Inter
national
J
our
nal
of
Electrical
and
Computer
Engineering
(IJECE)
V
ol.
8,
No.
5,
October
2018,
pp.
3149
–
3157
ISSN:
2088-8708
3149
I
ns
t
it
u
t
e
o
f
A
d
v
a
nce
d
Eng
ine
e
r
i
ng
a
nd
S
cie
nce
w
w
w
.
i
a
e
s
j
o
u
r
n
a
l
.
c
o
m
Motor
cycle
Mo
v
ement
Model
Based
on
Mark
o
v
Chain
Pr
ocess
in
Mixed
T
raffic
Rina
Mardiati
1
,
Bambang
R.
T
rilaksono
2
,
Y
udi
S.
Gondokary
ono
3
,
and
Sony
S.
W
ibo
w
o
4
1,2,3
School
of
Electrical
Engineering
and
Informatics,
Bandung
Institute
of
T
echnology
,
Indonesia
4
Department
of
Ci
vil
Engineering,
Bandung
Institute
of
T
echnology
,
Indonesia
Article
Inf
o
Article
history:
Recei
v
ed
August
9,
2017
Re
vised
June
25,
2018
Accepted
July
7,
2018
K
eyw
ord:
Mark
o
v
Chain
Maneuv
er
Intention
V
ehicle
Mo
v
ement
Mix
ed
T
raf
fic
ABSTRA
CT
Mix
ed
traf
fic
systems
are
dynamically
com
ple
x
since
there
are
man
y
parameters
and
v
ariables
that
influence
the
interactions
between
the
dif
ferent
kinds
of
v
ehicles.
Mod-
eling
the
beha
vior
of
v
ehicles,
especially
motorc
ycle
which
has
erratic
beha
vior
is
still
being
de
v
eloped
continuously
,
espe
cially
in
de
v
eloping
countries
which
ha
v
e
hetero-
geneous
traf
fic.
T
o
get
a
better
understanding
of
motorc
ycle
beha
vior
,
one
can
look
at
maneuv
ers
performed
by
dri
v
ers.
In
this
research,
we
tried
to
b
uild
a
model
of
motor
-
c
ycle
mo
v
ement
which
only
focused
on
maneuv
er
action
to
a
v
oid
the
obstacle
along
with
the
trajectories
using
a
Mark
o
v
Chain
approach.
In
Mark
o
v
Chain,
the
maneuv
er
of
motorc
ycle
will
described
by
state
transition.
The
state
transition
model
is
depend
on
probability
function
which
wil
l
use
for
determine
what
action
will
be
e
x
ecuted
ne
xt.
The
maneuv
er
of
motorc
ycle
using
Mark
o
v
Chain
model
w
as
v
alidated
by
comparing
the
analytical
result
with
the
naturalistic
data,
with
similarity
is
calculated
using
MSE.
In
order
to
kno
w
ho
w
good
our
proposed
method
can
describe
the
maneuv
er
of
mo-
torc
ycle,
we
try
to
compare
the
MSE
of
the
trajectory
based
on
Mark
o
v
Chain
model
with
those
using
polynomial
approach.
The
MSE
results
sho
wed
the
performance
of
Mark
o
v
Chain
Model
gi
v
e
the
smallest
MSE
which
0.7666
about
0.24
better
than
4
th
order
polynomial.
Copyright
c
201x
Institute
of
Advanced
Engineering
and
Science
.
All
rights
r
eserved.
Corresponding
A
uthor:
Rina
Mardiati
Department
of
Electrical
Engineering,
UIN
Sunan
Gunung
Djati
Bandung
Jl.
A.H.
Nasution
No.
105
Bandung
40614,
Indonesia
r
mardiati@uinsgd.ac.id
1.
INTR
ODUCTION
T
oday
,
Indonesian
transportation
problems
are
increasingly
being
encountered
in
our
daily
life.
Se
v-
eral
economic
and
social
moti
v
ations
can
be
rela
ted
to
the
need
to
minimize
the
time
s
pent
in
mot
orc
y
c
le
for
transportation
and
consequently
their
related
pollution
problems.
An
additional
problem
w
orth
mentioning
is
the
need
to
reduce
traf
fic
accidents,
a
human
and
social
cost
that
is
related
not
only
inadequate
dri
ving,
b
ut
also
to
the
planning
of
the
flo
w
conditions
[1].
Due
this
moti
v
ations,
the
literature
on
traf
fic
phenomena
is
already
v
ast
and
characterized
by
contri
b
ut
ions
co
v
ering
modeling
aspects,
statement
of
problems,
qualitati
v
e
analysis,
and
particularly
de
v
eloped
simulation
generated
by
applications.
Continuing
this
ef
forts,
there
are
man
y
litera-
ture
of
traf
fic
flo
w
theories
and
models
ha
v
e
been
de
v
eloped,
b
ut
researcher
generally
agree
that
modeling
has
not
yet
reached
a
satisfying
le
v
el.
Study
of
traf
fic
flo
w
are
become
important
since
man
y
reasons
behind
that,
such
as:
1)
it
is
necessary
to
de
v
elop
traf
fic
model
which
can
describes
the
real
phenomena,
2)
traf
fic
model
can
support
for
de
v
eloping
intelligent
transportation
system
(ITS)
whose
related
with
safety
dri
ving
syste
m,
3)
traf
fic
model
support
for
future
i
ssue
about
intelligent
car
[2].
So,
there
are
man
y
researchers
who
de
v
eloped
intelligent
transportation
system
in
order
to
minimize
the
number
of
traf
fic
accident
[3]
[4]
[5].
Mix
ed
traf
fic
systems
are
dynamically
comple
x
since
there
are
man
y
parameters
and
v
ariables
t
hat
influence
the
interactions
between
the
dif
ferent
kinds
of
motorc
ycles.
These
interactions
can
be
described
as
the
beha
vior
occurring
in
traf
fic.
V
ehicle
beha
vior
is
influenced
by
internal
and
e
xternal
f
actors.
Internal
J
ournal
Homepage:
http://iaescor
e
.com/journals/inde
x.php/IJECE
I
ns
t
it
u
t
e
o
f
A
d
v
a
nce
d
Eng
ine
e
r
i
ng
a
nd
S
cie
nce
w
w
w
.
i
a
e
s
j
o
u
r
n
a
l
.
c
o
m
,
DOI:
10.11591/ijece.v8i5.pp3149-3157
Evaluation Warning : The document was created with Spire.PDF for Python.
3150
ISSN:
2088-8708
f
actors
are
the
status
of
the
motorc
ycle
such
a
s
dri
v
er
psychology
,
motorc
ycle
position,
steering
angle,
speed,
etc.
Meanwhile,
e
xternal
f
actors
are
en
vironmental
conditions
such
as
the
position
of
other
v
ehicles,
road
borders,
pedestrians,
etc.
Modeling
the
beha
vior
of
the
motorc
ycle
is
still
being
continuously
de
v
eloped,
especially
in
de
v
el-
oping
countries
which
ha
v
e
the
characteristics
of
mix
ed
traf
fic.
T
o
get
a
better
understanding
of
motorc
ycle
beha
vior
,
one
can
look
at
maneuv
ers
performed
by
dri
v
ers.
Maneuv
er
here
can
also
be
called
as
a
mo
v
ement
performed
by
the
motorc
ycle
which
is
a
part
of
motorc
ycle’
s
mo
v
ement
to
change
a
lane
and
also
to
a
v
oid
obstacles
or
slo
wer
v
ehicle
in
front
of
him.
Modeling
the
maneuv
er
t
hat
can
adequately
simulate
actual
situations
has
man
y
benefits,
especially
to
solv
e
problems
in
t
he
field
of
transportation.
M
odeling
of
maneuv
ers
can
also
describe
ho
w
a
dri
v
er
interacts
within
a
traf
fic
system.
In
order
to
reach
this
goal,
v
arious
methods
ha
v
e
been
de
v
eloped
to
obtain
a
model
that
can
adequately
simulate
actual
situations.
Based
on
the
literature,
se
v
eral
methods
ha
v
e
been
proposed
for
modeling
v
ehicle
mo
v
ement,
such
as:
Rule
Based
Model
[6],
Cellular
Automata
[7]
[8]
[9],
and
Social
F
orce
Model
[10]
[11]
[12].
Each
method
has
its
o
wn
deficiencies
and
adv
antages.
Rule
Based
Model
is
quite
good
at
describing
v
ehicle
maneuv
ers
in
traf
fic
b
ut
it
only
w
orks
well
at
lo
w
traf
fic
density
.
Problems
will
occur
when
Rule
Based
Model
is
implemented
on
high-density
traf
fic
systems,
making
the
simulation
beha
v
e
in
an
unrealistic
manner
.
Celullar
Automata
(CA)
describes
v
ehicle
maneuv
ers
better
than
the
rule-based
system
b
ut
has
the
disadv
antage
that
its
position
updating
rules
are
deterministic.
Lately
,
Social
F
orce
Model
(SFM)
has
been
applied
to
describe
v
ehicl
e
maneuv
ers
and
it
performed
better
than
both
Rule
Based
Model
and
Celullar
Automata.
SFM
can
describe
the
mo
v
ement
of
v
ehicles
based
on
a
v
ector
-based
approach.
In
SFM,
which
includes
a
mo
v
ement
model
and
the
decision-making
process
of
the
dri
v
er
,
v
ehicle
beha
vior
is
described
by
the
sum
of
se
v
eral
v
ector
forces
(acceleration
force,
repulsi
v
e
force
and
attracti
v
e
force).
Although
SFM
can
describe
v
ehicle
mo
v
ement
better
than
Rule
Based
or
CA,
this
method
does
not
input
some
parameters
that
ha
v
e
a
lar
ge
ef
fect
on
ac
h
i
e
ving
a
realistic
model,
such
as
dri
v
er
psychology
or
a
probabilistic
model
for
dri
v
er
characteristics.
Mark
o
v
Chain
w
as
broadly
use
for
modeling
traf
fic
for
dif
ferent
purposes,
such
as
to
reduce
v
ehicle
emission
[13],
short-term
traf
fic
flo
w
forecasting
[14],
modeling
the
multi-traf
fic
signal-control
synchronization
[15],
traf
fic
state
prediction
[16],
tra
v
el
time
estimation
[17],
etc.
Ho
we
v
er
,
no
one
has
modeled
the
motorc
ycle
maneuv
er
using
the
Mark
o
v
Chain
model.
In
this
research,
we
tried
to
b
uild
a
model
of
motorc
ycle
mo
v
ement
which
only
focused
on
maneuv
er
action
to
a
v
oid
the
obstacle
along
with
the
trajectorie
s
using
a
Mark
o
v
Chain
approach.
In
Mark
o
v
Chain,
the
maneuv
er
of
motorc
ycle
will
described
by
state
transition.
The
state
transition
model
is
depend
on
probability
function
which
will
use
for
determine
what
action
will
be
e
x
ecuted
ne
xt.
The
maneuv
er
of
motorc
ycle
using
Mark
o
v
Chain
model
w
as
v
alidated
by
comparing
the
analytical
result
with
the
naturalistic
data,
with
similarity
is
calculated
using
MSE.
In
order
to
kno
w
ho
w
good
our
proposed
method
can
describe
the
maneuv
er
of
motorc
ycle,
we
try
to
compare
the
MSE
of
the
trajectory
based
on
Mark
o
v
Chain
model
with
those
using
polynomial
approach
that
has
been
done
in
pre
vious
research
[18].
This
paper
is
or
g
anized
as
follo
ws.
In
Section
2,
modeling
motorc
ycle
maneuv
ers
using
a
Mark
o
v
chain
process
is
presented.
The
simulation
and
analysis
of
applying
this
model
are
discussed
in
Section
3.
Finally
,
in
Section
4
this
paper
is
concluded
by
stating
that
the
Mark
o
v
chain
process
can
b
e
implemented
in
modeling
motorc
ycle
mo
v
ement
trajectories
for
certain
maneuv
ers.
2.
THE
PR
OPOSED
METHOD
This
research
tries
to
b
uild
a
model
of
maneuv
er
beha
vior
along
with
the
trajectories
using
a
Mark
o
v
Chain
approach.
In
a
Mark
o
v
Chain
process,
the
dri
v
er
will
mak
e
decisions
to
determine
what
actions
will
be
carried
out
in
a
certain
en
vironment.
V
ehicle
will
be
in
a
certain
position,
mo
ving
at
a
certain
speed
and
steering
angle
at
time
t
,
denoted
as
motorc
ycle
state
(internal
information).
Ne
xt,
motorc
ycle
will
recei
v
e
e
xternal
information
from
its
en
vironment,
s
uch
as
the
positions
and
v
elocities
of
v
ehicles
around
it,
for
further
processing
to
determine
what
action
is
to
be
tak
en
by
motorc
ycle
.
There
are
se
v
eral
steps
must
be
e
x
ecuted
to
b
uild
a
motorc
ycle
maneuv
er
model
using
Mark
o
v
Chain:
1)
Define
set
of
states
and
actions,
2)
Define
the
transition
probability
function,
and
3)
State
mapping
function.
2.1.
Set
of
States
and
Actions
A
state
is
the
condition
or
status
of
motorc
ycle
agent,
which
is
a
function
of
position
(
x;
y
)
,
speed
(
v
)
and
direction
of
steering
maneuv
ers
(
).
Meanwhile,
action
is
described
by
a
function
of
the
maneuv
er
m
2
IJECE
V
ol.
8,
No.
5,
October
2018:
3149
–
3157
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
3151
f
0,1,2
g
with
0,
1,
and
2
corresponds
to
no
maneuv
er
,
right
maneuv
er
,
and
left
maneuv
er
respecti
v
ely
.
Simi
larly
,
acceleration
a
2
f
0,1,2
g
with
0,
1,
and
2
corresponds
to
fix
ed
speed,
acceleration,
and
deceleration.
In
the
Mark
o
v
Chain
process,
the
action
of
the
motorc
ycle
is
determined
by
the
probability
of
a
transition
states.
There
are
ten
states
to
describe
the
mo
v
em
ent
of
motorc
ycle,
along
with
nine
actions.
The
set
of
actions
(
A
i
)
and
states
(
S
i
)
that
are
used
for
the
motorc
ycle
maneuv
ers
can
be
seen
in
T
able
1.
T
able
1.
Set
of
States
and
Actions
States
Definition
Actions
Definition
S
0
V
ehicle
with
speed
0
km/h
A
1
Increases
speed
with
no
maneuv
er
S
1
V
ehicle
with
speed
3
km/h
A
2
Increases
speed
with
right
maneuv
er
S
2
V
ehicle
with
speed
6
km/h
A
3
Increases
speed
with
left
maneuv
er
S
3
V
ehicle
with
speed
9
km/h
A
4
Decreases
speed
with
no
maneuv
er
S
1
L
V
ehicle
at
a
speed
of
3
km/h
with
left
steering
angle
A
5
Decreases
speed
with
right
maneuv
er
S
2
L
V
ehicle
at
a
speed
of
6
km/h
with
left
steering
angle
A
6
Decreases
speed
with
left
maneuv
er
S
3
L
V
ehicle
at
a
speed
of
9
km/h
with
left
steering
angle
A
7
Decreases
speed
with
right
maneuv
er
S
1
R
V
ehicle
at
a
speed
of
3
km/h
with
left
steering
angle
A
8
Fix
ed
v
elocity
with
right
maneuv
er
S
2
R
V
ehicle
at
a
speed
of
6
km/h
with
left
steering
angle
A
9
Fix
ed
v
elocity
with
left
maneuv
er
S
3
R
V
ehicle
at
a
speed
of
9
km/h
with
left
steering
angle
2.2.
T
ransition
Pr
obabilities
A
maneuv
er
is
the
mo
v
ement
to
change
the
direction
of
a
v
ehicle
to
mo
v
e
from
one
lane
to
another
.
Maneuv
ers
in
this
study
are
infl
u
e
nced
by
se
v
eral
parameters,
including
the
internal
state
of
the
v
ehicle
as
well
as
the
e
xternal
conditions,
as
sho
wn
in
Figure
1.
d
αβ
v
β
d
β
1
D
B
r
D
B
l
v
β
1
w
β
D
O
V
D
O
H
V
ehicle
β
V
ehicle
β
1
V
ehicle
α
W
R
Figure
1.
Illustration
of
Maneuv
er’
s
P
arameters
The
parameter
in
Figure
1
are
d
;
v
;
w
;
d
1
;
d
2
;
:
:
:
;
d
n
;
v
1
;
v
2
;
d
B
r
;
d
B
l
;
D
O
H
;
D
O
V
;
and
W
R
.
Where
d
is
a
distance
between
motorc
ycle
and
v
ehicle
,
v
is
speed
of
v
ehicle
,
w
is
the
width
of
v
ehicle
;
(
d
1
;
d
2
;
:
:
:
;
d
n
),
distance
between
motorc
ycle
and
v
ehicles
to
the
left
and
right,
v
1
is
a
speed
of
v
ehicle
to
the
left,
v
2
is
speed
of
v
ehicle
to
the
right,
d
B
r
is
the
distance
between
v
ehicle
and
the
right
border
,
d
B
l
,
distance
between
v
ehicl
e
and
the
left
border
,
D
O
H
is
a
horizontal
of
fset
distance
to
objects
that
are
in
front
of
the
v
ehicle,
D
O
V
is
the
v
ertical
of
fset
distance
to
v
ehicles
to
the
left
or
the
right
of
the
v
ehicle,
and
W
R
is
the
width
of
road.
There
are
tw
o
steps
performed
to
b
uild
the
transition
probability
function.
Firstly
,
to
model
the
prob-
ability
of
maneuv
erability
function
of
the
rider
.
Secondly
,
to
model
the
probability
of
maneuv
ering
direction
(left
or
right)
to
be
tak
en
by
the
motorc
ycle.
2.2.1.
T
ransition
Pr
obability
of
V
ehicle
Maneuv
ers
The
biggest
f
actor
that
influenced
the
maneuv
er
is
if
there
is
an
obstacle
in
front
of
the
motorc
ycle.
As
function
to
model
the
desire
of
maneuv
ering
of
the
dri
v
er
,
ne
g
ati
v
e
e
xponential
function
can
be
used
to
obtain
a
great
v
alue
for
small
distances
and
a
small
v
alue
for
lar
ge
distances
.
In
addition,
a
critical
distance
f
actor
(
D
c
)
should
be
added,
which
represents
the
critical
distance
where
the
motorc
ycle
must
do
maneuv
er
.
If
a
maneuv
er
V
ehicle
Mo
vement
Model
Based
on
Mark
o
v
Chain
Pr
ocess
in
Mixed
T
r
af
fic
(Rina
Mar
diati)
Evaluation Warning : The document was created with Spire.PDF for Python.
3152
ISSN:
2088-8708
cannot
be
carried
out
for
e
xample
because
there
are
v
ehicle
s
to
the
right
and
left,
then
the
action
is
decelerate
the
speed.
The
maneuv
er
intention
probability
of
motorc
ycle
stated
as
P
can
be
e
xpressed
as
follo
ws.
P
=
(
e
k
1
(
d
D
c
)
;
for
d
>
D
c
1
;
for
d
D
c
(1)
Where
d
is
the
distance
between
motorc
ycle
and
the
v
ehicle
in
front
of
it
(v
ehicle
),
D
c
is
represents
the
critical
distance
where
the
motorc
ycle
must
do
maneuv
er
or
decelerate
the
speed,
k
1
is
a
constant
that
sho
ws
an
increased
slope
rate
of
a
ne
g
ati
v
e
e
xponential
curv
e
as
a
function
of
the
distance
between
the
tw
o
v
ehicles.
Ph
ysically
,
a
small
v
alue
of
k
1
indicates
that
the
dri
v
er
w
ants
to
maneuv
er
despite
the
distance
to
the
v
ehicle
in
front
still
being
quite
lar
ge
(alert
dri
v
er),
while
a
high
v
alue
of
k
1
indicates
that
the
dri
v
er
w
ants
to
maneuv
er
when
the
distance
to
the
v
ehicle
in
front
is
already
v
ery
small
(a
dri
v
er
who
is
reluctant
to
maneuv
er
unless
pressed).
2.2.2.
Pr
obability
Estimation
of
Maneuv
er
Dir
ection
When
the
decision
to
maneuv
er
based
on
P
e
xceeds
a
certain
threshold
v
alue,
the
ne
xt
step
is
mod-
eled
the
optimum
maneuv
er
direction.
The
possible
directions
for
the
maneuv
er
are
determined
by
the
condition
of
the
road,
the
borders
to
the
right
and
left,
and
other
v
ehicles
in
the
vicinity
.
In
this
subsection,
we
b
uild
ma-
neuv
er
direction
probability
model
for
six
scenarios
that
commonly
occur
in
traf
fic,
as
sho
wn
in
Figure
2.
This
scenario
is
stil
l
possible
to
be
de
v
eloped
further
in
a
follo
w-up
study
,
which
will
be
tailored
to
observ
ations
in
the
field
and
processing
the
data
obtained.
The
deri
v
at
ion
process
of
the
probability
model
has
been
done
in
a
pre
vious
research
[18].
T
able
2.
Maneuv
er
Direction
Probabilities
Scenario
Pr
obability
of
Maneuv
er
Dir
ection
There
are
no
obstructions
on
the
left
nor
right
P
L
=
e
k
3
(
d
B
l
d
C
l
)
W
R
k
2
l
P
R
=
e
k
3
(
d
B
r
d
C
r
)
W
R
k
2
r
There
is
an
obstruction
to
the
right
of
the
motorc
yle
P
L
=
e
k
3
(
d
B
l
d
C
l
)
W
R
k
2
l
P
R
=
(
1
e
k
4
(
d
1
d
1
C
)
,
for
d
1
>
d
1
C
0
,
for
d
1
d
1
C
There
is
an
obstruction
to
the
left
of
the
motorc
ycle
P
L
=
(
1
e
k
5
(
d
1
d
1
C
)
,
for
d
1
>
d
1
C
0
,
for
d
1
d
1
C
P
R
=
e
k
3
(
d
B
r
d
C
r
)
W
R
k
2
r
There
are
obstruction
to
the
right
and
to
the
left
of
the
motorc
ycle
P
L
=
(
1
e
k
5
(
d
1
d
1
C
)
,
for
d
1
>
d
1
C
0
,
for
d
1
d
1
C
P
R
=
(
1
e
k
4
(
d
1
d
1
C
)
,
for
d
1
>
d
1
C
0
,
for
d
1
d
1
C
There
is
a
border
on
the
left
or
on
the
right
P
L
=
0
P
R
=
1
There
are
boarders
and
obstructions
P
L
=
e
k
3
(
d
B
l
d
C
l
)
W
R
k
2
l
P
R
=
(
1
e
k
4
(
d
1
d
1
C
)
,
for
d
1
>
d
1
C
0
,
for
d
1
d
1
C
T
able
2
sho
ws
the
maneuv
er
direction
probability
function.
P
L
denotes
the
probability
of
a
left
direction
maneuv
er
and
P
R
denotes
the
probability
of
a
right
direction
maneuv
er
,
with
k
2
;
k
3
;
k
4
and
k
5
as
constants.
P
arameter
denotes
as
a
function
of
horizontal
of
fset
distances
to
objects
that
are
in
front
of
motorc
ycle
(
D
O
H
l
;
D
O
H
r
)
,
where
w
=
d
O
H
l
+
d
O
H
r
,
l
=
D
O
H
l
D
O
H
l
+
d
O
H
r
and
r
=
D
O
H
r
D
O
H
l
+
D
O
H
r
.
d
C
r
and
d
C
l
denotes
critical
distance
with
motorc
ycle
on
the
left
and
right.
The
v
alue
can
be
set
to
the
v
alue
that
describes
the
habits
of
maneuv
ering
direction
of
Indonesian
dri
v
er
.
IJECE
V
ol.
8,
No.
5,
October
2018:
3149
–
3157
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
3153
2.3.
State-Mapping
Function
After
the
transiti
on
probability
function
were
de
termined,
then
we
need
to
b
uild
a
mapping
func-
tion
which
describes
the
state-action
mapping
of
decision
making
beha
vior
of
dri
v
er
.
Mathematically
,
this
is
e
xpressed
by
P(
S
t
+1
j
S
t
;
S
1
;
S
;
S
3
;
;
m;
a
)
=
P(
S
i
+1
j
S
t
;
m;
a
)
(2)
with
P(
S
i
+1
j
S
i
;
m
i
;
a
i
)
denotes
probability
function
of
maneuv
er
and
acceleration
at
time
t
.
(a)
(b)
(c)
(d)
(e)
(f
)
Figure
2.
T
raf
fic
Scenario
3.
RESEARCH
METHOD
The
research
method
consists
of
three
main
steps:
1)
Data
preprocessing,
2)
Simulation,
and
3)
Com-
paring
the
simulation
results
with
actual
data.
Data
preprocessing
w
as
be
gin
with
doing
the
naturalistic
obser
-
v
ation
using
video
streaming
which
ensures
that
the
normal
beha
vior
can
be
observ
ed
and
the
data
collected
are
not
af
fected
by
the
presence
of
researchers.
Ne
xt,
in
order
to
obtained
pix
el
coordinate
which
represents
motorc
ycle’
s
trajectory
,
coordinates
transformation
to
con
v
ert
the
pix
el
coordinate
to
real-w
orld
coordinate
as
we
seen
on
Figure
3.
The
preprocessing
of
data
in
this
research
is
similar
to
those
in
our
pre
vious
research
[18].
(a)
(b)
(c)
Figure
3.
The
flo
w
of
data
preprocessing,
(a)
T
ar
geted
motorc
ycle;
(b)
Maneuv
ers
trajectory;
(c)
Photo
coordi-
nate
in
a
pix
el
4.
SIMULA
TION
RESUL
TS
AND
DISCUSSION
The
simulation
results
sho
w
that
motorc
ycle
mo
v
ement
modeling
based
on
a
Mark
o
v
Chain
process
w
as
successf
ully
implemented.
The
Mark
o
v
Chai
n
model
w
as
e
v
aluated
by
looking
at
a
plotti
ng
diagram.
There
are
three
diagrams
in
Figure
4
which
describes
that
Mark
o
v
Chain
w
as
successfully
implemented
to
modeled
the
manue
v
er
of
motorc
ycle
(a
maneuv
er
diagram,
a
state
transition
diagram,
and
an
acceleration
diagram).
In
Figure
4
(a),
the
changes
of
each
state
and
speed
for
e
v
ery
v
ehicle
during
the
simulation
period
can
be
seen
clearly
.
The
blue
motorc
ycle
change
his
state
in
order
to
get
a
sa
v
ely
dri
ving.
Besides,
Figure
4
(c),
we
can
see
the
beha
vior
of
the
blue
motorc
ycle
making
three
times
maneuv
er
during
simulation
time,
at
the
5
th
,
the
19
th
,
and
the
40
th
time
step
of
the
duration.
In
that
time,
the
blue
motorc
ycle
sees
a
v
ehicle
in
front
of
V
ehicle
Mo
vement
Model
Based
on
Mark
o
v
Chain
Pr
ocess
in
Mixed
T
r
af
fic
(Rina
Mar
diati)
Evaluation Warning : The document was created with Spire.PDF for Python.
3154
ISSN:
2088-8708
him
and
try
to
a
v
oid
the
v
ehicle
by
making
a
maneuv
er
,
while
considering
the
border
beside
him
to
calculate
the
maneuv
er
probability
.
The
performance
of
Mark
o
v
Chain
model
is
v
erified
by
comparing
actual
maneuv
er
trajectory
with
the
trajectory
from
Mark
o
v
Chain
model.
Figure
5
(a)
sho
ws
the
actual
trajectory
data
of
ten
motorc
ycle’
s
that
has
tak
en
independently
at
dif
ferent
time
b
ut
plotted
in
one
figure.
It
is
interesting
to
note
that
there
are
dif
ferent
types
of
maneuv
er
style
in
that
trajectory
,
for
e
xample
motorc
ycle
1
mak
es
a
comple
x
maneuv
er
pattern
as
to
a
v
oid
the
obstacle.
In
the
other
hand,
motorc
ycle
3
and
motorc
ycle
6
tak
e
a
simple
maneuv
er
with
considerable
distance
to
complete
this
maneuv
er
.
Figure
5
(b)
sho
ws
the
track
of
Mark
o
v
Chain
model
using
(a)
(b)
(c)
Figure
4.
Mark
o
v
Chain
simulation,
(a)
State-transition
diagram;
(b)
Acceleration
diagram;
(c)
Maneuv
er
decision
diagram
that
only
accommodate
one
maneuv
ering
angle
which
is
45
.
W
e
observ
ed
in
this
figure
that
Mark
o
v
Chain
model
successfully
follo
w
the
trend
of
actual
maneuv
er
b
ut
cannot
produce
a
smooth
pattern.
T
o
impro
v
e
the
situation,
in
Figure
5
(c),
we
add
additional
states
of
maneuv
er
s
teering
angle
which
are
22
:
5
;
45
;
62
:
5
to
accommodate
a
smoother
trajectory
pattern
which
is
v
ery
close
to
the
actual
one.
As
a
comparison
we
plot
the
curv
e
fitting
of
the
track
using
static
polynomial
fitting
which
are
2
nd
order
polynomial,
3
r
d
order
polynomial,
and
4
th
order
polynomial
as
sho
wn
in
Figure
5
(d)
(e)
(f).
In
those
figures,
we
observ
ed
that
polynomial
can
fit
the
actual
trajectory
as
f
ar
as
the
trajectory
is
short
and
simple
curv
e
such
as
the
maneuv
er
of
motorc
ycle
2,
4,
and
9.
Ho
we
v
er
,
polynomial
fitting
has
dif
ficulty
to
track
a
comple
x
maneuv
er
such
as
the
maneuv
er
of
motorc
ycle
1.
This
polynomial
fitting
also
cannot
track
the
slight
maneuv
er
such
as
maneuv
er
of
motorc
ycle
7
and
10.
IJECE
V
ol.
8,
No.
5,
October
2018:
3149
–
3157
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
3155
(a)
(b)
(c)
(d)
(e)
(f)
Figure
5.
Manually
track
of
ten
dif
ferent
motorc
ycle’
s
maneuv
er
,
(a)
actual
motorc
ycle’
s
maneuv
er
track
(b)
approximation
of
maneuv
er’
s
track
using
Mark
o
v
Chain
with
=
45
(c)
approximation
of
maneuv
er’
s
track
using
Mark
o
v
Chain
with
=
22
:
5
;
45
;
62
:
5
(d)
approximation
of
maneuv
er’
s
track
using
2
nd
order
polynomial
(e)
approximation
of
maneuv
er’
s
track
using
3
r
d
order
polynomial
(f)
approximation
of
maneuv
er’
s
track
using
4
th
order
polynomial
Performance
of
each
fitting
method
is
sho
wn
in
T
able
3.
In
T
able
3,
we
observ
ed
that
Mark
o
v
Chain
model
gi
v
e
the
s
mallest
MSE
which
0.7666
about
0.24
better
than
4
th
order
polynomial.
Ev
en
though
in
most
cases
the
Mark
o
v
Chain
model
using
22.5,
45,
62.5
has
better
MSE
b
ut
in
particular
case,
for
e
xample
in
motorc
ycle
6,
the
performance
of
this
model
is
w
orse
than
polynomial
fitting,
since
the
maneuv
er
is
v
ery
comple
x.
T
able
3.
Mean
Square
Error
of
Motorc
ycle
Maneuv
er
T
rajectory
T
racking
V
ehicle
2
nd
order
3
r
d
order
4
th
order
=
45
=
22
:
5
;
45
;
62
:
5
1
2.7076
2.6887
2.1865
1.8073
0.6907
2
0.0849
0.0640
0.0626
0.4998
0.1472
3
1.0267
0.8822
0.8655
1.6302
0.4086
4
0.2525
0.2521
0.2509
0.9108
0.4336
5
0.8472
0.8092
0.6743
1.1847
0.2783
6
1.7213
1.3810
1.2796
3.8066
2.3199
7
1.9405
1.9405
1.9344
2.0455
0.6516
8
1.4145
1.4056
1.3684
2.0102
0.5179
9
0.2622
0.2022
0.1966
2.0889
0.7340
10
2.1705
1.6028
1.2978
3.8687
1.4841
A
v
erage
1.2428
1.1228
1.0117
1.9853
0.7666
5.
CONCLUSION
In
this
research,
a
simple
model
based
on
a
Mark
o
v
Chain
process
w
as
implemented
to
descr
ibe
motorc
ycle
maneuv
er
using
ten
states
and
nine
actions.
Besides,
the
probability
function
of
maneuv
er
intention
V
ehicle
Mo
vement
Model
Based
on
Mark
o
v
Chain
Pr
ocess
in
Mixed
T
r
af
fic
(Rina
Mar
diati)
Evaluation Warning : The document was created with Spire.PDF for Python.
3156
ISSN:
2088-8708
w
as
introduced
in
six
scenarios
which
inte
grated
to
the
state-mapping
function
of
Mark
o
v
Chain
Model.
The
maneuv
er
of
motorc
ycle
using
Mark
o
v
Chain
model
w
as
v
erified
by
comparing
the
analytical
result
with
the
naturalistic
data
which
gi
v
e
small
MSE.
This
methods
also
compare
with
static
polynomial
fitting
to
kno
w
ho
w
good
is
Mark
o
v
Chain
can
describe
a
motorc
ycle
maneuv
er
mo
v
ement.
As
a
result,
we
found
that
MSE
of
Mark
o
v
Chain
model
is
smaller
than
static
polynomial
fitting
approach.
A
CKNO
WLEDGMENT
The
authors
w
ould
lik
e
to
thank
the
Indonesia
Endo
wment
Fund
for
Education
(LPDP
Indonesia)
and
the
Indonesian
Ministry
of
Religious
Af
f
airs
for
their
financial
support
of
this
study
.
REFERENCES
[1]
N.
Bellomo
and
C.
Dogbe,
“On
the
modeling
of
traf
fic
and
cro
wds:
A
surv
e
y
of
models,
speculations,
and
perspecti
v
es,
”
SIAM
Re
vie
w
,
v
ol.
53,
no.
3,
pp.
409–463,
2011.
[Online].
A
v
ailable:
https://doi.or
g/10.1137/090746677
[2]
R.
Mardiati,
N.
Ismail,
and
A.
F
aroqi,
“Re
vie
w
of
microscopic
model
for
t
raf
fic
flo
w
,
”
ARPN
J
ournal
of
Engineering
and
Applied
Sciences
,
v
ol.
9,
no.
10,
pp.
1794–1800,
October
2014.
[3]
K.
Gopalakrishna
and
S.
Hariprasad,
“Real-time
f
atigue
analysis
of
dri
v
er
through
iris
recognition,
”
In-
ternational
J
ournal
of
Electrical
and
Computer
Engineering
(IJECE)
,
v
ol.
7,
no.
6
,
pp.
3306–3312,
2017.
[4]
S.
K.
V
enkata,
“
An
intelligent
online
v
ehicle
tyre
pressure
monitoring
system,
”
International
J
ournal
of
Electrical
and
Computer
Engineering
,
v
ol.
2,
no.
3,
p.
301,
2012.
[5]
T
.
S.
Guna
w
an,
A.
Mutholib,
and
M.
Kartiwi,
“Performance
e
v
aluation
of
automatic
number
plate
recog-
nition
on
android
smartphone
platform,
”
International
J
ournal
of
Electrical
and
Computer
Engineering
,
v
ol.
7,
no.
4,
p.
1973,
2017.
[6]
L.
Chong,
M.
M.
Abbas,
A.
M.
Flintsch,
and
B.
Higgs,
“
A
rule-based
neural
net-
w
ork
approach
to
model
dri
v
er
naturalistic
beha
vior
in
traf
fic,
”
T
r
ansportation
Resear
c
h
P
art
C:
Emer
ging
T
ec
hnolo
gies
,
v
ol.
32,
pp.
207
–
223,
2013.
[Online].
A
v
ailable:
http://www
.sciencedirect.com/science/article/pii/S0968090X12001210
[7]
Y
.
Zhang
and
H.
Duan,
“Modeling
mix
ed
traf
fic
flo
w
at
crossw
alks
in
microsimulations
using
cellular
automata,
”
Tsinghua
Science
and
T
ec
hnolo
gy
,
v
ol.
12,
no.
2,
pp.
214–222,
April
2007.
[8]
J.
X.
Ding,
H.
J.
Huang,
and
Q.
T
ian,
“
A
mix
ed
traf
fic
flo
w
m
od
e
l
based
on
a
modified
cellular
automaton
in
tw
o-lane
system,
”
in
2009
International
J
oint
Confer
ence
on
Computational
Sciences
and
Optimi
za-
tion
,
v
ol.
2,
April
2009,
pp.
124–126.
[9]
Z.
dong
Zhang,
Y
.
f
ang
yang,
W
.
Qi,
A.
Chariete,
and
X.
xiang
Lin,
“
A
cellular
automata
traf
fic
flo
w
model
considering
b
us
lane
changing
beha
vior
wit
h
scheduling
parameters,
”
Discr
ete
Dynamics
in
Natur
e
and
Society
,
v
ol.
2015,
p.
7
pages,
januari
2015.
[10]
T
.
Ming
and
J.
Hongfei,
“
An
approach
for
calibration
and
v
alidation
of
the
social
force
pedestrian
model,
”
2011,
kalibrasi
dan
V
alidasi
P
arameter
pada
SFM.
[11]
D.
N.
HUYNH,
M.
BOL
TZE,
and
A.
T
.
VU,
“Modelling
mix
ed
traf
fic
flo
w
at
signalized
intersectionusing
social
force
model,
”
J
ournal
of
the
Eastern
Asia
Society
for
T
r
ansportation
Studies
,
v
ol.
10,
pp.
1734–
1749,
2013.
[12]
W
.
Zeng,
H.
Nakamura,
and
P
.
Chen,
“
A
modified
social
force
model
for
pedestrian
beha
vior
simulation
at
signalized
crossw
alks,
”
Pr
ocedia
-
Social
and
Behavior
al
Sciences
,
v
ol.
138,
pp.
521
–
530,
2014.
[Online].
A
v
ailable:
http://www
.sciencedirect.com/science/article/pii/S1877042814041536
[13]
J.
Sun,
C.
Li,
J.
Ding,
J.
Y
ang,
and
Z.
Liu,
“
A
mark
o
v
chain
based
traf
fic
flo
w
control
model
for
re-
ducing
v
ehicles’
co
2
emissions,
”
in
V
ehicular
Electr
onics
and
Safety
(ICVES),
2015
IEEE
International
Confer
ence
on
.
IEEE,
2015,
pp.
250–255.
[14]
S.
Sun,
G.
Y
u,
and
C.
Zhang,
“Short-term
traf
fic
flo
w
forecasting
using
sampling
mark
o
v
chain
method
with
incomplete
data,
”
in
Intellig
ent
V
ehicles
Symposium,
2004
IEEE
.
IEEE,
2004,
pp.
437–441.
[15]
J.
B.
Clempner
and
A.
S.
Pozn
yak,
“Modeling
the
multi-traf
fic
signal-control
synchronizat
ion:
A
mark
o
v
chains
g
ame
theory
approach,
”
Engineering
Applications
of
Artificial
Intellig
ence
,
v
ol.
43,
pp.
147–156,
2015.
[16]
C.
Antoniou,
H.
N.
K
outsopoulos,
and
G.
Y
annis,
“T
raf
fic
state
prediction
using
mark
o
v
chain
models,
”
in
Contr
ol
Confer
ence
(ECC),
2007
Eur
opean
.
IEEE,
2007,
pp.
2428–2435.
[17]
J.
Y
eon,
L.
Elefteriadou,
and
S.
La
wphongpanich,
“T
ra
v
el
time
estimation
on
a
free
w
ay
using
discrete
IJECE
V
ol.
8,
No.
5,
October
2018:
3149
–
3157
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISSN:
2088-8708
3157
time
mark
o
v
chains,
”
T
r
ansportation
Resear
c
h
P
art
B:
Methodolo
gical
,
v
ol.
42,
no.
4,
pp.
325–338,
2008.
[18]
R.
Mardiati,
B.
R.
T
rilaksono,
Y
.
S.
Gondokaryono,
and
S.
S.
Sulaksono,
“Mot
orc
ycl
es
trajectory
tracking
model
based
on
polynomial
least-squares
approximation,
”
Advanced
Science
Letter
s
,
v
ol.
23,
no.
5,
pp.
4537–4541,
2017.
BIOGRAPHY
OF
A
UTHORS
Rina
Mardiati
recei
v
ed
her
Bachelor
de
gree
in
Mathematics
Education
from
Uni
v
ersitas
Pen-
didikan
Indonesia
(UPI)
in
2006
and
Master
de
gree
in
Electrical
Engineering
from
Institut
T
eknologi
Bandung
(ITB)
in
2009.
She
is
no
w
a
PhD
student
at
the
School
of
Electrical
Engi-
neering
and
Informatics,
ITB.
Her
research
is
in
the
fields
of
mathematical
modeling,
simulation,
traf
fic
modeling
and
intelligent
systems.
She
is
af
filiated
with
IEEE
as
a
student
member
.
Bambang
Riyanto
T
rilaksono
recei
v
ed
his
Bachelor
de
gree
in
Electrical
Engineering
from
ITB,
and
Master
and
Doctorate
de
gre
es
from
W
aseda
Uni
v
ersity
,
Japan.
He
is
no
w
a
professor
at
the
School
of
Electrical
Engineering
and
Informatics,
ITB.
His
research
interests
include
rob
ust
control,
intelligent
contr
ol
&
intelligent
systems
,
discrete
e
v
ent
systems,
control
applications,
telerobotics,
embedded
control
systems,
and
robotics.
Y
udi
Satria
Gondokary
ono
Gondokaryono
recei
v
ed
his
Bachelor
de
gree
in
Electrical
Engineering
from
ITB,
and
Master
and
Doctorate
de
gree
from
Ne
w
Me
xico
State
Uni
v
ersity
.
He
is
no
w
an
Assistant
Professor
in
School
of
Electrical
Engineering
and
Informatics
at
School
of
Electrical
Engineering
and
Informatics,
ITB.
His
resea
rch
interests
include
human
computer
interaction,
real-
time,
and
embedded
systems.
Sony
Sulaksono
W
ibo
w
o
recei
v
ed
his
Bachelor
and
Master
de
grees
in
Ci
vil
Engineering
from
ITB,
and
Doctorate
de
gree
from
Chualalangk
orn
Uni
v
ersity
.
He
is
no
w
an
assistant
professor
at
the
Department
of
Ci
vil
Engineering,
ITB.
His
research
interests
include
tra
v
el
beha
vior
,
public
trans-
portation
and
transport
planning,
non-motorized
transportation,
transportation
and
the
en
vironment.
V
ehicle
Mo
vement
Model
Based
on
Mark
o
v
Chain
Pr
ocess
in
Mixed
T
r
af
fic
(Rina
Mar
diati)
Evaluation Warning : The document was created with Spire.PDF for Python.