta
bili
ty
pr
oble
m
of
con
st
raints
pla
nts
in
cl
os
e
d
-
l
oop
with
MPC
co
ntr
ollers
is
show
n
in
Figure
1.
Figure
1. A
rtific
ia
l p
anc
reas t
echn
i
qu
e
s
3.
METHO
DOL
OGY
Mod
el
pr
e
dicti
ve
c
on
tr
ol
is
a
m
et
ho
d
of
co
ntr
ol
in
w
hic
h
the
prese
nt
co
ntr
ol
sig
nal
is
ob
ta
ine
d
by
so
lvi
ng,
at
each
sam
pling
tim
e,
a
finite
co
nst
raints
open
-
lo
op
optim
al
con
trol
pro
blem
,
us
ing
the
rece
nt
sta
te
of
t
he
syst
em
as
the
init
ia
l
sta
te
,
the
op
ti
m
i
zat
ion
pro
du
ce
s
an
optim
al
c
on
t
ro
l
si
gn
al
s
equ
e
nce
a
nd
t
he
fi
rst
con
t
ro
l
sig
nal
in
this
se
qu
e
nc
e
is
app
li
ed
t
o
the
syst
e
m
.
M
PC
is
able
to
c
op
e
with
hard
const
raints
on
con
t
rol
sign
al
s
a
nd
sta
te
s
as
sho
wn
in
Fi
gure
2.
Th
eref
or
e
,
been
a
pp
li
ed
in
m
edical
app
li
cat
io
n
w
her
e
ap
pro
va
l
of
cons
trai
nts
is
m
os
tly
ver
y
sign
i
ficant
since
eff
ic
ie
ncy
re
quest
s
operati
ng
points
on
or
cl
os
e
to
the
li
m
it
of
the set
of
ac
ce
ptable stat
es a
nd c
on
t
ro
l si
gn
a
ls.
Figure
2. MPC
for
c
onstrai
nts
syst
e
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Arti
fi
ci
al p
anc
reas
te
c
hn
i
qu
e
s base
d on rob
us
t m
odel
pr
ed
ic
ti
ve con
troll
e
r
(
Waleed
Kh
al
id Al
-
Azz
awi
)
191
4.
MO
DELIN
G
AND SIM
UL
ATIO
N
In
t
his
pap
e
r,
t
he
MPC
desig
n
will
be
em
plo
ye
d
for
t
he
re
gu
la
ti
on
of
c
on
strai
ned
syst
em
s.
A
rob
us
t
sta
bili
ty
of
MPC
has
bee
n
a
vo
i
d
vio
la
ti
ons
of
in
pu
t
a
nd
ou
t
pu
t
c
onstrai
nts
dri
ve
certai
n
ou
t
pu
t
var
ia
bles
to
their
optim
al
set
points,
wh
il
e
kee
pi
ng
oth
e
r
ou
tp
uts
withi
n
sp
eci
fie
d
val
ue.
A
void
re
dunda
nt
m
ov
em
ent
of
the
in
put
var
ia
bles.
Esti
m
a
ted
values
of
ou
tpu
t
var
ia
bles
are
pr
e
dicte
d
us
in
g
a
li
nea
r
qu
a
drat
ic
re
gu
l
at
or
of
the
plant
a
nd
cu
rr
e
nt
m
easur
em
ents.
T
he
co
ntr
ol
si
gn
a
l
cal
culat
ion
s
are
acc
ordin
g
t
o
bo
th
up
com
ing
pr
e
dicti
on
s
a
nd r
ece
nt
m
easur
em
ents.
The
c
on
t
ro
l
sig
nals
(
)
,
at
the
ℎ
sam
pling
tim
e
are
cal
c
ul
at
ed
s
o
t
hat
they
m
ini
m
iz
e
a p
e
rfor
m
ance
ind
e
x
J
. th
e
constraints
of
op
t
i
m
al
co
ntro
l
syst
e
m
as
(1)
:
{
(
)
≤
(
)
≤
(
)
(
)
≤
(
)
≤
(
)
(1)
The In
su
li
n ki
ne
ti
c equ
at
io
ns
are
(2)
:
̇
=
1
[
−
(
)
+
(
)
]
̇
=
1
[
−
(
)
+
(
)
]
(2
)
̇
=
1
[
−
(
)
+
(
)
]
Wh
e
re,
I
-
ins
ul
in p
la
sm
a, X
-
insu
li
n
inter
sti
tial, S
-
ins
ulin s
ensiti
vity
an
d
U
-
co
ntr
ol sig
na
l
Wh
il
e, t
he gluc
os
e
kin
et
ic
e
quat
ion
s a
re
(3)
[
9]:
̇
=
(
)
+
(
)
+
1
[
(
)
+
(
)
]
(3
)
̇
=
[
(
)
−
(
)
]
−
(
)
(
)
w
he
re,
G
-
glu
c
os
e
plasm
a, an
d Y
-
subc
utane
ou
s
c
om
par
t
m
e
nt
.
Accor
ding
t
o
m
edical
ex
per
i
ence
,
,
,
,
,
can
be
c
onside
red
c
onsta
nt
.
Acc
ordin
g
t
o
insu
li
n
a
nd
gluc
os
e
kin
et
ic
eq
uation
the
sta
te
sp
ace
represe
ntati
on
for
bo
t
h
ins
ulin
a
nd
gl
uco
se
e
quat
io
ns
ar
e
(4
-
5)
:
[
̇
̇
̇
]
=
[
−
1
0
1
−
1
0
0
0
−
1
]
[
]
+
[
0
0
1
]
(
)
(4)
[
̇
̇
]
=
[
−
1
1
−
(
+
(
)
]
[
]
(5)
Accor
ding
to
m
edical
analy
sis
the
he
patic
ba
la
nce
eq
uatio
ns
was
de
si
gne
d
to
reduce
t
he
eff
ect
ive
of
m
eal
(
disturba
nce)
(
6)
=
{
0
.
88
≥
0
.
88
0
≤
≤
0
.
88
0
ℎ
=
0
.
06
−
0
.
25
=
{
≥
0
.
25
0
ℎ
=
0
.
25
+
(6
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
, No
.
3
,
Septem
ber
201
9
:
1
89
–
1
9
3
192
=
0
.
87
+
ℎ0
.
0045
(
−
175
)
=
0
.
11
Fr
om
the
c
onst
raints
syst
em
outp
ut,
t
he
as
sum
ed
in
pu
t
an
d
ou
t
pu
t
are
gi
ve
n
as
known
pa
ram
et
er
to
the
co
ns
tr
ai
nts
syst
e
m
identific
at
ion
t
oo
l
box.
It
prov
i
des
the
best
a
ppr
opriat
e
tra
ns
fe
r
f
unct
ion
by
us
in
g
the
ada
ptive
tu
ning
m
et
ho
d
to
get
the
val
ues
of
po
l
es
a
nd
z
ero
es
a
nd
li
nea
rized
the
c
onst
raints
syst
em
wh
ic
h
is
sho
wn
in
Fi
gure
3.
T
he
ac
hieve
d
syst
e
m
m
od
el
is
us
e
d
to
li
nea
rise
t
he
co
ns
trai
nts
sy
stem
and
the
m
od
el
pr
e
dicti
ve
is
de
sign
e
d
t
o
sim
ul
at
e the syst
em
as prese
nted
i
n Fi
gure
4.
Figure
3. Me
as
u
re
d
a
nd sim
ul
at
ed
syst
em
o
utp
ut
Figure
4. I
nput
stat
e (in
s
ulin
) t
o
outp
ut stat
e
(b
l
ood gluc
os
e
)
f
or
plant
Af
te
r
fi
nd
i
ng
the
tra
diti
onal
(P
I
D)
c
on
t
ro
ll
er,
the
pla
nt
is
ob
viousl
y
check
e
d
a
nd
est
im
at
es
the
pe
rfo
rm
ance.
H
oweve
r,
the
set
tl
ing
tim
e
ta
ken
is
in
fi
nite
an
d
t
he
process
is
not
st
able
f
or
the
es
tim
a
te
d
ou
t
pu
t.
Wh
il
e,
w
hen
the
R
MPC
is
co
nne
ct
ed
vice
P
ID
co
ntr
oller.
T
he
RM
PC
t
ha
t
ha
ve
been
de
sign
e
d
the
c
ontrolle
d
var
ia
bles
a
nd
m
od
el
outp
ut
base
d
o
n
M
PC
co
ntr
oller
to
olb
ox.
T
he
in
herent
desig
ne
d
R
MP
C
con
t
ro
ll
er
had
m
od
el
w
hich
is
s
how
n
in
Fig
ur
e
5.
RM
PC
c
an
i
ncr
ea
se
syst
e
m
sp
eed
a
nd
est
i
m
ation
acc
ur
acy
.
This tech
ni
qu
e
g
ive
s a
fairly
pro
per
e
stim
ate
of the
sp
ee
d and i
s ro
bust t
o any
par
am
et
er v
a
riat
io
n.
Figure
5. RM
P
C o
utput f
or in
su
li
n
a
nd
bloo
d gluc
os
e stat
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Arti
fi
ci
al p
anc
reas
te
c
hn
i
qu
e
s base
d on rob
us
t m
odel
pr
ed
ic
ti
ve con
troll
e
r
(
Waleed
Kh
al
id Al
-
Azz
awi
)
193
5.
CONCL
US
I
O
NS
In
this
pa
per
,
we
desc
ribe
d
RM
PC
based
on
a
li
nea
rized
form
of
con
st
raints
syst
em
.
The
sta
te
is
increase
d
with
an
i
ns
erte
d
m
eal
(d
ist
ur
ban
c
e)
to
e
nsure
a
sta
te
con
tr
ol
of
the
bloo
d
glu
co
se.
We
us
e
MPC
te
chn
iq
ues
to
cal
culat
e
ins
ulin
qua
ntit
y
in
t
he
cases
w
he
r
e
the
in
su
li
n
s
ensiti
vity
re
duct
ion
s
an
d
in
t
he
case
wh
e
re
th
e
an
y
m
eal
(d
ist
urba
nce)
is
ta
ke
n.
T
he
sim
ul
at
ion
resu
lt
s
pr
ese
nted
in
t
his
pa
pe
r
s
how
tha
t
the
desig
ne
d
RM
PC
is
m
or
e
ef
fecti
ve,
it
has
faster
set
tl
ing
tim
es.
Th
e
unce
rtai
n
pa
ra
m
et
er
in
i
nput
an
d
it
s
eff
ect
on
the
diff
e
re
nt
facto
rs
of
t
he
real
-
tim
e
syst
e
m
a
re
al
so
m
easur
ed
.
T
he
desig
ned
RM
PC
no
t
on
l
y
consi
der
e
d
of
the
un
e
xpect
ed
distu
r
ban
ce
s
in
syst
em
,
howe
ver
al
s
o
recover
t
he
c
ertai
n
pa
ram
eter
s
t
o
the d
e
sired
v
al
ues, he
nce
pr
e
s
enting t
he r
ob
ust
ness
i
n per
for
m
ance.
REFERE
NCE
S
[1]
Boiroux,
Dim
itri;
Finan
,
Dani
el
Aaron
,
“
Strat
eg
ie
s
for
glu
cose
cont
ro
l
in
peopl
e
wi
th
t
y
pe
1
dia
b
etes
,
”
In
Proc
ee
dings
of
th
e
18
th
Wo
rld
Congress
of
the
Int
ernati
on
al
F
ede
ration
o
f
Au
tomati
c
Co
ntrol
(
IFA
C)
DO
I:
10.
3182/201108
28
-
6
-
IT
-
1002.
03714
,
2011.
[2]
Gianni
Marc
h
et
t
i
,
Mass
imili
ano Barol
o
,
Loi
s Jovanovi
č
,
“
A fe
edf
orward
–
fee
db
ack gl
ucose
con
tro
l
strateg
y
for
t
y
p
e 1
dia
be
te
s m
el
l
it
us
,
”
El
se
vi
er,
Jour
nal
of
Proc
ess
Control
,
Vol
ume
1
8,
Iss
ue
2
,
Fe
bru
ary
2008,
Page
s
149
-
162.
[3]
Mojgan
Esna
-
A
shari
,
Ma
r
y
am
Ze
kri
,
Masood
As
kar
i
,
“
Predic
t
ive
Control
of
t
he
B
lood
Gluco
se
L
evel
in
T
y
p
e
I
Diabe
t
ic
Pa
tient
Us
ing
Del
a
y
D
iffe
ren
ti
a
l
Equa
t
ion
W
ang
Mode
l
,
”
Journal
of
Me
dic
a
l
Signa
ls
and
Sensors
,
2
017
Jan
-
Mar;
7(
1)
:
8
–
20.
[4]
Luka
s
Ortm
ann,
Dawei
Shi
,
E
y
a
l
Dass
au,
“
Gauss
ia
n
Proce
ss
-
Bas
ed
Model
Predictive
Con
trol
of
Blood
Glucose
f
or
Pati
ent
s
with T
ype
1
Di
abete
s M
el
litus
,
”
arXi
v:1
707.
09948v1
[
cs
.
S
Y
]
31
Jul 2017
.
[5]
C.
Toff
an
in,
R
.
Visenti
n,
M
.
Mess
ori,
F.
Di
Palma,
L
.
Mag
ni,
and
C
.
Cob
el
li,
“
Towa
rds
a
run
-
to
-
run
ad
a
pti
ve
art
if
ic
i
al pa
ncr
eas
:
In
si
l
ic
o
results
,”
IE
EE Tr
ans. Biomed. E
ng
.
,
2
017.
[6]
B.
J.
Misgeld
,
P
.
G.
T
enbr
ock
,
and
S.
Le
onh
ar
dt,
“
Redu
ce
d
-
or
der
fi
ltering
for
insuli
n
sensi
ti
vi
t
y
esti
m
ation
un
der
ext
ern
al
d
isturbance
s
,
”
in
Am
er.
Control
Conf
.
I
E
EE
,
2017,
pp.
14
44
–
1449.
[7]
C.
D.
Man
M.
C
amill
er
i
C
.
Cob
e
ll
i
,
"
A
s
y
stem
m
odel
o
f
or
al
glu
c
ose
absorpt
ion:
Vali
da
ti
on
on
g
old
stand
ard
data
,
"
IEE
E
Tr
ans.
Bi
o
med.
Eng
.
,
vo
l. 5
3
no.
12
pp
.
247
2
-
2478
Dec
.
200
6.
[8]
F.
Chee T.
Fern
a
ndo,
“
Closed
-
Lo
op
Control of
B
l
ood
Glucose
,
”
G
ermany
Be
rl
in:
Springer
-
Ve
rlag
2007.
[9]
A.
Colosim
o,
“
Medic
a
l
Dat
a
A
naly
s
is
,
”
3
rd
int
e
rnational
S
ymposium,
ISMDA
20
02,
Rome
,
Italy,
2002
Evaluation Warning : The document was created with Spire.PDF for Python.