TELKOM
NIKA
, Vol.14, No
.2, June 20
16
, pp. 489~4
9
6
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.3156
489
Re
cei
v
ed
No
vem
ber 2
6
, 2015; Re
vi
sed
March 22, 20
16; Accepted
April 6, 2016
Segway Line Tracer Using Proportional-Integral-
Derivative Controllers
Wija
y
a
Kurnia
w
a
n*
1
, Mochammad Hannats Hanafi
Ichsan
2
, Eko Setia
w
an
3
Univers
i
t
y
Of Bra
w
ij
a
y
a, Jl. Ve
teran Mal
a
n
g
, East Java, Pho
ne: +
62 34
1 55
161
1/ F
a
x +
62
341 5
6
5
420
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
w
j
a
y
kur
n
ia
@
ub.ac.id
1
, ha
na
s.hanafi
@
ub.
a
c
.id
2
, ekosetia
w
a
n@
ub.ac.i
d
3
A
b
st
r
a
ct
Intelli
gent c
ont
rol, se
nsors
a
nd
hardw
ar
e i
n
tegr
ati
o
n
are
exp
e
cted
to
gen
erate
an
efficient
transportation system
and and
mi
nimum
effort, to carry goods from
one
location to anot
her loc
a
tion. Line
tracer use
d
by
robot to trans
p
o
rt follow
the p
a
th; it has
a sy
stem that us
es
a lig
ht sens
or to read th
e col
o
r
from a li
ne that
represe
n
ts
the path to mak
e
specific
dir
e
cti
on. Segw
ay is tw
o
w
heele
d
transp
o
rtatio
n ite
m
that have a
n
ef
ficient en
er
gy
used. Now
a
da
ys line trac
er can on
ly w
o
rk if it has three or
mor
e
w
heels
an
d
segw
ay can o
n
ly w
o
rk w
i
th
riders. T
h
is re
search
se
gw
a
y
desig
ne
d by
lego ro
bot, P
I
D (Proportio
n
al,
Integral, D
e
riv
a
tive) co
ntrol u
s
ed to co
ntrol
an i
nput fr
o
m
gyrosco
pe se
n
s
or in for
m
of
elev
ation
an
gl
e of
the earth. The
control syste
m
is expec
te
d to control tw
o w
heel
ed Se
gw
ay to
reach stea
d
y
state rapid
l
y. So
the Segw
ay w
ould ru
n w
i
thout invo
lv
in
g hu
ma
n or w
i
thout rid
e
r.
Ke
y
w
ords
: Se
gw
ay, line trac
er, gyroscop
e
and PID
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Two
-
wheel
ed
vehicle that i
s
commo
nly calle
d
a Seg
w
ay, being
a
resea
r
ch topi
c that is
gro
w
ing, and
Line Tra
c
er
is a robot th
at can wal
k
on his o
w
n b
y
reading the
lines and pa
ths
have bee
n d
e
termin
ed an
d balan
ce
au
tomatically b
a
se
d on
cha
nge
s and
shi
fts in its bala
n
c
e
point [1]. This study aim
s
to expand re
sea
r
ch t
hat has be
en do
n
e
previou
s
ly, whi
c
h combi
n
e
s
the two whe
e
l balan
cing
robot an
d make a two
-
wheele
d
rob
o
t can
walk o
n
his o
w
n by usin
g
Line T
r
a
c
e
r
[2]. Segway i
s
expecte
d to
prod
uce a
de
sign th
at can
run
itself
(au
t
omatic) carrie
s
the good
s wit
hout huma
n
a
ssi
stan
ce [3].
In the previo
us
studie
s
co
ndu
cted, two
-
whe
e
led
robo
t lego su
cce
s
sfully created
using
the Model Predictive Control by
using in
verted pen
dul
um [5]. Another study Q
-
Le
arnin
g
to make
a rob
o
t ca
n l
earn
ho
w to
solve the
pro
b
lems fin
d
ing
the route
by way of
the robot itself aft
e
r
some time le
arnin
g
in via the sam
e
path
[6].
This
research will be conduct
ed by m
a
king the design of t
he Segway that use
the robot
as re
se
arch o
b
ject
s. Lego
NXT is a ro
b
o
t which
is u
s
ed as a mo
de
l Segway, an
d will be give
n a
colo
r se
nsor t
hat use
s
a PID co
ntrolle
r for the testing
pro
c
e
ss.
Lego
NXT ro
bot will be
co
me a mod
e
l a
nd a
s
a me
an
s of test the
control al
gorith
m
s that
will b
e
exami
ned. PID con
t
rol al
gorithm
progra
mme
d
with
high
-lev
el lan
gua
ge
C a
nd
ha
s ni
ce
performance
to tracking st
ability [4
]. Th
e color sensor is a device t
hat can di
stinguish colors, in
this ca
se
colo
r sen
s
o
r
u
s
ed
to make the robot
re
cog
n
ize the colo
r of the path that has
colo
r [5].
The
suitabilit
y of the perfor
mance of
the control
algo
rithm, color sensors
and
perfo
rman
ce
the robot will
be investigat
ed in ord
e
r
Segway robot that use
s
the con
c
e
p
t that run
s
on a track th
at has bee
n provide
d
and
can ru
n it
sel
f
automaticall
y
without the driver. Ho
w to
orga
nize the
perfo
rman
ce
of two m
o
tors, each of
whi
c
h i
s
own
ed
by both
whe
e
l
s, so the
mot
o
r
can alte
rnatel
y proce
s
sed
balan
ce a
nd
make the
shif
t place
s
.
2. Rese
arch
Metho
d
Lego Mi
nd
storm
s
NXT
ki
t is an p
a
ckage that
con
t
ains p
a
rts such
as
Leg
o
robot
s,
sen
s
o
r
s, a
c
tu
ators and
a small co
mpute
r
-
NXT Bri
ck.
In ord
e
r to a
b
le to devel
o
p
and
uplo
ad
the
prog
ram to
th
e NXT Bri
c
k required a
n
al
gorithm
. T
h
is re
sea
r
ch u
s
ed Le
go b
e
cause it used
32-
bit ARM micropro
c
e
s
so
r a
nd the user
can p
r
og
ram
m
ed it as we
ll using the
compile
r Leg
o
C
.
Lego
C contai
ns a compl
e
te developm
e
n
t environm
e
n
t [7]. The reaso
n
for cho
o
sin
g
Leg
oC
is in
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 489 – 49
6
490
addition to
ha
ving a hig
h
p
r
oce
s
sing
sp
e
ed is also very easy to
work with
an
d ha
s ma
ny built-i
n
feature
s
, su
ch a
s
dat
a loggin
g
, gamep
ad
supp
ort,
debu
gging,
Bluetooth wirele
ss
comm
uni
cati
on, sen
s
o
r
dri
v
ers an
d so o
n
[8].
Figure 1. Leg
o Mindsto
rm
NXT
Figure 2. Col
o
rs
circuit se
nso
r
Lego h
a
s a
color
sen
s
o
r
to detect multi
p
le col
o
rs in
clude red, gre
en, and blu
e
[9]. Step
in Figure 2 re
d diode is tu
rned on for a
few mome
nts, the red light of an object and the obje
c
t
will reflect
an certain li
ght i
n
tens
ity
characteri
stics. T
h
e intensit
y i
s
detected
by
LDR
whi
c
h i
s
on
the active su
rface of the se
nso
r
.
This p
r
o
c
e
s
s is rep
eated
on the othe
r two diode
s
are g
r
ee
n an
d blue so th
at it will
obtain 3 valu
es are then p
r
ocesse
d in the pro
g
ra
m.
With a particular refe
re
nce value that
has
been determi
ned
in previo
us calib
ration
pro
c
e
s
s,
the
colo
r of a
n
obje
c
t ca
n b
e
dete
r
mine
d
or
read by the
sen
s
o
r
. In the diagram it can
be
see
n
in the image col
o
r det
ection p
r
o
c
e
ss
pro
c
ed
ure bel
ow.
Figure 3. Col
o
r Dete
ction
Procedu
re
Although
co
mbine
thre
e
colors
on th
e
LED
will
pro
d
u
ce
s
white
c
ol
or, it d
o
e
s
n
o
t
mean
3
LEDs
ca
n be
repla
c
e
d
wit
h
an white L
E
D. The re
as
on is that in
orde
r to ident
ify the color o
f
an
obje
c
t, it take
s a
minimu
m
of 2
comp
on
ent si
gnal.
T
h
is
wa
s do
ne
by mea
s
u
r
ing
the inte
nsity
of
each freque
n
c
y of the
refle
c
ted li
ght. In t
h
is
ca
se
it
is
the color
sen
s
or
circuit i
s
only u
s
ing
three
basi
c
colors o
f
red, gree
ncolor, and bl
ue
[10].
The
p
r
o
c
e
s
s of
acquiri
ng colo
r sen
s
o
r
to
rea
d
the
si
gnal
s, the
sig
nals
coll
ecte
d at the
digitize
d se
nsor anal
og inp
u
t to a value in the rang
e 0
- 1023.
Ran
ge si
gnal
area d
e
fined
by the followi
ng equ
ation:
(
1
)
In orde
r to obtain a hig
h
-resolution
sen
s
o
r
,
the resi
stan
ce of
the shunt resi
stor i
s
cal
c
ulate
d
to
suit the
cha
r
a
c
teri
stics
of the L
D
R.
Th
e
ci
rcuit con
s
ists of
L
D
R a
n
d
shunt
re
si
stor
form a voltag
e divider ci
rcuit.
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Segwa
y
Lin
e
Tra
c
er
Using
Propo
rtional
-I
ntegral
-Deri
v
ative Controll
ers
(Wij
a
ya Kurnia
wa
n)
491
In orde
r to se
arch for the o
p
timal value shu
n
t resi
sto
r
so that the area si
gnal b
e
c
ome
s
as
wide
a
s
p
o
ssible th
en t
he ab
ove e
q
uation
sho
u
ld
given by diff
erential
ag
ain
s
t R
sh
unt
which
is then filled a
value of zero
(0) in o
r
de
r to get the equ
ation.
0
,
∗
∗
(
3
)
The meth
od
use
d
in thi
s
resea
r
ch, the
move
ment
of the motor is de
scrib
e
d
in the
followin
g
flowcha
r
t. At the
begin
n
ing
of
the robot is e
x
ecuted, de
cl
ared
∆
T of 0.5 s and n
are
integers. T
h
e
next p
r
ocess i
s
to
multip
ly
∆
T
with
n
to p
r
odu
ce
odd
or
even
numbe
rs. If the
results
of od
d
n the
process p
e
rf
o
r
med
by
the sen
s
o
r
is rea
d
ing
lin
es right after the
completio
n
of the p
r
o
c
e
s
s of
rea
d
ing t
he ri
ght, the l
e
ft sen
s
o
r
re
ad line
s
. If gi
ven input li
ne
s a
r
e
bla
ck
a
nd
white enviro
n
m
ent,
the rea
d
ing pro
c
e
s
s if
it
finds
bla
c
k lin
es, th
en t
he
con
d
ition
1, if it doe
s
n
o
t
find the color
black (whi
ch i
n
this
ca
se
white) t
hen t
h
e
con
d
ition 0. I
f
conditio
n
s
a
r
e ri
ght sen
s
o
r
readi
ng
s an
d
sen
s
o
r
=1 lef
t
=0 then th
e
motor ri
ght-b
ack an
d left motor forwa
r
d, and vice versa.
Ho
wever, if the sen
s
o
r
rig
h
t and left=0 th
en both the m
o
tor will move
forwa
r
d.
St
a
r
t
Mo
v
e
Fo
r
w
a
r
d
or
Ba
c
k
w
a
r
d
Re
a
d
Ri
g
h
t
Se
n
s
o
r
Re
a
d
Le
f
t
Se
n
s
o
r
If
Ri
g
h
t
=
1
Le
f
t
=
0
If
Ri
g
h
t
=
0
Le
f
t
=
1
Ri
g
h
t
Mo
t
o
r
Fo
r
w
a
r
d
Le
f
t
Mo
t
o
r
Fo
r
w
a
r
d
Ri
g
h
t
Mo
t
o
r
Ba
c
k
w
a
r
d
Le
f
t
Mo
t
o
r
Fo
r
w
a
r
d
Ri
g
h
t
Mo
t
o
r
Fo
r
w
a
r
d
Le
f
t
Mo
t
o
r
Ba
c
k
w
a
r
d
If
n
Od
d
No
Ye
s
Ye
s
Ba
l
a
n
c
e
Re
a
d
Gyr
o
s
c
o
p
e
Se
n
s
o
r
If
Ba
l
a
n
c
e
on
U=K
p
.
e
+
Ki
∫
e.
d
t
+K
d
No
St
o
p
∆
T
.
n
If
n
Ev
e
n
De
c
l
a
r
e
∆
T
=
0.
5
S
n
=i
n
t
En
d
En
d
Ye
s
Ye
s
No
No
Figure 4. Algorithm which Implemente
d
in Rob
o
t
In orde
r to tune the sy
stem, requi
red
kno
w
led
ge
about the inf
l
uen
ce of of the PID
controlle
r tra
n
sie
n
t re
spo
n
se
and
stea
dy para
m
et
ers of the
syst
em that
wo
ul
d be
controll
ed.
Such kno
w
le
dge can be seen
fro
m
the
mathemati
c
a
l
equatio
ns t
hat of the PID controll
er [
11].
The natu
r
e of
PID paramet
ers a
r
e:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 489 – 49
6
492
P :
effect on the tran
sient state
(settling
time
=T
ss) and
ste
ady (stea
d
y state erro
r=Ess).
I
:
improve
ste
a
d
y state
(st
eady
state
error=
Ess) b
u
t usually
worsen
tran
si
ent (ari
sing
overshoot
=M
p).
D :
improve
the
t
r
an
sient
state
of the syste
m
(settling time=T
s,
overshoot=M
p) b
u
t
no effect at
all on the syst
ems
steady state.
Expresse
d co
ntrol actio
n
a
s
:
(
4
)
This type
of
cont
rolle
r u
s
ed to
imp
r
o
v
e the spee
d of re
sp
on
se, preve
n
t errors a
n
d
maintain a st
able ste
ady state [5].
3. Results a
nd Analy
s
is
In this chapt
er, the robot will
be tested three times. The firs
t test is a test gyroscope
sen
s
o
r
, thi
s
sensor are u
s
ed to
that the
ro
bot
can
st
a
nd u
p
well.
T
he
se
con
d
t
e
st
is
t
h
e
t
e
st
of
a
config
uratio
n
value Ki, Kp and Kd
with trial erro
r
met
hod of. Whil
e
the third te
st is a test a
gai
nst
the colo
r pat
h and the en
vironme
n
t, with the
test expectation
ca
n be determi
ned pe
rcenta
ge
rat
e
of
sy
st
e
m
s.
3.1. G
y
roscope Sensor T
est
First te
sting i
n
this re
se
arch is to
exami
ne the val
u
e of input
s obtaine
d
by Gyro
scop
e
sen
s
o
r
. Te
st
s p
e
rfo
r
me
d
by putting
a
robot
on
the
floor
with th
e
ro
bot p
o
sitio
n
"sl
e
epi
ng"
as
s
h
ow
n
a
t
F
i
gu
r
e
5(
a
)
.
Figure 5. Rob
o
t at a) “sle
ep
ing” conditio
n
;
b)
“stan
d
ing
”
con
d
ition wit
h
held; c)
“sta
nding
”
con
d
ition with
out being h
e
l
d
a
b
c
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Segwa
y
Lin
e
Tra
c
er
Using
Propo
rtional
-I
ntegral
-Deri
v
ative Controll
ers
(Wij
a
ya Kurnia
wa
n)
493
Then
do
the
captu
r
e value
by Gyrosco
p
e
50 ti
mes, after taking the
value of the
positio
n
of the bed 50 times the
value obtai
ned for the
equilibri
um, af
ter the
robot is pl
aced in the
positio
n of "
s
tandi
ng" Fi
g
u
re
5(b).
or
less tha
n
th
e an
gle
of 9
0
o
wh
en
the
ro
bot i
s
b
e
i
ng
positio
ned in
the "sleep" a
s
in Figure 5
(
a
)
.
Table 1. Gyro
scope Sen
s
o
r
Count
Test nu
mber -
G
y
rosc
op
e
Valu
e
1 594
2 593.5
3 599
--
--
48 595
49 599
50 592
A
v
e
r
age
596
After calculati
on 50 time
s, the average
was ta
ken an
d
the value re
sults 596. Thi
s
value
is the valu
e whe
n
the rob
o
t on sl
eep p
o
sition
be
ca
me
initial set value. After that the ro
bot
ca
n
stand
s a
s
sh
own at Figu
re
5.c, r
obot ca
n stand
s with
out being tou
c
he
d.
3.2. Ki, Kp dan Kd Te
st a
nd Analy
s
is
Giving value
for Kp, Ki an
d Kd is u
s
e
d
to balan
ce th
e rob
o
t thro
u
gh two
wh
eel
s that is
locate
d o
n
ri
ght an
d left
of the
rob
o
t, the th
ree
parameters
are
i
n
terconn
ecte
d. The
minim
u
m
value for Kp of 0.2 and ma
ximum 0.6 is whe
r
e t
he co
ndition of the robot can sta
nds eve
n
though
the tilt. The smaller the val
ue Ki increa
si
ngly l
eaning f
o
rward and if
the greate
r
the value of Kp
robot
s in
crea
singly l
eanin
g
ba
ckwa
rd th
ough
it can
st
and
s. As fo
r
Ki given mini
mum valu
e a
nd
a
maximum val
ue 3.8 3.2.
The
smalle
r
the rob
o
t
to
balan
ce th
e
con
d
ition of
the body mo
re
quickly, otherwise the g
r
ea
ter the valu
e
Ki Kp re
sp
o
n
s
e to the
slo
w
er. While the
value of Kd i
s
to
improve
rel
a
tions Kp
and
Ki in the
bod
y balan
ce
th
e robot. Th
e
followin
g
tabl
e sho
w
s the
trial
error inp
u
t of each value to
the equilibri
u
m
conditio
n
s
of the robot.
Table 2. Ro
b
o
t Trial Again
s
t Kp, Ki and Kd Value
Trial nu
mb
er
Kp Value
Ki Value
Kd Value
R
o
bo
t
Co
nd
i
t
ion
A
g
a
i
n
s
t
-
Kp
Ki
Kd
1
0.2
3.2
0.003
Lean
Back Rapid
Slow
Respo
n
se
2
0.2
3.4
0.005
Lean
Back Rapid
Fast
Response
3
0.2
3.6
0.003
Lean
Back Slow
Slow
Respo
n
se
4
0.2
3.8
0.005
Lean
Back Slow
Fast
Response
5
0.2
3.2
0.003
Lean
Back Rapid
Slow
Respo
n
se
6
0.2
3.4
0.005
Lean
Back Rapid
Fast
Response
7
0.2
3.6
0.003
Lean
Back Slow
Slow
Respo
n
se
8
0.2
3.8
0.005
Lean
Back Slow
Fast
Response
9 0.4
3.2
0.003
Perpendicular
Rapid
Slow
Respo
n
se
10 0.4
3.4
0.005
Perpendicular
Rapid
Fast
Response
11 0.4
3.6
0.003
Perpendicular
Slow
Slow
Respo
n
se
12 0.4
3.8
0.005
Perpendicular
Slow
Fast
Response
13 0.4
3.2
0.003
Perpendicular
Rapid
Slow
Respo
n
se
14 0.4
3.4
0.005
Perpendicular
Rapid
Fast
Response
15 0.4
3.6
0.003
Perpendicular`
Slow
Slow
Respo
n
se
16 0.4
3.8
0.005
Perpendicular
Slow
Fast
Response
17
0.6
3.2
0.003
Lean Fo
r
w
ar
d
Slow
Slow
Respo
n
se
18
0.6
3.4
0.005
Lean Fo
r
w
ar
d
Slow
Fast Response
19
0.6
3.6
0.003
Lean Fo
r
w
ar
d
Slow
Slow
Respo
n
se
20
0.6
3.8
0.005
Lean Fo
r
w
ar
d
Slow
Fast Response
21
0.6
3.2
0.003
Lean Fo
r
w
ar
d
Slow
Slow
Respo
n
se
22
0.6
3.4
0.005
Lean Fo
r
w
ar
d
Slow
Fast Response
23
0.6
3.6
0.003
Lean Fo
r
w
ar
d
Slow
Slow
Respo
n
se
24
0.6
3.8
0.005
Lean Fo
r
w
ar
d
Slow
Fast Response
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 489 – 49
6
494
De
scription of
robot conditi
ons a
gai
n
s
t the ea
ch pa
ra
meter on T
a
b
l
e 2:
The re
sp
on
se
to the Kp :
a.
Lean Ba
ck : the angl
e bet
wee
n
floor by
robot 70
o
-84
o
b.
Perpe
ndi
cula
r : the angle b
e
twee
n floor
by robot 85
o
-95
o
c.
Lean Fo
rward : the angle
betwe
en floor by robot 96
o
-105
o
d.
For an
other a
ngle ro
bot falls
The re
sp
on
se
to the Ki :
a.
Rapi
d : equi
libration
process conditio
n
s
r
obot i
s
f
a
st so that t
he movem
e
n
t
of
equilib
ration l
ooks ro
ugh a
nd it took 4 to 7 swin
g rob
o
t
to equilibrat
e
.
b.
Slow : e
quili
bration
p
r
oce
s
s is slo
w
so
that the
m
o
vement of t
he robot
co
n
d
ition
equilib
ration
smooth a
nd o
n
ly took 2 to 3 swi
ng ro
bot
to equilibrate
.
The re
sp
on
se
to the Kd
a.
Fast Response : robot equilibration
process response to Ki is fast
b.
Slow Response : robot equilibration
process response to Ki is slow
Based
on te
st result
s of tri
a
l error
by en
teri
ng the
re
spective value
of the com
b
i
nation of
Ki, Kp and Kd obtained th
e best co
mbi
nation of 0.
4; 3.8; and 0.005. In these con
d
ition
s
the
robot
ca
n
stand u
p
well.
Stand
well
with the
se
nse
that the
rob
o
t sta
n
d
s
u
p
rig
h
t, g
ood
equilibration process
a
nd quick
response.
3.2. Perform
ance Te
sting
In the
sub
s
e
quent te
sting
pro
c
e
s
s, testing the p
e
rfo
r
man
c
e
of th
e color sen
s
or. Colo
r
sen
s
o
r
is use
d
to detect the colo
r of the track
und
er the rob
o
t to recogni
ze the tracks. In
the
followin
g
ima
ge rea
d
ing
s
colo
r co
nfigu
r
ation is
d
o
n
e
by robots.
This test is u
s
ed to re
ad the
su
ccess rate sen
s
o
r
readi
n
g
s of
col
o
r when re
cog
n
iz
i
ng
six col
o
rs
inclu
d
ing: bl
a
c
k (colo
r
p
a
th
),
white (enviro
n
ment
p
a
th) as well as red,
yellow,
g
r
een
an
d blu
e
is th
e
colo
r di
sturb
a
n
c
e
is
locate
d between the environment an
d p
a
thway
s
su
ch
as the Figu
re
6.
Figure 6. Rob
o
t and the Pa
th
Path Specification :
a.
1 Path=2 Straight Pat + 4 Cure (2 L
e
ft and 2 Rig
h
t Cu
rve)
b.
1 Straight Path=3
1cm
c.
1 Curve Path
=21
c
m
d.
Colo
r Di
sturb
ance, rep
r
e
s
e
n
t with a box, each b
o
x 2cm
2
Tes
t
s
carried out ten times
a round to t
he ri
ght, and
ten times round to the left. Tes
t
ing
is don
e ea
ch
for the main
color
(the color line a
nd
environ
ment),
and col
o
r di
sturb
a
n
c
e (re
d
,
gree
n, blue, and re
d. To test the
colo
r
of the line, has four
kind
s
of testing that curve
s
to the
right, turn left, and on track
straight, each of wh
ich tested between
c
onditions of by time.
Con
d
ition, re
pre
s
ent
s the con
d
ition of the rob
o
t whe
n
to reco
gni
ze the line, giving the
para
m
eter i
s
if the robot to recogni
ze
the pat
h well
and can wa
lk it will be repre
s
e
n
ted b
y
para
m
eter "
p
ath", if the robot doe
s not
recogni
ze th
e
path prope
r
l
y
(out of line) then it will b
e
rep
r
e
s
ente
d
by para
m
ete
r
"fail". Whil
e the len
g
th
of time rep
r
ese
n
ts the ti
me u
s
ed i
n
the
para
m
eter "
w
ork", for the p
a
ram
e
te
r " fail" time is
not rec
o
rded.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Segwa
y
Lin
e
Tra
c
er
Using
Propo
rtional
-I
ntegral
-Deri
v
ative Controll
ers
(Wij
a
ya Kurnia
wa
n)
495
Table 3. Te
st by Path
Test N
u
m
b
er
Righ
t Cur
v
e
Left
Cur
v
e
Straigh
t
K
o
nd
i
s
i Wa
k
t
u K
o
nd
i
s
i
Wa
k
t
u
K
o
nd
i
s
i
Wa
k
t
u
1
Work 00.00.47
Work
00.01.00
Work
00.00.36
2
Work 00.00.44
Work
00.00.51
Work
00.00.20
3
Work 00.00.42
Work
00.00.51
Work
00.00.15
4
Fail -
Work
00.00.46
Work
00.00.21
5
Work 00.00.44
Fail
-
Work
00.00.16
6
Work 00.00.41
Work
00.01.35
Work
00.00.20
7
Fail -
Work
00.00.54
Work
00.00.23
8
Work 00.00.43
Work
00.00.48
Work
00.00.21
9
Work 00.00.44
Work
00.00.46
Work
00.00.22
10
Fail -
Work
00.00.42
Work
00.00.23
Success R
a
te
70
%
90
%
100
%
A
v
e
r
age
Success R
a
te
00.00.43,57
00.00.54,77
00.00.21,7
Re
sults of testing on the
Ri
ght cu
rve
s
along the 29
cm
have a su
ccess rate
of 70%
take
s a
n
ave
r
age
of 43.5
7
se
con
d
s,
whil
e when
curve
s
L
e
ft ha
s a
su
ccess
rate
of 90% a
nd
a
n
averag
e time
of 54.77 se
con
d
s. Fo
r the test
on a
straig
ht track ha
s a 10
0
%
succe
s
s b
y
an
averag
e time of 21.7 se
con
d
s.
Test
s for
col
o
r di
sturban
ce we
re d
one
10 times to t
he rig
h
t or lef
t. The su
ccess of the
detectio
n
sen
s
or to dete
c
t
is if the di
sturban
ce i
s
read
ing di
sturb
a
n
c
e that di
stu
r
ban
ce i
s
at th
e
right
or left. Dis
r
uption if the delay
is
more than
5
s
e
conds
(Fail / *)
, if the delay is
not more than
5 se
con
d
s wi
thout any disturban
ce. Parameter d
a
sh
(-) in the table is whe
n
no
t testing. If
the
test is don
e o
n
the diso
rde
r
on the Right,
then on
the left there is no
disturb
a
n
c
e
and othe
rwi
s
e.
Table 4. Trial
Against Di
stu
r
ban
ce
Colo
r
Trial
Red
Yello
w
Blue
Green
Righ
t
Left
Righ
t
Left
Righ
t
Left
Righ
t
Left
1
00.00.01
- 00.00.01
- 00.00.29*
-
00.00.01
-
2
00.00.02
- 00.00.01
- 00.00.55*
-
00.00.01
-
3
00.00.01
- 00.00.02
-
00.00.04
-
00.00.01
-
4
00.00.02
- 00.00.03
-
Fail
-
00.00.01
-
5
00.00.01
- 00.00.01
- 00.02.20*
-
00.00.02
-
6
-
00.00.01
-
00.00.01
- 00.00.54*
-
00.00.54
7
- 00.00.01
-
00.00.02
-
Fail
-
00.04.12
8
-
00.00.02
-
00.00.01
- 00.01.07*
-
00.01.07*
9
- 00.00.01
-
00.00.02
-
00.00.02
-
00.00.45
10
-
00.00.01
-
00.00.02
- 00.04.12*
-
Fail
Success
Rate
100%
100%
100%
100%
20%
20%
100%
60%
success
rate
agains
t the
color
100%
100%
20%
80%
A
v
e
r
age
Success
Rate
75%
Re
sults
of te
sting to
dete
c
t tampe
r
ing
colo
rs for
ea
ch
colo
r of b
o
th rig
h
t and
left each
colo
r (rig
h
t + left
/
2).
F
o
r red colo
r
di
stu
r
ban
ce
can
b
e
dete
c
ted
wi
th accu
ra
cy o
f
100%, 1
0
0
%
yellow, 20% blue an
d gre
e
n
colo
rs
can
be dist
in
gui
sh
ed by the level of accura
cy of 80%.
4. Conclusio
n
After doin
g
some te
sting
on the
ro
bot, it wa
s
con
c
l
uded
in thi
s
study. PID al
gorit
h
m
su
ccessfully impleme
n
ted
on the
ro
bot so that ro
bot
doesn’t fall.
PID is used to set the ro
b
o
t
equilib
rium, i
n
this resea
r
ch the be
st co
mbination val
ue of Ki, Kp and Kd a
r
e u
s
ed to
adju
s
t th
e
balan
ce of e
a
ch robot is
0.4;
3.8; and 0.005. At th
e a straig
ht
path, succe
s
s rate the rob
o
t
in
passin
g
alon
g a strai
ght line 31 cm is 100% by
an
averag
e tra
v
el time 00.00.21,7 se
co
n
d
s,
w
h
er
e
a
s
wh
en
te
s
t
e
d
u
n
der
c
o
nd
itio
ns
cu
r
v
e
s
to
the ri
ght has a succe
ss rate of 70% by averag
e
travel time 0
0
.00.43,57
a
nd curve
s
to
the left
by 90
% by avera
g
e
time 00.0
0
.54,77. Th
e ro
bot
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 489 – 49
6
496
can distin
guish
between co
lors distu
r
b
a
n
c
e
a
gain
s
t th
e col
o
r p
a
th, for the red a
nd yello
w col
o
r
has a
succe
s
s
rate th
at i
s
very g
ood
a
t
100%, fo
r t
he
colo
r
hurd
l
e g
r
ee
n h
a
s a
su
cce
s
s rate
whi
c
h i
s
pretty good
by
80
%, while
for the bl
ue
color is not
goo
d
only for 20%,
and
so th
at t
h
e
averag
e
su
ccess
rate i
s
7
5
%. So it
ca
n be
ca
tego
ri
zed
a
s
wo
rki
ng
well, the
robot d
o
e
s
n’t
fall
with the
com
b
ination
valu
e of PID an
d
and
rob
o
ts th
at wo
rk
well
whe
n
p
a
ssin
g ob
sta
c
le
s
colors
except blu
e
color.
Referen
ces
[1]
Saputro BY, A
mri D, Ja
yanti
SD.
Co
mpar
is
on of C
ontrol
Methods P
D
, PI, and PID on
T
w
o W
heele
d
Self Bala
nci
ng
Rob
o
t
. EECSI
201
4. Yog
y
ak
a
r
ta, Indonesi
a
. 201
4.
[2]
Jung T
,
Kim HW
, Jung S.
Impl
e
m
e
n
tatio
n
an
d C
ontro
l
of Bal
anci
n
g
Lin
e
T
r
acer
Using
Visi
on
.
Internatio
na
l C
onfere
n
ce o
n
Ubiq
uito
us Ro
bots and
Am
bi
ent Intelli
ge
nce
.
Incheon. 20
1
1
; 8: 858-8
62.
[3]
Ngu
y
e
n
HG, K
ogut G,
Barua
R, Burmeister
A.
A Segw
ay RMO-Based
Rob
o
tic T
r
ans
port Syste
m
.
SPIE Proc. Philad
hel
phi
a. 20
0
4
: 5609.
[4]
Li L, Xi
e J, Li W
.
F
u
zz
y
Ada
p
t
ive PID
Contro
l of a Ne
w
H
y
d
r
aulic Erecti
ng
Mecha
n
ism.
TELKOMNIKA
T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol.
20
13; 11(4): 7
15-
724
[5]
Canale M, Brunet SC.
A Leg
o Mindstor
m
N
X
T
experi
m
e
n
t for Model Pre
d
ictive C
ontrol
Educatio
n
.
Europ
e
a
n
Cont
rol Co
nferenc
e
.
Z
u
rich. 2013: 254
9-25
54.
[6]
Ricard
o
V
C
A, Enri
que
HP,
Gabri
e
l
HAM
.
A Li
ne
foll
o
w
er robot
i
m
p
l
e
m
e
n
tation
u
s
ing
Le
go
’
s
Mindstor
m
s Kit
and Q-Lear
ni
n
g
. ACT
A
Unive
r
sitaria. Guan
aj
uato. 22: 11
3-1
18.
[7]
Oliveira G, Silva R, Lira T
,
Reis
LP.
Env
i
r
o
n
m
e
n
t Map
p
i
ng
usin
g th
e
Leg
o Mi
ndstor
m
s
NXT
an
d
leJOS NXJ
. Un
iversid
a
d
e
do
Porto. Porto.
[8]
Henr
yra
nu
BP
, Nur
w
ars
i
to
H, Kurni
a
w
a
n
W
.
One-T
i
me Pass
w
o
rd I
m
pleme
n
tatio
n
on
Leg
o
Mindstorms NX
T
.
T
E
LKOMNIKA T
e
leco
mmu
n
ic
ation C
o
mp
utin
g Electr
onics a
nd Co
ntrol
. 201
4;
12(3): 68
9-6
9
4
.
[9]
Konca
n
d
a
M, W
ilke MB, Ball
antin
e DS
.
Using
Le
go
Mindstorms
NXT
Rob
o
tics
Kits as a
Spectrop
hom
etric Instrument.
Internatio
na
l Journ
a
l o
n
Sma
r
t Sensing
and
Intellig
ent Sys
t
ems.
20
10;
3(3): 201
0.
[10]
T
r
ussel
M
.
Engagi
ng
w
i
t
h
the
NXT
Protot
y
p
e
Board. Spri
ng
T
e
rm. S
w
iss F
eder
al Institute
T
e
chnol
og
y
of Z
u
rich. 200
9
.
[11]
Gunterus, F
.
F
a
lsafah
Dasa
r Sistem Pen
g
end
ali
an Pr
os
es. Jakarta: P
T
Elex Med
i
a
Komputi
n
d
o
.
199
4.
Evaluation Warning : The document was created with Spire.PDF for Python.