T
E
L
KO
M
N
I
KA
T
e
lec
om
m
u
n
icat
ion
,
Com
p
u
t
i
n
g,
E
lec
t
r
on
ics
an
d
Cont
r
ol
Vol.
18
,
No.
1
,
F
e
br
ua
r
y
2020
,
pp.
19
~
29
I
S
S
N:
1693
-
6930,
a
c
c
r
e
dit
e
d
F
ir
s
t
G
r
a
de
by
Ke
me
nr
is
tekdikti
,
De
c
r
e
e
No:
21/E
/KP
T
/2018
DO
I
:
10.
12928/
T
E
L
KO
M
NI
KA
.
v18i1.
13760
19
Jou
r
n
al
h
omepage
:
ht
tp:
//
jour
nal.
uad
.
ac
.
id/
index
.
php/T
E
L
K
OM
N
I
K
A
A
p
p
li
c
a
t
io
n
an
d
e
val
u
at
io
n
o
f
t
h
e
n
e
u
r
al
n
e
t
w
or
k
i
n
ge
a
r
b
o
x
Dheyaa
S
h
ah
e
e
d
Al
-
Az
z
awi
Co
l
l
eg
e
o
f
C
o
mp
u
t
er
Sci
e
n
c
es
a
n
d
In
f
o
rmat
i
o
n
T
ec
h
n
o
l
o
g
y
,
W
as
i
t
U
n
i
v
ers
i
t
y
,
Iraq
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
J
ul
30
,
2019
R
e
vis
e
d
Aug
24
,
2019
Ac
c
e
pted
S
e
p
1
3
,
20
19
W
e
d
ev
e
l
o
p
ed
o
l
d
d
e
s
i
g
n
e
d
o
f
a
Back
-
Pro
p
a
g
at
i
o
n
n
eu
ral
n
et
w
o
r
k
(BPN
N
),
w
h
i
ch
i
t
w
as
d
e
s
i
g
n
e
d
b
y
o
t
h
er
res
earc
h
ers
,
an
d
w
e
mad
e
mo
d
i
fi
ca
t
i
o
n
i
n
t
h
e
i
r
s
t
r
u
ct
u
re.
T
h
e
1
s
t
v
e
l
o
c
i
t
y
rat
i
o
w
a
s
d
i
s
cr
i
mi
n
at
e
d
b
y
l
o
w
e
s
t
s
p
ee
d
,
an
d
h
i
g
h
e
s
t
t
w
i
s
t
.
T
h
e
6
th
v
el
o
c
i
t
y
rat
i
o
w
a
s
d
i
s
cri
m
i
n
a
t
ed
b
y
h
i
g
h
es
t
s
p
ee
d
,
an
d
l
o
w
es
t
t
w
i
s
t
.
T
h
e
a
i
m
o
f
t
h
i
s
p
ap
er
i
s
t
o
d
e
s
i
g
n
n
eu
r
al
s
t
r
u
ct
u
re
g
e
t
b
es
t
p
erfo
rma
n
ce
t
o
co
n
t
ro
l
an
el
ect
r
i
cal
au
t
o
m
o
t
i
v
e
t
ra
n
s
p
o
rt
a
t
i
o
n
s
i
x
-
s
p
eed
g
earb
o
x
o
f
t
h
e
v
eh
i
cl
e.
W
e
f
o
cu
s
o
n
t
h
e
e
v
al
u
at
i
o
n
o
f
t
h
e
BPN
N
t
o
s
el
ec
t
t
h
e
s
u
i
t
ab
l
e
n
u
m
b
er
o
f
l
ay
er
s
an
d
n
eu
r
o
n
s
.
E
x
p
er
i
men
t
al
l
y
,
t
h
e
s
t
r
u
ct
u
re
o
f
t
h
e
p
r
o
p
o
s
e
d
BPN
N
are
co
n
s
t
r
u
ct
e
d
fro
m
fo
u
r
l
ay
er
s
:
ei
g
h
t
i
n
p
u
t
n
o
d
es
i
n
t
h
e
f
i
rs
t
l
a
y
er
t
h
at
rec
ei
v
ed
d
at
a
i
n
b
i
n
ary
n
u
m
b
er
,
4
5
n
e
u
ro
n
s
i
n
1
s
t
h
i
d
d
e
n
-
l
a
y
er,
2
5
n
eu
ro
n
s
i
n
2
nd
h
i
d
d
en
-
l
ay
er,
an
d
6
n
eu
ro
n
s
i
n
t
h
e
fo
u
rt
h
l
ay
er.
T
h
e
MSE
an
d
n
u
m
b
er
o
f
E
p
o
c
h
s
are
t
h
e
m
ai
n
fact
o
rs
u
s
ed
fo
r
t
h
e
ev
al
u
at
i
o
n
o
f
t
h
e
p
r
o
p
o
s
e
d
s
t
ru
c
t
u
re,
a
n
d
c
o
mp
ared
w
i
t
h
t
h
e
o
t
h
er
s
t
ru
ct
u
res
w
h
i
ch
w
as
d
es
i
g
n
ed
b
y
o
t
h
er
res
earc
h
ers
.
E
x
p
eri
me
n
t
a
l
l
y
,
w
e
d
i
s
co
v
ered
t
h
at
t
h
e
b
es
t
v
a
l
u
e
o
f
E
p
o
c
h
an
d
MS
E
w
as
ch
o
s
e
n
w
h
en
t
h
e
BPN
N
co
n
s
i
s
t
ed
o
f
t
w
o
h
i
d
d
en
-
l
ay
er
s
,
4
5
,
an
d
2
5
n
eu
ro
n
s
i
n
t
h
e
1
st
an
d
2
nd
h
i
d
d
en
-
l
ay
er
re
s
p
ec
t
i
v
el
y
.
T
h
e
i
mp
l
emen
t
at
i
o
n
w
as
a
p
p
l
i
e
d
u
s
i
n
g
MA
T
L
A
B
s
o
ft
w
are.
K
e
y
w
o
r
d
s
:
Ar
ti
f
icia
l
ne
ur
a
l
ne
twor
k
Automatic
tr
a
ns
mi
s
s
ion
ge
a
r
box
Ba
c
k
pr
opa
ga
ti
on
ne
ur
a
l
ne
twor
ks
Ne
ur
a
l
ne
twor
k
c
la
s
s
if
ier
s
P
a
tt
e
r
n
r
e
c
ognit
i
on
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e
.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Dhe
ya
a
S
ha
he
e
d
Al
-
Az
z
a
wi
,
C
oll
e
ge
of
C
omput
e
r
S
c
ienc
e
s
a
nd
I
nf
or
mation
T
e
c
hnology
,
W
a
s
it
Unive
r
s
it
y,
I
r
a
q
.
E
mail:
da
laz
z
a
wi@uow
a
s
it
.
e
du.
iq
1.
I
NT
RODU
C
T
I
ON
T
he
tr
a
ns
por
tation
ge
a
r
box
c
a
n
be
a
c
qua
int
e
d
a
s
a
tool
uti
li
z
e
d
f
or
movi
ng
the
mec
ha
nica
l
moveme
nt
f
o
r
the
ve
hicle
’
s
e
ngine
a
nd
a
s
s
igni
ng
it
to
the
whe
e
ls
,
a
nd
th
is
oc
c
ur
r
e
d
by
c
onn
e
c
ti
ng
to
the
f
lywhe
e
l
that
wa
s
loca
ted
a
t
the
ba
c
k
e
nd
of
t
he
e
ngine.
T
he
a
im
of
us
ing
thi
s
tool
is
inc
r
e
a
s
ing
tor
que
a
nd
de
c
r
e
a
s
ing
the
r
ound
pe
r
hour
o
f
the
c
r
a
nks
ha
f
t
of
the
mot
o
r
e
ngine
to
be
s
uit
a
ble
s
pe
e
d
a
c
c
or
ding
to
s
pe
e
d
of
the
d
r
ivi
ng
whe
e
ls
[
1]
.
T
he
tr
a
ns
por
tation
ge
a
r
box
c
ons
is
ts
of
s
e
ve
r
a
l
s
e
r
r
a
ted
s
ha
f
ts
,
a
nd
nu
mb
e
r
of
ge
a
r
s
of
va
r
ious
ve
locity
r
a
ti
o,
e
a
c
h
of
one
a
ble
to
c
ontr
ol
a
nd
c
ha
nge
the
r
otation
ve
locity
to
dif
f
e
r
e
nt
ve
locity
r
a
ti
o
.
I
n
a
ny
ve
hicle
,
ther
e
a
r
e
two
kinds
of
ge
a
r
,
a
utom
oti
ve
a
nd
manua
l
ge
a
r
,
a
nd
in
thi
s
p
a
pe
r
,
we
c
onc
e
r
ne
d
a
t
the
a
utom
oti
ve
ge
a
r
box
type
[
2
,
3]
.
T
he
a
utom
oti
ve
tr
a
ns
mi
s
s
ion
ge
a
r
box
c
ons
is
ts
of
a
s
pe
c
if
ied
number
of
ge
a
r
whe
e
ls
,
a
nd
the
a
ppr
opr
iate
whe
e
l
is
s
e
lec
ted
f
or
e
a
c
h
s
pe
e
d
a
utom
a
ti
c
a
ll
y
a
nd
without
the
int
e
r
ve
nti
on
of
th
e
dr
iver
,
pr
ovidi
ng
the
c
om
f
or
t
o
f
d
r
iver
.
T
he
r
e
a
r
e
n
umer
ous
kinds
o
f
a
utom
oti
ve
tr
a
ns
mi
s
s
ion
ge
a
r
box:
s
e
mi
-
a
utom
a
ti
c
tr
a
ns
mi
s
s
ion,
hydr
a
uli
c
a
nd
c
onti
nuous
c
ha
nge
[
4
,
5]
.
T
he
hydr
a
uli
c
ge
a
r
box
us
e
d
a
f
lui
d
c
oupli
ng
ins
tea
d
of
f
r
iction
c
lut
c
h,
whic
h
wa
s
us
e
d
in
manua
l
(
tr
a
dit
ional)
ge
a
r
box
type.
T
he
s
e
mi
-
au
tom
a
ti
c
a
nd
c
onti
nuous
ly
va
r
iable
ge
a
r
box
tec
hnique
c
ha
n
ge
s
the
s
pe
e
d
r
a
ti
on
de
pe
nding
on
a
n
in
telli
ge
nt
c
omput
e
r
pr
ogr
a
m
not
a
s
the
t
r
a
dit
ional
hydr
a
ul
ic
type.
T
h
e
s
ys
tem
de
tec
t
the
s
pe
e
d
of
d
r
ivi
ng
whe
e
ls
a
nd
c
hoos
ing
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
1
,
F
e
br
ua
r
y
2020
:
19
-
29
20
the
a
ppr
opr
iate
ve
locity
r
a
ti
o
of
g
e
a
r
box,
whic
h
s
uit
a
ble
with
thi
s
s
pe
e
d
[
6
,
7]
.
E
xpe
r
im
e
ntally,
th
e
B
P
NN
wa
s
the
be
s
t
c
hoice
a
s
int
e
ll
igent
s
ys
tem
be
twe
e
n
c
las
s
if
ica
ti
on
manne
r
s
f
or
the
r
e
c
ognit
ion
p
r
oc
e
s
s
e
s
.
T
he
B
P
NN
us
e
d
gr
a
dient
de
s
c
e
nt
a
s
opti
mi
z
a
ti
on
method
in
the
tr
a
ini
ng
method,
it
de
tec
t
a
nd
s
um
the
los
s
f
unc
ti
on
of
gr
a
dient
o
f
a
ll
we
ight
s
in
the
ne
ur
a
l
ne
t,
a
nd
the
gr
a
dient
is
r
e
us
e
d
to
upda
te
the
we
ight
s
f
or
a
tt
e
mpt
ing
to
mi
nim
ize
the
e
r
r
o
r
f
unc
ti
on
[
8
-
10]
.
B
P
NN
is
c
las
s
if
ied
a
s
a
s
upe
r
vis
e
d
le
a
r
ning,
thus
,
it
ne
e
d
to
s
upply
r
e
qu
ir
e
d
output
to
e
a
c
h
da
ta
e
ntr
y
f
or
c
omput
ing
the
e
r
r
o
r
f
unc
ti
on.
I
t
wa
s
mul
ti
-
la
ye
r
f
e
e
d
f
or
wa
r
d
ne
ts
a
nd
us
e
d
c
ha
in
r
ule
in
it
e
r
a
ti
ve
manne
r
f
o
r
c
a
lcula
ti
ng
the
g
r
a
dient
f
unc
ti
on
f
or
e
a
c
h
laye
r
.
I
t
im
por
tant
to
c
a
ll
a
s
uit
a
ble
a
c
ti
va
ti
on
f
unc
ti
on
f
or
e
a
c
h
laye
r
[
11
-
14]
.
T
he
wor
k
o
f
B
P
NN
is
s
umm
a
r
ize
d
a
s
f
o
ll
ows
:
whe
n
da
ta
of
lea
r
ne
d
objec
t
is
a
ppli
e
d
the
outpu
t
node
is
c
omput
e
d
by
a
c
ti
va
ti
on
f
unc
ti
on
f
o
r
the
we
ight
e
d
s
ums
mul
ti
p
li
e
d
with
input
nod
e
s
,
then
the
c
ompar
is
on
be
twe
e
n
c
omput
e
d
output
a
nd
de
s
ir
e
d
output
wa
s
c
a
lcula
ted
to
ge
t
the
e
r
r
o
r
va
lue
,
a
nd
thi
s
va
lue
of
e
r
r
or
is
us
e
d
to
upda
te
the
we
ight
s
a
ga
in
f
o
r
opt
im
izing
the
ne
w
r
e
s
ult
o
f
output
c
omput
a
ti
ons
[
15
-
18]
.
T
he
B
P
NN
is
c
ons
tr
uc
ted
f
r
om
a
t
lea
s
t
thr
e
e
laye
r
s
,
whic
h
they
a
r
e
:
input
la
ye
r
(
f
i
r
s
t
laye
r
)
,
hidden
laye
r
(
s
)
(
s
e
c
ond
o
r
mor
e
laye
r
s
)
,
a
nd
f
inally
,
output
laye
r
(
the
las
t
laye
r
)
.
T
he
f
ir
s
t
laye
r
c
ons
is
ts
of
de
ter
mi
ne
d
node
s
(
ne
ur
ons
)
,
the
hidde
n
laye
r
is
c
ons
tr
uc
ted
of
s
ingl
e
or
mor
e
laye
r
s
,
whe
r
e
e
a
c
h
laye
r
uti
li
z
e
s
a
s
pe
c
if
ied
a
c
ti
va
ti
on
f
unc
ti
on
to
s
upply
it
s
output
the
ne
xt
laye
r
,
a
nd
the
las
t
(
out
put)
laye
r
c
ontains
a
t
a
s
pe
c
if
ied
number
o
f
node
s
a
nd
u
ti
li
z
e
s
a
n
a
c
ti
va
ti
on
f
unc
ti
on
[
19
-
24]
.
Ya
z
da
ni
M
.
&
R
a
s
s
a
f
i
A.
A.
,
[
25]
pr
e
s
e
nted
the
s
tudy
a
bout
e
va
luation
of
the
wa
r
ning
a
ppli
c
a
ti
on
map,
a
nd
the
s
a
ti
s
f
a
c
ti
on
of
the
dr
iver
to
thes
e
a
ppli
c
a
ti
ons
.
T
he
s
tudy
wa
s
im
pleme
nted
us
ing
32
dr
ier
s
in
the
two
-
wa
y
r
oa
d
in
the
nor
th
-
we
s
t
in
I
r
a
n.
T
he
s
tu
dy
s
hows
that
the
dr
iver
s
we
r
e
plea
s
e
d
with
a
lar
mi
ng
f
r
om
c
a
r
s
pe
a
ke
r
s
,
but
they
not
with
a
lar
ms
f
r
om
mobi
le
s
pe
a
ke
r
s
.
Ka
li
s
tr
a
tov
D.
[
26]
,
p
r
e
s
e
nted
a
global
mathe
matica
l
model
to
de
a
l
with
digi
tal
im
a
ge
s
c
omi
ng
f
r
om
the
tr
a
f
f
ic
c
ont
r
ol
s
it
e
in
the
tr
a
f
f
ic
c
onge
s
ti
on
withi
n
lar
ge
c
it
ies
,
whe
r
e
thi
s
model
c
a
n
ove
r
c
ome
the
pr
oblem
o
f
nois
e
on
the
tr
a
n
s
mi
tt
e
d
s
ignal,
the
F
our
ier
s
e
r
ies
c
a
n
a
ls
o
s
e
pa
r
a
te
tr
a
ns
mi
tt
e
d
s
ignal
va
r
iable
s
a
nd
us
e
them
with
c
omput
e
r
s
im
ulation.
Omidi
A,
e
t.
a
l
.
,
[
27
]
us
e
d
AD
S
s
of
twa
r
e
to
de
s
i
gn
the
low
nois
e
a
mpl
i
f
ier
c
i
r
c
uit
withi
n
the
low
f
r
e
que
nc
y
ba
nd.
T
his
c
ir
c
uit
wa
s
de
s
igned
a
nd
im
pleme
nted
in
wir
e
les
s
ne
twor
ks
a
nd
GPS
s
ys
tems
a
nd
pr
ove
d
to
be
e
f
f
icie
nt
a
nd
e
f
f
e
c
ti
ve
.
Khotbe
hs
a
r
a
E
.
&
S
a
f
a
r
i
H.
[
28
,
]
us
e
d
a
t
r
e
e
matr
ix
c
ontaining
pr
e
-
f
e
d
da
ta
a
bout
whe
r
e
to
buil
d
a
hos
pit
a
l,
a
nd
us
e
a
s
mar
t
da
ta
c
oll
e
c
ti
on
method
to
de
te
r
mi
ne
the
be
s
t
loc
a
ti
on
f
or
a
hos
pit
a
l
buil
ding
de
pe
nding
on
the
ps
yc
hologi
c
a
l
s
tate
of
the
pa
ti
e
nts
in
that
a
r
e
a
a
s
we
ll
a
s
the
a
v
a
il
a
bil
it
y
of
mate
r
ials
a
nd
e
quipm
e
nt
ne
c
e
s
s
a
r
y
in
the
tr
e
a
tm
e
nt.
Ga
mi
l
Y.
&
R
a
hman
I
.
A.
,
[
29
]
pr
e
s
e
nted
a
s
tudy
s
howing
the
ne
ga
ti
ve
im
pa
c
t
on
the
lac
k
of
c
omm
unica
ti
on
on
the
indus
tr
y
a
nd
a
f
ter
s
e
ve
r
a
l
que
s
ti
onna
ir
e
s
s
how
that
the
indus
tr
y
p
r
e
tends
a
n
d
of
f
e
r
s
s
omething
e
xc
e
ll
e
nt
whe
n
ther
e
is
c
omm
unica
ti
on
a
mong
the
owne
r
s
of
the
pr
ojec
t.
Our
wor
k
is
to
de
s
ign
a
s
im
ulation
s
o
f
twa
r
e
of
B
P
NN
a
nd
to
downloa
d
thi
s
s
of
twa
r
e
in
F
P
GA
,
the
wor
k
o
f
thi
s
s
im
u
lation
is
de
pe
nding
on
the
s
pe
e
d
o
f
dr
ivi
ng
whe
e
ls
of
v
e
hicle
the
s
uit
a
ble
va
lue
o
f
r
a
ti
o
s
pe
e
d
f
o
r
tr
a
n
s
mi
s
s
ion
ge
a
r
box
a
r
e
s
e
lec
ted.
F
o
r
il
lus
tr
a
ti
on
pur
pos
e
,
th
e
s
of
twa
r
e
c
hoos
e
s
the
s
pe
e
d
r
a
ti
o
(
1
)
if
the
e
nt
e
r
e
d
da
ta
va
lue
a
t
r
a
nge
(
00000000
–
00010100
)
,
whe
r
e
th
e
r
e
a
l
ve
locity
of
the
ve
hicle
is
r
a
nge
d
(
00
–
20
)
km/
h,
if
the
r
e
a
l
ve
locity
of
the
ve
hicle
is
r
a
nge
d
(
21
–
40)
km
/h,
the
e
nter
e
d
da
ta
to
the
s
of
twa
r
e
i
s
r
a
nge
d
(
00010101
–
00101000)
,
then
the
it
c
hoos
e
s
the
ve
l
oc
it
y
r
a
ti
o
nu
mber
(
2)
,
a
nd
s
o
on
.
T
he
pr
opos
e
d
B
P
NN
of
the
s
ys
tem
c
ons
tr
uc
ted
f
r
om
f
our
laye
r
s
:
the
input
laye
r
c
ons
is
ts
of
e
ight
ne
ur
ons
,
the
f
ir
s
t
laye
r
c
ompos
e
d
f
r
om
f
or
ty
-
f
i
ve
ne
ur
ons
,
s
e
c
ond
hidden
laye
r
s
c
ons
is
t
of
twe
nty
-
f
ive
ne
ur
ons
,
a
nd
the
f
or
th
(
las
t
laye
r
)
c
ons
is
ts
of
s
ix
n
e
ur
ons
.
T
he
M
AT
L
AB
s
of
twa
r
e
ha
s
be
e
n
uti
li
z
e
d
t
o
de
s
ign
a
nd
im
pleme
nt
the
p
r
opos
e
d
s
ys
tem.
T
he
r
e
s
ult
s
a
r
e
obtaine
d
by
us
ing
T
r
a
inl
m
(
)
M
a
tl
a
b
f
unc
ti
on
f
o
r
tr
a
ini
ng
the
B
P
NN
.
Us
ing
S
a
tl
ins
f
unc
ti
on
a
s
a
c
ti
va
ti
on
f
u
nc
ti
ons
f
or
hidden
laye
r
s
,
a
nd
S
a
tl
in
f
unc
ti
on
wa
s
us
e
d
f
or
li
ne
a
r
f
unc
ti
on
f
o
r
the
L
a
s
t
laye
r
.
T
he
pe
r
f
o
r
manc
e
of
the
pr
opos
e
d
s
of
twa
r
e
ha
s
be
e
n
e
va
luate
by
the
number
of
e
poc
h
of
tr
a
ini
ng
ne
twor
ks
a
nd
the
M
S
E
f
or
the
tr
a
ini
ng
a
nd
tes
ti
ng
pha
s
e
.
2.
T
HE
P
ROP
OS
E
D
S
I
M
UL
AT
I
ON
S
YST
E
M
DE
S
I
GN
T
he
pr
opos
e
d
s
ys
tem
de
pe
nde
d
mainly
a
t
the
B
P
NN
whic
h
wa
s
de
s
igned
by
Az
a
a
d
B
.
[
30]
,
a
n
d
w
e
make
s
ome
modi
f
ica
ti
on
a
t
the
ne
u
r
a
l
s
tr
uc
tur
e
,
t
he
s
tr
uc
tur
e
of
the
p
r
opos
e
d
B
P
NN
ha
s
be
e
n
il
lus
tr
a
ted
in
the
F
igur
e
1,
a
s
de
picte
d
be
ll
ow
.
T
he
de
s
ign
of
the
pr
opos
e
d
B
P
NN
s
tr
uc
tu
r
e
ha
s
pa
s
s
e
d
thr
oug
h
s
e
ve
r
a
l
s
tage
s
to
the
f
inal
de
s
ign
s
tatus
a
nd
is
a
c
c
e
pted
e
xpe
r
im
e
ntally.
T
he
r
e
is
no
pr
oblem
in
de
t
e
r
mi
ning
the
ne
ur
ons
c
ount
in
the
input
laye
r
or
in
the
ou
tput
laye
r
,
but
the
p
r
oblem
is
in
the
ne
ur
ons
c
ou
nt
to
be
pr
ovided
in
the
hidden
laye
r
s
a
s
we
ll
a
s
the
hidden
-
laye
r
s
c
ount
thems
e
lves
.
T
he
s
tr
uc
tur
e
c
ons
is
ts
of
f
our
laye
r
s
,
the
f
i
r
s
t
laye
r
na
med
input
laye
r
whe
r
e
it
c
ompos
e
d
f
r
om
e
ight
ne
ur
ons
(
node
s
)
,
it
r
e
c
e
ived
the
r
e
a
l
ve
locity
of
ve
hicle
whe
e
ls
in
binar
y
number
s
tyl
e
a
s
pr
e
s
e
nted
in
T
a
ble
1,
a
nd
the
e
ight
e
ntr
ies
of
B
P
NN
r
e
c
e
ive
it
s
da
ta
f
r
om
the
dr
iver
whe
e
l
s
pe
e
d,
whic
h
wa
s
e
nc
ode
d
int
o
e
ight
bit
s
b
inar
y
number
.
T
he
las
t
laye
r
na
me
wa
s
output
laye
r
,
it
c
ompos
e
d
f
r
om
s
ix
ne
ur
ons
,
e
a
c
h
on
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
A
ppli
c
ati
on
and
e
v
aluat
ion
of
the
ne
ur
al
ne
tw
or
k
i
n
ge
ar
box
(
Dhe
y
aa
Shahe
e
d
A
l
-
A
z
z
aw
i
)
21
r
e
pr
e
s
e
nt
the
s
pe
e
d
r
a
ti
on
of
the
ge
a
r
box,
f
or
e
xa
mpl
e
,
the
f
ir
s
t
ne
ur
on
de
voted
f
or
number
(
1)
r
a
ti
o
s
pe
e
d
of
ge
a
r
box,
the
s
e
c
ond
ne
ur
on
f
or
numbe
r
(
2
)
r
a
ti
o
s
pe
e
d
of
ge
a
r
box,
a
nd
s
o
on.
F
or
e
a
c
h
e
xe
c
uti
on
ther
e
mus
t
be
one
ne
ur
on
is
a
c
ti
va
ted
to
one
a
nd
a
ll
the
other
de
a
c
ti
va
ted
to
z
e
r
os
,
the
S
a
tl
in
f
unc
ti
on
us
e
d
a
s
li
ne
a
r
f
unc
ti
on
.
F
igur
e
1.
S
t
r
uc
tur
e
o
f
the
p
r
opos
e
d
B
P
NN
T
a
ble
1.
C
ode
s
of
the
p
r
opos
e
d
s
tr
uc
tur
e
V
e
lo
c
it
y R
a
ti
o
R
e
a
l
V
e
lo
c
it
y km/
h
I
nput
da
ta
P
r
opos
e
d S
im
ul
a
ti
on
S
of
twa
r
e
O
ut
put
s
A
B
C
D
E
F
1
00
-
20
00000000
-
00010100
1
0
0
0
0
0
2
21
-
40
00010101
-
00101000
0
1
0
0
0
0
3
41
-
60
00101001
-
00111100
0
0
1
0
0
0
4
61
-
100
00111101
-
01100100
0
0
0
1
0
0
5
101
-
130
01100101
-
10000001
0
0
0
0
1
0
6
>
=
131
10000010
-
11111111
0
0
0
0
0
1
At
f
ir
s
t,
the
de
s
ign
wa
s
e
xpe
r
im
e
ntally
ba
s
e
d
on
the
c
ompos
it
ion
of
the
ne
twor
k
c
ons
is
ti
ng
of
one
hidden
laye
r
a
nd
then
e
a
c
h
e
xpe
r
im
e
ntal
s
tage
,
the
number
of
c
e
ll
s
f
or
mi
ng
that
laye
r
is
s
e
lec
ted
f
r
o
m
the
c
e
ll
to
70
c
e
ll
s
.
At
e
a
c
h
s
tage
,
the
e
va
luation
is
ba
s
e
d
on
the
va
lue
of
M
S
E
a
nd
the
number
o
f
e
po
c
hs
,
s
e
e
T
a
ble
2
.
T
he
n
the
pr
opos
e
d
B
P
NN
wa
s
r
e
c
ons
tr
u
c
ted
f
r
om
two
hidden
laye
r
s
,
a
nd
r
e
va
luate
d
to
d
e
ter
mi
ne
the
mos
t
a
ppr
opr
iate
number
of
ne
ur
ons
de
pe
ndin
g
on
the
va
lue
of
M
S
E
f
o
r
the
tr
a
ini
ng
a
nd
tes
ti
ng
s
a
mpl
e
s
a
nd
the
number
of
e
poc
hs
in
e
a
c
h
tr
a
ini
ng
pha
s
e
is
s
e
lec
ted,
s
e
e
T
a
ble
3
.
At
las
t,
the
s
uit
a
ble
s
tr
uc
tur
e
of
B
P
NN
wa
s
c
hos
e
n,
a
nd
c
ons
is
ted
o
f
f
our
laye
r
s
,
th
e
f
ir
s
t
a
nd
las
t
laye
r
wa
s
mentioned
a
bove
a
nd
it
is
not
our
pr
oblem
in
our
r
e
s
e
a
r
c
h.
T
he
f
ir
s
t
hidden
laye
r
c
ompos
e
d
f
r
om
f
or
ty
-
f
ive
ne
ur
ons
,
the
s
e
c
ond
hid
de
n
laye
r
c
ons
tr
uc
ted
f
r
om
twe
nty
-
f
ive
ne
u
r
ons
,
a
nd
us
e
d
S
a
tl
ins
f
unc
ti
on
a
s
li
ne
a
r
f
unc
ti
on
f
or
e
a
c
h
hidden
lay
e
r
.
3.
T
HE
I
M
P
L
E
M
E
NT
AT
I
ON
OF
T
HE
P
ROP
OS
E
D
S
I
M
UL
AT
I
ON
S
YST
E
M
T
he
pr
opos
e
d
B
P
NN
im
pleme
nted
us
ing
M
AT
L
AB
s
of
twa
r
e
,
we
us
e
d
T
r
a
inl
m
f
unc
ti
on
(
L
e
ve
nbe
r
g
-
M
a
r
qua
r
dt
tr
a
ini
ng)
a
s
tr
a
ini
ng
a
lgo
r
it
hm.
W
he
n
we
im
pleme
nt
the
p
r
opos
e
d
s
of
twa
r
e
in
a
c
omput
e
r
,
the
block
of
the
p
r
opos
e
d
s
im
ulation
B
P
NN
will
be
s
hown
a
nd
il
lus
tr
a
ted
i
n
F
igur
e
2,
i
t
ha
s
e
ight
input
butt
ons
that
r
e
pr
e
s
e
nt
the
ve
locity
of
ve
hi
c
le,
the
da
ta
c
a
n
be
e
nter
e
d
to
the
s
im
ulation
s
ys
tem
by
a
c
ti
va
ti
ng
(
c
li
c
king
a
t)
the
butt
on
f
o
r
1
e
ntr
y
,
a
nd
de
a
c
ti
va
ti
ng
(
r
e
lea
s
ing)
the
butt
on
f
or
z
e
r
o
e
ntr
y.
T
he
s
im
ulation
ha
s
s
ix
output
butt
ons
,
whe
r
e
on
ly
one
butt
on
will
be
a
c
ti
ve
(
s
e
t
to
one
)
f
or
the
r
e
quir
e
d
ve
locity
a
nd
the
othe
r
butt
ons
will
be
de
a
c
ti
va
te
(
r
e
s
e
t
to
z
e
r
o)
.
W
he
n
the
s
im
ulation
c
ompl
e
te,
the
r
e
s
ult
will
be
one
o
f
ou
tput
bu
tt
ons
s
e
t
to
one
a
nd
the
othe
r
r
e
s
e
t
to
z
e
r
o.
T
o
be
gin
the
s
im
ulation,
we
s
e
lec
t
th
e
de
s
ir
e
d
input
butt
ons
,
f
or
ins
tanc
e
,
we
wa
nt
to
e
nter
the
ve
locity
va
lue
(
00111100)
,
whic
h
it
e
qua
l
to
(
60
km)
a
s
a
r
e
a
l
ve
locity
va
lue
in
de
c
im
a
l
r
e
pr
e
s
e
nta
ti
on,
c
li
c
k
a
t
the
butt
ons
3,
4,
5
a
nd
6,
then
c
li
c
k
a
t
th
e
B
P
NN
S
im
ulation
B
utt
on,
s
e
e
F
igur
e
3.
Ne
w
block
of
s
im
ulation
will
be
a
ppe
a
r
on
the
c
o
mput
e
r
s
c
r
e
e
n,
s
e
e
F
igur
e
4,
thi
s
block
il
lus
tr
a
tes
the
c
onne
c
ti
on
of
B
P
NN
laye
r
s
,
by
c
li
c
king
a
t
the
f
ir
s
t
hidden
laye
r
butt
on,
the
s
umm
a
ti
on
of
mul
ti
pli
c
a
ti
on
we
ight
s
by
input
s
va
lues
s
tar
ted,
a
nd
the
r
e
s
ul
ts
will
be
the
e
ntr
ies
va
lues
f
or
the
s
e
c
ond
hidden
laye
r
.
Now
,
I
nput
la
ye
r
(
8)
node
s
)
Output
laye
r
(
6)
node
s
)
T
wo
hidden
laye
r
s
(
45
f
o
r
the
f
ir
s
t
h
idden
laye
r
a
nd
25
node
s
f
o
r
e
a
c
h
laye
r
)
We
ight
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
1
,
F
e
br
ua
r
y
2020
:
19
-
29
22
we
c
li
c
k
a
t
the
s
e
c
ond
hidden
laye
r
butt
on
,
the
r
e
s
ult
will
be
a
c
ti
va
ti
on
of
bu
tt
on
C
,
that
mea
ns
the
ge
a
r
box
s
pe
e
d
r
a
ti
o
is
3,
s
e
e
F
igur
e
5
.
T
a
ble
2.
R
e
s
ult
s
of
e
poc
h
a
nd
M
S
E
f
o
r
im
pleme
nti
ng
the
B
P
NN
c
ons
is
ts
of
s
ingl
e
hidden
laye
r
with
many
ne
ur
ons
number
N
o. of
N
e
ur
ons
i
n
f
ir
s
t
hi
dde
n
la
ye
r
E
poc
h
M
S
E
i
n
tr
a
in
in
g
pha
s
e
M
S
E
i
n
te
s
ti
ng
pha
s
e
1
8
0.327
0.3507
2
6
0.308
0.2991
3
54
0.236
0.2521
4
8
0.234
0.2798
5
35
0.226
0.2481
6
45
0.226
0.2676
7
16
0.220
0.2619
8
17
0.224
0.2533
9
16
0.228
0.2570
10
15
0.232
0.2644
11
15
0.220
0.2559
12
16
0.224
0.2718
13
11
0.212
0.2505
14
14
0.240
0.2651
15
11
0.210
0.2400
16
11
0.218
0.2630
17
12
0.218
0.2508
18
14
0.224
0.2615
19
12
0.212
0.2511
20
11
0.218
0.2462
21
13
0.212
0.2707
22
13
0.216
0.2803
23
11
0.228
0.2744
24
6
0.216
0.2495
25
12
0.211
0.2606
26
12
0.212
0.3079
27
10
0.214
0.2488
28
15
0.218
0.2743
N
o. of
N
e
ur
ons
i
n
f
ir
s
t
hi
dde
n l
a
ye
r
E
poc
h
M
S
E
i
n
tr
a
in
in
g
pha
s
e
M
S
E
i
n
te
s
ti
ng
pha
s
e
29
10
0.244
0.3047
30
14
0.211
0.2470
31
10
0.220
0.2503
32
12
0.218
0.2434
33
11
0.212
0.2646
34
10
0.211
0.2398
35
10
0.214
0.2453
36
11
0.214
0.2521
37
13
0.212
0.2603
38
12
0.214
0.2768
39
11
0.212
0.2527
40
12
0.216
0.2584
45
5
0.210
0.220
50
9
0.214
0.2472
55
14
0.214
0.2466
60
8
0.216
0.2341
65
9
0.214
0.2428
70
11
0.218
0.2627
T
a
ble
3.
R
e
s
ult
s
of
e
poc
h
a
nd
M
S
E
f
o
r
im
pleme
nti
ng
the
B
P
NN
c
ons
is
ts
of
double
h
idden
-
laye
r
s
,
1
s
t
hidden
-
laye
r
f
ixed
a
t
45
ne
ur
ons
,
a
nd
with
man
y
ne
ur
ons
number
f
or
the
2
nd
hidden
-
laye
r
N
o. of
N
e
ur
ons
i
n
s
e
c
ond hidde
n l
a
ye
r
E
poc
h
M
S
E
i
n
tr
a
in
in
g
pha
s
e
M
S
E
i
n
te
s
ti
ng
pha
s
e
1
52
0.0982
0.1422
2
44
0.0877
0.1787
3
52
0.0502
0.1599
4
84
0.0414
0.1268
5
38
0.0472
0.1695
6
40
0.156e
-
32
0.1319
7
47
1.64e
-
33
0.1246
8
31
7.36e
-
33
0.1960
9
27
9.12e
-
33
0.1130
10
58
7.12e
-
22
0.1357
11
53
8.20e
-
32
0.1116
12
20
1.03e
-
32
0.0936
13
26
1.03e
-
32
0.1950
14
35
8.51e
-
33
0.1024
15
30
1.51e
-
32
0.1512
16
17
1.49e
-
32
0.1287
17
18
7.34e
-
33
0.1433
18
23
6.87e
-
33
0.1540
19
18
1.09e
-
32
0.1407
20
19
1.14e
-
32
0.1969
21
19
1.75e
-
32
0.1204
22
23
1.00e
-
32
0.0992
23
19
1.13e
-
32
0.1538
24
23
9.42e
-
33
0.1187
25
12
1.85e
-
32
0.0182
26
15
1.62e
-
32
0.1195
20
19
1.06e
-
32
0.1061
28
16
1.86e
-
32
0.1320
N
o. of
N
e
ur
ons
i
n
s
e
c
ond hidde
n l
a
ye
r
E
poc
h
M
S
E
i
n
tr
a
in
in
g
pha
s
e
M
S
E
i
n
te
s
ti
ng
pha
s
e
29
18
1.52e
-
32
0.1298
30
20
1.94e
-
32
0.1558
31
16
2.00e
-
32
0.1394
32
15
1.86e
-
32
0.1263
33
16
1.63e
-
32
0.1181
34
17
1.65e
-
32
0.1018
35
17
2.03e
-
32
0.1073
36
18
2.65e
-
32
0.1323
37
20
1.68e
-
32
0.1213
38
21
1.59e
-
32
0.1804
39
16
2.05e
-
32
0.1412
40
16
2.78e
-
32
0.1132
45
15
2.31e
-
32
0.1115
50
15
2.88e
-
32
0.1596
55
24
2.91e
-
32
0.1368
60
17
3.75e
-
32
0.1254
65
15
3.59e
-
32
0.1508
70
17
3.37e
-
32
0.1728
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
A
ppli
c
ati
on
and
e
v
aluat
ion
of
the
ne
ur
al
ne
tw
or
k
i
n
ge
ar
box
(
Dhe
y
aa
Shahe
e
d
A
l
-
A
z
z
aw
i
)
23
F
igur
e
2.
B
lock
of
the
p
r
opos
e
d
s
im
ulation
s
ys
tem
F
igur
e
3.
T
he
s
im
ulation
block
a
f
ter
c
li
c
king
the
de
s
ir
e
d
input
a
nd
B
P
NN
s
im
ulation
butt
ons
F
igur
e
4.
B
lock
of
s
im
ulation
s
hows
the
B
P
NN
laye
r
s
F
igur
e
5.
B
lock
of
s
im
ulation
a
f
ter
the
e
xe
c
uti
on
c
ompl
e
ted
4.
RE
S
UL
T
S
A
ND
DI
S
CU
S
S
I
ON
As
we
mentioned
a
t
the
pr
e
vious
s
e
c
ti
on,
the
s
tr
uc
tur
e
of
the
pr
opos
e
d
B
P
NN
c
ons
tr
a
ined
a
t
the
s
e
lec
ti
ng
the
c
ount
of
hidden
-
laye
r
s
a
s
we
ll
a
s
t
he
c
ount
of
it
s
ne
ur
ons
.
T
he
pr
opos
e
d
ne
twor
k
wa
s
tr
a
ined
a
t
31
input
s
a
mpl
e
s
,
tes
ted
a
t
151
input
s
a
mpl
e
s
,
a
nd
the
e
va
luation
of
the
pr
opos
e
d
s
tr
uc
tur
e
de
pe
nde
d
on
the
va
lue
of
M
S
E
f
or
the
tr
a
ined
a
nd
tes
ted
s
a
mpl
e
s
in
a
ddit
ion
to
the
number
of
e
poc
h
o
f
the
t
r
a
ini
ng
pha
s
e
.
W
e
us
e
d
two
pha
s
e
s
of
de
s
igni
ng
pr
os
e
s
s
e
s
,
the
f
i
r
s
t
one
im
pleme
nted
the
s
tr
uc
tur
e
of
ne
ur
a
l
c
ons
is
ts
of
one
s
ingl
e
hidden
laye
r
only,
a
nd
in
the
s
e
c
ond
pha
s
e
we
im
pleme
nted
the
s
tr
uc
tur
e
o
f
ne
ur
a
l
ne
t
c
ons
is
ts
of
two
hidden
laye
r
s
.
E
xpe
r
im
e
ntally
,
in
the
f
ir
s
t
pha
s
e
,
we
s
e
lec
ted
f
or
ty
-
f
ive
ne
u
r
ons
f
o
r
the
f
ir
s
t
hidden
laye
r
,
a
nd
the
e
poc
h
a
nd
S
M
E
f
or
the
tr
a
ini
ng
a
nd
tes
ti
ng
s
a
mpl
e
s
we
r
e
the
be
s
t
va
lue
a
t
that
numbe
r
of
ne
ur
ons
.
S
e
e
F
igur
e
6,
to
noti
c
e
the
s
tr
uc
tur
e
of
the
pr
opos
e
d
B
P
NN
with
s
ingl
e
hidden
laye
r
a
nd
the
e
poc
h
va
lue
f
or
us
ing
45
ne
ur
ons
a
t
the
f
ir
s
t
h
idden
laye
r
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
1
,
F
e
br
ua
r
y
2020
:
19
-
29
24
F
igur
e
6.
M
e
n
u
s
hows
the
va
lue
of
e
poc
h
a
t
t
r
a
ini
n
g
pha
s
e
f
or
B
P
NN
c
ons
is
ts
of
s
ingl
e
hidden
laye
r
with
45
ne
ur
ons
F
igur
e
7
il
lus
tr
a
te
the
p
lot
o
f
Gr
a
dient,
M
omentums
a
nd
va
li
da
ti
on
c
he
c
ks
whe
n
tr
a
ini
ng
the
B
P
N
N
with
s
ingl
e
hidden
laye
r
o
f
45
ne
ur
ons
.
F
igu
r
e
8
il
lus
tr
a
tes
the
plot
of
tr
a
ini
ng
pe
r
f
o
r
manc
e
f
or
th
e
B
P
NN
with
s
ingl
e
hidden
laye
r
o
f
45
ne
ur
ons
,
a
nd
the
be
s
t
va
li
da
ti
on
pe
r
f
o
r
manc
e
va
lue
wa
s
a
t
the
1
s
t
e
poc
h.
F
igur
e
9
s
hows
s
how
the
plot
of
r
e
gr
e
s
s
ion
dur
i
ng
the
t
r
a
ini
ng
o
f
the
B
P
NN
of
s
ingl
e
d
hidden
l
a
ye
r
with
45
ne
ur
ons
.
F
igur
e
7.
P
lot
of
gr
a
dient,
mom
e
ntum
a
nd
va
li
da
ti
on
c
he
c
ks
f
or
the
t
r
a
ini
ng
B
P
NN
c
ons
is
ts
of
s
ingl
e
hidden
laye
r
with
45
ne
ur
ons
F
igur
e
8.
P
lot
of
pe
r
f
o
r
manc
e
f
or
t
r
a
ini
ng
the
B
P
NN
with
s
ingl
e
laye
r
of
45
ne
ur
ons
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
A
ppli
c
ati
on
and
e
v
aluat
ion
of
the
ne
ur
al
ne
tw
or
k
i
n
ge
ar
box
(
Dhe
y
aa
Shahe
e
d
A
l
-
A
z
z
aw
i
)
25
F
igur
e
9.
P
lot
of
r
e
gr
e
s
s
ion
f
or
tr
a
ini
ng
B
P
NN
wit
h
s
ingl
e
hidden
laye
r
with
45
ne
ur
ons
All
the
a
bove
thr
e
e
f
igu
r
e
s
of
plot
ing
,
s
hown
the
45
ne
ur
ons
in
the
B
P
NN
is
the
be
s
t
c
hoice
,
whe
r
e
the
va
lue
of
the
tr
a
ini
ng
e
poc
h
wa
s
5,
a
nd
the
va
lue
of
M
S
E
f
o
r
the
t
r
a
ini
ng
a
nd
tes
ti
ng
pe
r
f
o
r
ma
nc
e
wa
s
0.
210
a
nd
0.
220
r
e
s
pe
c
ti
ve
ly.
T
he
n
towa
r
d
mo
r
e
e
nha
nc
ing
the
ne
twor
k,
be
ga
n
to
the
s
e
c
ond
pha
s
e
of
the
de
s
igni
ng
pr
oc
e
s
s
,
we
r
e
c
ons
tr
uc
ted
the
s
tr
uc
t
ur
e
of
the
ne
ur
a
l
ne
t
by
a
dding
a
nother
one
hidden
-
laye
r
to
be
th
e
ne
twor
k
c
ons
is
ts
of
two
hidden
-
laye
r
s
.
E
xpe
r
im
e
ntally,
we
dis
c
ove
r
e
d
that
the
be
s
t
ne
ur
ons
c
ounts
wa
s
25
ne
ur
ons
in
the
s
e
c
ond
hidden
lay
e
r
,
a
s
il
lus
tr
a
ted
by
the
be
s
t
va
lue
of
E
poc
h
a
nd
M
S
E
wa
s
c
hos
e
n
whe
n
the
B
P
NN
c
ons
is
ted
of
two
h
idden
-
laye
r
s
,
45,
a
n
d
25
ne
ur
ons
in
the
1
st
a
nd
2
nd
hidd
e
n
-
laye
r
r
e
s
pe
c
ti
ve
ly,
s
e
e
F
ig
ur
e
10,
whic
h
it
s
hows
menu
that
dis
plays
the
s
tr
uc
tur
e
a
nd
the
e
poc
h
va
lue
e
q
ua
l
to
12
whe
n
tr
a
ini
ng
the
pr
opos
e
d
B
P
NN
c
ons
is
ted
of
tw
o
hidden
-
laye
r
s
,
45,
a
nd
25
ne
ur
ons
r
e
s
pe
c
ti
ve
ly
.
F
igur
e
11
,
s
hows
the
plot
of
Gr
a
dient,
M
omentu
m
a
nd
va
li
da
ti
on
c
he
c
ks
f
or
the
tr
a
ini
ng
B
P
NN
c
ons
is
ts
of
two
hidden
laye
r
,
1
st
hidden
-
laye
r
f
ixed
a
t
the
45
ne
ur
ons
,
a
nd
the
2
nd
hidden
-
laye
r
s
e
t
to
25
ne
ur
ons
,
a
nd
the
be
s
t
va
lues
wa
s
a
t
the
12
th
e
p
oc
h.
F
igu
r
e
12
,
s
hows
the
p
lot
of
pe
r
f
o
r
manc
e
f
o
r
tr
a
ini
ng
the
B
P
NN
with
two
hidden
-
laye
r
s
,
45
a
nd
25
ne
ur
ons
f
or
the
f
ir
s
t
a
nd
s
e
c
ond
hidden
-
laye
r
r
e
s
pe
c
ti
ve
ly,
a
nd
the
be
s
t
va
li
da
ti
on
pe
r
f
or
manc
e
va
lue
wa
s
0.
15258
a
t
the
3
rd
e
poc
h
.
F
igur
e
13
,
s
hows
the
plot
of
r
e
gr
e
s
s
ion
f
or
t
r
a
ini
ng
B
P
NN
with
two
h
idden
-
laye
r
s
,
45
a
nd
25
ne
ur
ons
f
or
the
1
st
a
nd
2
nd
h
idden
-
laye
r
r
e
s
pe
c
ti
ve
ly.
F
inally
a
nd
e
xpe
r
im
e
ntally,
f
r
om
a
ll
the
a
bove
t
hr
e
e
mentioned
f
igur
e
s
of
plot
ing
,
we
noti
c
e
that
the
be
s
t
va
lues
f
or
e
poc
h
e
qua
l
to
12,
a
nd
va
lue
of
M
S
E
f
o
r
tr
a
ini
ng
a
nd
tes
ti
ng
pe
r
f
or
manc
e
wa
s
e
qua
l
to
1.
85e
-
32
a
nd
e
qua
l
to
0.
0181r
e
s
pe
c
ti
ve
ly,
we
r
e
e
xis
t
in
the
B
P
NN
whic
h
s
tr
uc
tur
e
d
f
r
om
two
hidd
e
n
laye
r
s
(
45
a
nd
25
ne
ur
ons
in
the
1
st
a
nd
2
nd
hidden
-
laye
r
r
e
s
pe
c
ti
ve
ly)
.
C
ompar
ing
with
p
r
e
v
ious
wor
k
of
other
r
e
s
e
a
r
c
he
r
s
,
f
or
ins
tanc
e
,
Az
z
a
d
B
.
S
a
e
e
d
[
30]
,
ha
s
de
s
igned
a
nd
pr
e
s
e
nted
a
s
im
ulation
s
ys
tem
a
s
il
lus
tr
a
ted
in
F
igur
e
14,
a
s
s
hown
in
thi
s
f
igur
e
,
the
be
s
t
va
lue
of
M
S
E
e
qua
l
to
3.
9392e
-
25
is
r
e
a
c
he
d
a
t
E
poc
h
1
5.
R
a
mya
J
.
e
t
a
l
[
31]
a
n
d
W
e
i
L
.
e
t
a
l
[
32
]
ha
ve
de
s
igned
a
nd
ga
ve
a
s
im
ulation
s
ys
tem
a
s
il
lus
tr
a
ted
in
F
igu
r
e
15
a
nd
F
igu
r
e
16
r
e
s
pe
c
ti
ve
ly,
whic
h
th
e
y
s
hown
that
the
be
s
t
va
lue
of
M
S
E
e
qua
l
to
4
.
3
515e
-
14
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
1
,
F
e
br
ua
r
y
2020
:
19
-
29
26
a
nd
e
qua
l
to
1.
3205e
-
15
we
r
e
r
e
a
c
he
d
a
t
the
e
poc
h
24
a
nd
36
r
e
s
pe
c
ti
ve
ly.
T
hus
,
f
r
om
the
other
pr
e
vious
wor
ks
,
we
c
onc
lude
that
our
pr
opos
e
d
B
P
NN
ge
ts
the
be
s
t
r
e
s
ult
s
in
the
tr
a
ini
ng
tes
ti
ng
pe
r
f
o
r
manc
e
va
lue,
whe
r
e
the
be
s
t
S
M
E
va
lue
wa
s
e
qua
l
to
1
.
85e
-
32
a
t
the
12
th
e
poc
h
o
f
t
r
a
ini
ng.
F
igur
e
10
.
M
e
nu
dis
plays
the
va
lue
of
e
poc
h
a
t
tr
a
i
ning
pha
s
e
f
or
B
P
NN
c
ons
is
ts
of
two
hidden
laye
r
s
with
45
a
nd
25
ne
ur
ons
f
o
r
the
f
i
r
s
t
a
nd
s
e
c
ond
hid
de
n
laye
r
s
r
e
s
pe
c
ti
ve
ly
F
igur
e
11.
P
lot
of
Gr
a
dient,
M
omentum
a
nd
va
li
da
ti
on
c
he
c
ks
f
or
the
t
r
a
ini
ng
B
P
NN
c
ons
is
ts
of
two
hidd
e
n
laye
r
s
,
1
st
h
idden
-
laye
r
f
ixed
a
t
the
45
ne
u
r
ons
,
a
n
d
2
nd
hidden
-
laye
r
s
e
t
to
25
ne
ur
ons
F
igur
e
12.
P
lot
of
pe
r
f
o
r
manc
e
f
or
tr
a
ini
ng
the
B
P
NN
with
two
hidden
-
laye
r
s
,
45
a
nd
25
ne
ur
ons
f
or
the
1
st
a
nd
2
nd
hidden
-
laye
r
r
e
s
pe
c
ti
ve
ly
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
A
ppli
c
ati
on
and
e
v
aluat
ion
of
the
ne
ur
al
ne
tw
or
k
i
n
ge
ar
box
(
Dhe
y
aa
Shahe
e
d
A
l
-
A
z
z
aw
i
)
27
F
igur
e
13.
P
lot
of
r
e
gr
e
s
s
ion
f
or
tr
a
ini
ng
B
P
NN
wi
th
two
hidden
-
la
ye
r
s
,
45
a
nd
25
ne
ur
ons
f
or
the
1
st
a
nd
2
nd
hidden
-
laye
r
r
e
s
pe
c
ti
ve
ly
F
igur
e
14.
P
lot
of
a
z
z
a
d
B
.
S
a
e
e
d
pe
r
f
or
manc
e
S
M
E
r
e
s
ult
s
F
igur
e
15.
P
lot
of
R
a
mya
J
.
e
t
a
l.
pe
r
f
or
manc
e
S
M
E
r
e
s
ult
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
1693
-
6930
T
E
L
KO
M
NI
KA
T
e
lec
omm
un
C
omput
E
l
C
ontr
o
l
,
Vol.
18
,
No
.
1
,
F
e
br
ua
r
y
2020
:
19
-
29
28
F
igur
e
1
6
.
P
lot
of
W
e
i
L
.
e
t
a
l.
pe
r
f
o
r
manc
e
S
M
E
r
e
s
ult
s
5.
CONC
L
USI
ON
T
he
main
r
ole
of
the
int
e
l
li
ge
nt
s
ys
tems
of
pr
ogr
a
med
a
utom
a
ti
c
tr
a
ns
mi
s
s
ion
ge
a
r
box
is
ba
s
e
on
the
pe
r
f
o
r
manc
e
of
the
s
of
twa
r
e
r
e
s
ult
s
that
i
ns
talled
f
or
that
pur
pos
e
,
a
nd
in
the
ne
ur
a
l
n
e
twor
ks
,
the
pe
r
f
or
manc
e
de
pe
nds
on
the
M
S
E
tr
a
ini
ng
a
n
d
tes
ti
ng
p
e
r
f
o
r
manc
e
r
e
s
ult
s
,
a
nd
on
the
ti
me
of
tr
a
ini
ng
whic
h
r
e
pr
e
s
e
nted
in
e
poc
h
o
f
t
r
a
ini
ng.
I
n
our
p
r
o
pos
e
d
s
ys
tem,
the
B
P
NN
wa
s
c
ons
tr
uc
ted
with
de
t
e
r
mi
ne
d
number
of
hidden
laye
r
s
a
nd
number
of
ne
ur
ons
that
a
ble
the
s
ys
tem
r
e
a
c
he
d
th
e
be
s
t
pe
r
f
or
manc
e
r
e
s
ult
s
.
T
he
e
xpe
r
im
e
ntal
r
e
s
ult
s
s
how
the
be
s
t
pe
r
f
or
m
a
nc
e
S
M
E
va
lues
a
r
e
1.
85e
-
32
a
nd
e
qua
l
to
0.
0181
f
or
tr
a
ini
ng
a
nd
tes
ti
ng
pe
r
f
o
r
manc
e
r
e
s
pe
c
ti
ve
ly
wit
h
e
poc
h
number
a
t
12
.
I
n
the
f
utu
r
e
,
I
pr
e
f
e
r
to
us
e
s
ome
methods
of
A
I
in
the
int
e
ll
igent
s
ys
tem
s
uc
h
a
s
ge
ne
ti
c
a
lgor
it
hm
ins
tea
d
of
ne
ur
a
l
ne
twor
ks
that
ma
y
s
hows
ne
w
good
pe
r
f
o
r
manc
e
r
e
s
ult
s
.
RE
F
E
RE
NC
E
S
[1
]
Bag
ameri
N
.
,
V
arg
a
B.
&
Mo
l
d
o
v
an
u
D
.
,
“Co
mp
a
rat
i
v
e
A
n
a
l
y
s
i
s
o
f
A
u
t
o
mat
i
c
T
ran
s
i
s
s
i
o
n
n
a
n
d
Man
u
a
l
T
ran
s
i
s
s
i
o
n
Beh
a
v
i
o
u
r
o
n
t
h
e
W
o
rl
d
w
i
d
e
H
a
rmo
n
i
ze
d
L
i
g
h
t
D
u
t
y
T
e
s
t
C
y
cl
e,
”
M
A
T
E
C
W
e
b
o
f
C
o
n
f
er
e
n
ces
,
v
o
l
.
1
8
4
,
n
o
.
1
2
,
J
an
u
ary
2
0
1
8
.
[2
]
D
ark
o
St
a
n
o
j
ev
i
,
V
l
a
d
i
m
i
r
Sp
a
s
o
j
ev
i
,
Ig
o
r
St
e
v
an
o
v
i
,
an
d
A
l
ek
s
an
d
ar
N
e
d
i
,
"
T
h
e
Co
n
t
em
p
o
rar
y
A
u
t
o
ma
t
i
c
G
earb
o
x
e
s
-
rev
i
ew
o
f
t
h
e
cu
rre
n
t
s
t
a
t
e
an
d
i
n
t
er
p
ret
a
t
i
o
n
o
f
ad
v
an
t
ag
es
an
d
d
i
s
a
d
v
a
n
t
a
g
e
s
o
f
t
h
ei
r
u
s
e
w
i
t
h
re
s
p
ect
t
o
v
e
h
i
c
l
e
p
erf
o
rman
ce
an
d
t
err
i
fi
c
s
afet
y
,
"
Jo
u
r
n
a
l
o
f
A
p
p
l
i
e
d
E
n
g
i
n
eer
i
n
g
S
c
i
en
ce
,
v
o
l
.
1
1
,
n
o
.
2
,
p
p
.
89
-
9
7
,
2
0
1
3
.
[3
]
A
n
d
rew
Mo
s
k
a
l
i
k
,
A
aro
n
H
u
l
a,
D
an
i
el
Bar
b
a,
an
d
J
o
h
n
K
arg
u
l
,
"
In
v
e
s
t
i
g
a
t
i
n
g
t
h
e
E
ffec
t
o
f
A
d
v
a
n
ced
A
u
t
o
m
at
i
c
T
ran
s
mi
s
s
i
o
n
s
o
n
Fu
e
l
Co
n
s
u
mp
t
i
o
n
U
s
i
n
g
V
e
h
i
c
l
e
T
es
t
i
n
g
an
d
Mo
d
el
i
n
g
,
"
S
A
E
In
t
e
r
n
a
t
i
o
n
a
l
Jo
u
r
n
a
l
o
f
E
n
g
i
n
es
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
1
9
1
6
-
1
9
2
8
,
2
0
1
6
.
[4
]
N
g
o
V
.
D
.
,
H
o
fman
T
.
,
St
ei
n
b
u
ch
M.
&
Serraren
s
A
.
,
“
G
ear
Sh
i
ft
Map
D
es
i
g
n
Met
h
o
d
o
l
o
g
y
fo
r
A
u
t
o
mat
i
v
e
T
ran
s
i
s
s
i
o
n
s
,
”
i
n
P
r
o
ceed
i
n
g
s
o
f
t
h
e
In
s
t
i
t
u
t
i
o
n
o
f
M
ech
a
n
i
ca
l
E
n
g
i
n
eer
s
P
a
r
t
D
Jo
u
r
n
a
l
o
f
A
u
t
o
m
o
b
i
l
e
E
n
g
i
n
eer
i
n
g
,
v
o
l
.
2
2
8
,
n
o
.
1
,
p
p
.
5
0
-
7
2
,
J
a
n
u
ar
y
2
0
1
3
.
[5
]
W
en
c
h
en
S
h
en
,
H
u
i
l
o
n
g
Y
u
,
Y
u
h
u
i
H
u
,
a
n
d
J
u
n
q
i
a
n
g
X
i
,
"
O
p
t
i
m
i
zat
i
o
n
o
f
S
h
i
f
t
Sc
h
ed
u
l
e
f
o
r
H
y
b
ri
d
E
l
ec
t
ri
c
V
eh
i
cl
e
w
i
t
h
A
u
t
o
mat
e
d
Man
u
a
l
T
ra
n
s
m
i
s
s
i
o
n
,
"
E
n
e
r
g
i
es
Jo
u
r
n
a
l
,
v
o
l
.
9
,
n
o
.
3
,
p
p
.
1
-
1
1
,
March
2
0
1
6
.
[6
]
J
i
a
n
Y
ao
,
L
i
Ch
en
,
an
d
Fen
g
y
u
L
i
u
,
"
E
x
p
eri
me
n
t
al
St
u
d
y
o
n
Imp
ro
v
emen
t
i
n
t
h
e
Sh
i
ft
Q
u
al
i
t
y
fo
r
an
A
u
t
o
ma
t
i
c
T
ran
s
mi
s
s
i
o
n
U
s
i
n
g
a
M
o
t
o
r
-
D
r
i
v
e
n
W
e
d
g
e
C
l
u
t
ch
,
"
P
r
o
cee
d
i
n
g
s
o
f
t
h
e
In
s
t
i
t
u
t
i
o
n
o
f
M
ec
h
a
n
i
c
a
l
E
n
g
i
n
ee
r
s
P
a
r
t
D
Jo
u
r
n
a
l
o
f
A
u
t
o
m
o
b
i
l
e
E
n
g
i
n
eer
i
n
g
,
v
o
l
.
2
2
8
,
n
o
.
6
,
p
p
.
6
6
3
-
6
7
3
,
A
p
ri
l
2
0
1
4
.
[7
]
L
o
n
g
-
C
h
an
g
H
s
i
e
h
,
an
d
H
s
i
u
-
C
h
en
T
a
n
g
,
"
T
h
e
In
n
o
v
at
e
D
es
i
g
n
o
f
A
u
t
o
ma
t
i
c
T
ran
s
mi
s
s
i
o
n
fo
r
E
l
e
ct
ri
c
Mo
t
o
r
cy
c
l
e,
"
Tr
a
n
s
a
c
t
i
o
n
s
-
Ca
n
a
d
i
a
n
S
o
c
i
et
y
f
o
r
M
ech
a
n
i
c
a
l
E
n
g
i
n
ee
r
i
n
g
,
v
o
l
.
3
7
,
n
o
.
3
,
p
p
.
7
4
1
-
7
5
3
,
Sep
t
em
b
er
2
0
1
3
.
[8
]
J
i
n
g
L
i
,
J
i
-
h
an
g
Ch
e
n
g
,
J
i
n
g
-
y
u
a
n
Sh
i
,
an
d
Fei
H
u
an
g
,
"
Bri
ef
In
t
ro
d
u
c
t
i
o
n
o
f
Bac
k
Pro
p
ag
a
t
i
o
n
(BP)
N
e
u
ral
N
et
w
o
r
k
A
l
g
o
r
i
t
h
m
an
d
It
s
Imp
ro
v
emen
t
,
"
A
d
va
n
ces
i
n
Co
m
p
u
t
e
r
S
ci
e
n
ce
a
n
d
In
f
o
r
m
a
t
i
o
n
E
n
g
i
n
ee
r
i
n
g
,
v
o
l
.
1
6
9
,
p
p
.
5
5
3
-
5
5
8
,
J
an
u
ary
2
0
1
2
.
[9
]
Ch
i
h
-
Y
a
o
L
o
,
"
Back
Pro
p
ag
at
i
o
n
N
eu
ra
l
N
e
t
w
o
rk
o
n
t
h
e
Fo
recas
t
i
n
g
Sy
s
t
em
o
f
Sea
Fo
o
d
Mat
er
i
al
D
ema
n
d
,
"
Co
m
m
u
n
i
c
a
t
i
o
n
s
i
n
Co
m
p
u
t
er
a
n
d
In
f
o
r
m
a
t
i
o
n
S
ci
e
n
ce
,
v
o
l
.
2
0
2
,
p
p
.
1
4
7
-
1
5
4
,
J
an
u
ary
2
0
1
1
.
[1
0
]
D
h
e
y
aa
Sh
ah
eed
Sab
r,
“
Re
v
i
e
w
o
n
Ch
al
l
en
g
es
i
n
B
u
i
l
d
i
n
g
K
n
o
w
l
ed
g
e
Bas
e
d
So
c
i
et
i
es
T
h
r
o
u
g
h
E
-
G
o
v
er
n
a
n
ce
Framew
o
r
k
:
T
ech
n
o
l
o
g
y
a
n
d
Ma
n
ag
eme
n
t
I
s
s
u
es
fr
o
m
G
l
o
b
a
l
Pers
p
ect
i
v
e,
”
In
d
i
a
n
J
o
u
r
n
a
l
o
f
S
c
i
en
ce
a
n
d
Tech
n
o
l
o
g
y
,
v
o
l
.
1
0
,
n
o
.
2
8
,
p
p
.
1
-
7
,
A
u
g
u
s
t
2
0
1
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.