Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
10
,
No.
3
,
June
2020
,
pp. 3
066~
3073
IS
S
N: 20
88
-
8708
,
DO
I: 10
.11
591/
ijece
.
v10
i
3
.
pp3066
-
30
73
3066
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
T
raject
ory
r
econ
struction
for r
obot pr
ogra
mm
ing
by dem
on
stration
Red
a
H
an
i
fi El
ha
chemi
Am
ar
1
, L
ared
j
Be
nchikh
2
, H
ak
i
ma Dermec
he
3
,
Ouamr
i
Bachi
r
4
,
Z
ou
bir A
h
med
-
F
oitih
5
1
,3,5
Dépa
rtment
d
’él
e
ct
roniqu
e, Unive
rsit
é
des
Sci
enc
es
et de la Te
chnol
ogi
e
d’Ora
n,
Alger
ia
2
Univer
sité d’ E
vr
y
-
Va
l
-
d’Essonne,
La
bor
at
oir
e
I
BISC
,
Franc
e
4
Dépa
rte
m
ent de
T
e
chnol
ogi
e, U
nive
rsit
é
de
Be
c
har
,
Alg
eria
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r 11
, 201
9
Re
vised
Jan
8
,
2020
Accepte
d
Ja
n 2
0
, 2
020
The
re
pro
du
ct
i
on
of
h
a
nd m
o
vem
ents b
y a r
obot r
em
ai
ns
d
ifficult
and
conve
ntion
al
le
arn
i
ng
m
et
ho
ds
do
not
al
low
us
to
fait
hfu
ll
y
recreate
these
m
ov
em
e
nts
because
it
is
ve
ry
diff
ic
ult
w
he
n
the
nu
m
ber
of
cro
ssi
ng
po
i
nts
is
ve
ry
la
r
ge
.
P
rogr
am
m
ing
by
Dem
on
stra
ti
on
giv
e
s
a
be
tt
er
oppo
rtun
it
y
f
or
s
olv
i
ng
this
pro
blem
by
tracking
the
us
er
’
s
m
ov
e
m
ents
with
a
m
otion
c
aptu
re
syst
e
m
and
c
reati
ng
a
robo
ti
c
pro
gr
am
to
rep
r
oduce
the
perform
ed
ta
sk
s.
T
his
pa
pe
r
presents
a
Program
m
ing
by
Dem
on
strat
ion
syst
em
in
a
traj
ect
or
y
le
vel
f
or
t
he
re
pro
duct
ion
of
ha
nd/t
ool
m
ov
em
ent
by
a
m
anipu
l
at
or
r
obot;
this
was
reali
z
ed
by
trac
king
the
us
e
r’
s
m
ove
m
ent
with
the
Ar
T
oolkit
and
reconst
ru
c
ti
ng
the
tra
j
ec
tories
by
us
in
g
the
c
on
st
rained
c
ubic
sp
li
ne.
T
he
re
su
lt
s
obta
ined
with
the
co
ns
t
raine
d
cu
bic
s
pline
we
re
com
par
ed
wit
h
cu
bic
sp
li
ne
interp
olati
on
.
Finall
y
the
ob
ta
ine
d
trajecto
ries
ha
ve
been
sim
ulate
d
in
a
virtu
al
e
nviro
nm
ent
on
the P
um
a 6
00 robot
.
Ke
yw
or
d
s
:
In
te
r
pola
ti
on
Moti
on capt
ur
e
Pr
og
ram
m
ing
by d
em
on
strat
i
on
Traj
ect
or
y
rec
onstr
uction
Copyright
©
202
0
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
:
Re
da Hanifi
Elhachem
i Am
ar
,
Dép
a
rtem
ent d
’élect
roniq
ue
,
Un
i
ver
sit
é
des Sci
ences et
de l
a Tech
no
l
og
ie
d’Oran
,
El m
nao
uar
B
P
1505, B
ir El
D
j
irs
3100
0 (Ora
n) A
l
gér
ie
.
Em
a
il
:
red
a.h
a
nifi@
un
i
v
-
us
to
.d
z
1.
INTROD
U
CTION
Since
the
in
ve
ntion
of
t
he
fir
st
robo
ts,
t
he
r
epro
du
ct
io
n
of
hu
m
an
m
ov
em
ent
is
sti
l
l
a
chall
eng
i
ng
su
bject
i
n
r
ob
otics.
T
his
r
e
pro
du
ct
io
n
ca
n
be
div
ide
d
into
t
wo
cat
egories:
the
f
irst
o
ne
is
i
m
it
a
ti
on
of
the
hu
m
an
m
ov
e
m
ent
as
i
t
is
by
the
ro
bot
to
rea
li
ze
the
ta
sk
.
Re
fer
e
nces
[
1,
2]
they
i
m
it
at
e
the
hu
m
an
m
ov
e
m
ent
to
reali
ze
diff
e
re
nt
ta
sk
s
li
ke
wr
it
in
g
an
d
openi
ng
doors,
their
ap
proac
h
was
de
rive
d
from
the hum
an
f
unct
ion
in
g base
d on t
he b
od
y sc
hem
a and
the
percept.
I
n
the
work of
Jie
a
nd al.
[3
]
,
they
f
ocus
on
the
pose
im
it
ation
betwee
n
a
hu
m
an
an
d
a
hum
ano
id
r
obot
and
pro
po
se
d
a
po
se
sim
il
ari
ty
m
e
tric
based
on
the
sha
re
d
str
uc
ture
of
the
m
otion
spa
ces
of
hum
an
and
r
obot
.
T
he
sec
ond
re
pro
duces
t
he
hu
m
an
m
ov
e
m
ent
by
ta
king
as
r
efere
nce
only
the
ha
nd
(the
end
-
ef
fecto
r)
a
nd
ne
glect
ing
how
a
nd
wh
e
r
e
the
oth
e
r
joi
nts
are
po
sit
io
ne
d
(s
houl
der,
el
bow
,
and
wr
ist
)
.
I
n
the
work
of
Be
nch
i
kh
[
4
]
,
in
w
hich
he
reali
zed
a
syst
e
m
of
a sync
hrono
us
repr
od
uction o
f
the
hum
an
m
ov
em
ent, foll
owin
g
the
m
ov
em
ent o
f
a sin
gl
e point
of inter
est
.
On
e
can
noti
ce
two
cat
eg
or
i
es
of
r
obot
pro
gr
am
m
ing
m
e
t
hods
,
the
first
on
e
is
m
anu
al
li
ke
the
te
xt
base
d
syst
em
s
,
gr
a
phic
al
syst
e
m
s
and
the
te
ach
-
penda
nt
program
m
ing;
the
second
on
e
is
the
a
ut
om
atic
pro
gr
am
m
ing
su
c
h
as
the
i
nst
ru
ct
io
n
a
nd
the
obse
r
vatio
n
ba
sed
pro
gra
m
m
ing
.
T
he
F
igure
1
re
pr
e
sent
s
the
cat
egorizat
ion
of
these
m
et
ho
ds
m
ade
by
[
5
]
.
F
or
a
pp
li
cat
io
ns
suc
h
as
im
it
ation
an
d
re
pro
duct
ion
of
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Tra
je
ct
or
y
rec
on
str
uctio
n
fo
r
robo
t
p
r
ogrammi
ng b
y
d
e
m
on
str
atio
n
(
Re
da Ha
nifi
Elhachemi
Am
ar
)
3067
hu
m
an
m
ov
em
ent;
con
ve
nt
ion
al
le
ar
ning
m
et
ho
ds
as
t
ext
pro
gr
am
m
i
ng,
grap
hical
syst
e
m
s
or
the
te
ach
-
penda
nt
progr
a
m
m
ing
a
re
not
s
uitable
.
E
sp
eci
al
ly
wh
e
n
th
e
tra
j
ect
or
ie
s
to
repr
oduc
e
are
com
plex
a
nd
the
num
ber
of
cro
ssi
ng
poin
ts
are
high.
M
os
t
of
the
rece
nt
w
orks
f
or
hum
an
m
ov
em
e
nt
repr
oductio
n
us
e
the
r
obot
ob
s
erv
at
io
n
base
d
pr
ogram
m
ing
.
In
t
his
ca
te
gory,
we
fi
nd
w
hat
is
c
al
le
d
Program
m
ing
by D
em
on
strat
ion
or Lea
rn
i
ng
by w
at
c
hing
in o
t
her rea
dings
[1
–
4
,
6
,
7
].
Figure
1.
Ro
bot pro
gr
am
m
ing
m
e
tho
ds
Pr
og
ram
m
ing
by
Dem
on
stra
t
ion
(
PbD
)
is
a
robo
t
progra
m
m
ing
m
et
ho
d
base
d
on
the
extracti
on
of
data
dire
ct
ly
from
the
visu
al
iz
at
ion
of
t
he
use
r’
s
pe
rfor
m
ance.
It
is
a
pro
m
isi
ng
autom
at
ic
m
et
ho
d,
w
hi
ch
can
per
m
it
to
a
us
e
r
with
li
tt
le
or
no
e
xperti
se
to
program
ro
bot
ta
sk
s
[
8
,
9
]
.
In
Pb
d
syst
em
s,
t
he
te
acher
pe
rfor
m
s
the
ta
sk
w
hile
a
le
ar
ning
int
erf
ace
rec
ords
the
m
ov
em
ent
and
act
io
ns
carried
out
du
rin
g
the
pe
rfo
r
m
ance.
Diff
e
re
nt
inter
faces
ca
n
be
us
e
d.
Kinest
hetic
guida
nc
e
te
acher
m
ov
es
the
r
obot
li
nk
s
m
anu
al
ly
and
the
traject
or
ie
s
are
rec
orde
d
[
7
]
.
The
di
rect
co
ntro
l
a
nd
te
le
ope
rati
on
i
nterf
ace
s
are
al
so
us
ed
in
Pbd
syst
e
m
s
[
10
]
.
In
te
r
faces
base
d
on
sens
ors
(
visio
n,
m
agn
et
ic
and
inerti
a)
are
widely
use
d
in
P
bd
syst
e
m
s.
Ther
e
is
a
va
riet
y
of
trackin
g
te
chn
iq
ues
in
visio
n
syst
e
m
s
[1
1
–
1
4
]
and
i
t
has
an
ad
va
nt
age
over
the
oth
e
r
sens
or
-
bas
ed
m
et
ho
ds,
beca
us
e
no
sens
ors
nee
d
to
be
at
ta
ched
to
the
te
acher
.
A
nd
the
visio
n
syst
em
s
al
lo
w
the
te
ache
r
to
ha
ve
m
or
e
na
tural
pe
r
form
ance
without
bein
g
disturbe
d
by
the
m
ater
ia
l.The
seq
ue
nce
m
easur
e
m
ents
data
usual
ly
goes
t
hroug
h
a
scal
ing
ste
p
th
rou
gh
i
nter
po
ll
at
ion
s.
F
or
t
he
tim
e
series
da
ta
an
d
from
diff
eren
t
ty
pes
of
inter
po
la
ti
on
te
chn
i
qu
e
s,
poly
no
m
ia
l
interpo
la
ti
ons
are
the
m
os
tl
y
us
ed.
The
cub
ic
sp
li
ne
inter
pol
at
ion
wa
s
us
e
d
to
r
eco
ns
tr
uc
t
the
Ca
rtesi
an
tra
j
ect
or
ie
s
i
n
[
1
5
]
a
nd
N
on
-
Un
i
form
Ra
t
ion
al
B
-
Sp
li
ne
s
(
N
URBS)
for
th
e
trajecto
ry
a
ppr
ox
im
at
ion
[1
6
]
.
T
he
L1
sp
li
nes
for
the
inter
pola
ti
on
a
nd
pr
ese
r
vation o
f
the tra
j
ect
ory
[1
7
-
1
9
].
In
this
pap
e
r,
we
pro
pose
a
Pb
d
syst
em
in
trajecto
ry
le
vel
b
ased
on
visu
al
trac
king
syst
em
(ArT
oo
l
Kit)
to
trac
k
the
ha
nd
/
to
ol
m
ov
e
m
ent.
In
this
work
we
us
e
d
the
c
ub
ic
s
pl
ine
inter
pola
ti
on
to
reconstr
uct
the
Ca
rtesi
an
traje
ct
or
ie
s.
Com
par
ed
to
previ
ous
wo
r
ks
;
the
co
ns
trai
ne
d
cu
bic
sp
li
ne
interp
ol
at
ion
is
si
m
ple
to
i
mp
le
m
ent
and
ef
fici
ent.
Finaly
the
cap
uted
m
ov
em
ents
are
rep
r
oduce
d
by
a m
anipu
la
to
r
r
obot
in
a sim
ulate
d
env
ir
on
m
ent.
2.
THE
TR
ACK
ING
SYST
EM
Fo
r
t
he
ha
nd/
To
ol
trackin
g,
we
pro
pose
to
us
e
a
visi
on
-
ba
sed
trac
king
syst
e
m
.
These
syst
e
m
s
us
e
i
m
age
-
processi
ng
m
e
tho
ds
to
cal
culat
e
the
ca
m
era
pose
r
el
at
ive
to
real
world
ob
j
ect
s
and
giv
e
t
he
posit
ion
and
ori
entat
io
n
of
t
hese
obje
ct
s.
I
n
our
w
ork,
we
us
e
t
he
ARToolKit
for
trac
king
t
he
us
er
’s
ha
nd/t
oo
l.
We
fi
xed
the m
ark
ers
on
the
faces o
f
a
cu
be
(the
sam
e
m
a
rk
e
r
f
or
al
l
fac
es).
T
his w
il
l
al
low
the
cam
era
to
se
e
at
le
ast
on
e
m
ark
e
r
an
d
perm
it
us
to
avo
i
d
occlusi
ons.
Ba
sed
on
the
works
[
20
,
21
]
we
hav
e
use
d
si
m
ple
m
ark
ers wit
h
a
3
0% bor
der
w
idth to
ha
ve
a
bette
r
detect
io
n
of the m
ark
er and
av
oid
at the m
axi
m
u
m
t
he
false
identific
at
ions.
In
the
F
i
gure
2,
we
ca
n
se
e
an
e
xam
ple
of
t
he
m
ark
ers
that
are
us
in
g.
T
he
operat
or
will
perform
the
m
ov
em
ents
an
d
the
syst
e
m
wil
l
track
the
po
s
it
i
on
of
the
m
ark
e
r
f
or
t
he
hand/
to
ol
an
d
giv
es
the
posit
ion
(
x,
y,
z)
e
ve
r
y
sa
m
pling
ti
m
e
(th
e
posit
ion
is
cal
c
ulate
d
f
ro
m
the
centre
of
the
cub
e
).
The
ac
qu
isi
ti
on
cam
era
was
set
at
20
f
ram
e
s/s
to
lim
i
t
m
a
rk
e
r
m
iss
identific
at
ion
s.
T
he
slow
f
ram
e
rate
and
so
m
e
m
i
ss
identific
at
ion
s
cre
at
ed
gap
es
an
d
the
re
fore,
a
n
inter
po
la
ti
on
of
the
ac
quire
d
trac
k
in
g
dat
a
wa
s
m
and
at
ory
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, N
o.
3
,
J
une
2020
:
30
66
-
3073
3068
Figure
2
.
Exa
m
ple o
f
the
ArTo
olKit m
ark
ers
3.
THE
C
O
NS
T
RA
I
NE
D CUB
IC
SPLI
NE
The
pr
i
nciple
beh
i
nd
co
ns
tra
ined
c
ubic
sp
li
ne
is
to
preve
nt
ov
e
rs
hootin
g
and
el
i
m
inati
ng
os
ci
ll
at
ion
by
sacrifici
ng
sm
oo
thn
ess
.
I
n
this
interp
ol
at
ion
,
we
re
pl
ace
the
equ
al
it
y
of
the
secon
d
order
de
riv
at
ives
at
ever
y
point
by
a
sp
eci
fie
d
first
orde
r
der
i
vatives
[
22,
23
].
The
co
ns
tr
uc
ti
on
of
the
c
on
strai
ned
c
ubic
sp
li
ne
functi
on
is bas
ed on t
he f
ollo
wing c
rite
ria:
-
Curves a
re th
i
r
d order
poly
no
m
ia
ls
(
)
=
3
+
2
+
+
(1)
-
Curves
pass
th
rou
gh all
the
know
n po
i
nts
-
First o
r
der
de
rivati
ve,
is
the
s
a
m
e fo
r
both
fun
ct
io
ns o
n
ea
ch
si
de of
a
poi
nt.
′
(
)
=
+
1
′
(
)
(2)
-
Boun
dar
y c
ondi
ti
on
s ar
e
the
s
a
m
e as
for
t
he natu
ral cu
bic s
pline.
1
′′
(
0
)
=
′′
(
)
=
0
(3)
-
The
sec
ond o
r
der de
rivati
ve
is re
placed
by a
sp
eci
fie
d first
order de
rivati
ve
at eve
ry poin
t.
′
(
)
=
+
1
′
(
)
=
′
(
)
(4)
The
m
ai
n
ste
p
beco
m
es
the
ca
lc
ulati
on
of
the
slo
pe
f
or
eac
h
po
i
nt.
Natu
rall
y,
we
know
t
he
slo
pe
will
be
betwee
n
th
e
slop
e
s
of
t
he
ad
j
acent
st
raigh
t
li
nes
,
a
nd
it
sh
ould
a
ppr
oach
zer
o
if
t
he
slop
e
of
ei
th
er
li
ne
appr
oach
es
zer
o.
′
(
)
=
2
/
(
+
1
−
+
1
−
+
−
−
1
−
−
1
)
(5
)
′
(
)
=
0
, if
t
he
slo
p
c
ha
ng
e
s sig
n
at
t
his point.
The
e
quat
ion (
5) is use
d o
nly
for
the
interm
ediat
e points,
in
the e
nd points:
1
′
(
0
)
=
3
(
1
−
0
)
2
(
1
−
0
)
−
′
(
1
)
2
(6)
′
(
)
=
3
(
−
−
1
)
2
(
−
−
1
)
−
′
(
−
1
)
2
(7)
In
t
his
inte
rpol
at
ion
,
t
her
e
is
no
nece
ssit
y
to
s
olv
e
a
syst
em
of
equat
ion
because
the
slop
e
at
each
po
i
nt
is
known.
Ba
sed
on
the
two
a
djace
nt
po
i
nts
on
e
ach
side.
We
can
cal
cu
la
te
ever
y
sp
li
ne
f
un
ct
io
n;
as g
i
ven b
y
(1)
b
y
us
in
g
(8)
t
o (
13).
′′
(
−
1
)
=
2
[
′
(
)
+
2
′
(
−
1
)
]
(
−
−
1
)
+
6
(
−
−
1
)
(
−
−
1
)
2
(8
)
′′
(
)
=
2
[
2
′
(
)
+
′
(
−
1
)
]
(
−
−
1
)
+
6
(
−
−
1
)
(
−
−
1
)
2
(9
)
Finall
y, eve
ry
po
ly
nom
ia
l i
s
cal
culat
ed
f
r
om
the foll
ow
i
ng
par
am
et
ers:
=
′′
(
)
−
′′
(
−
1
)
6
(
−
−
1
)
(10)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Tra
je
ct
or
y
rec
on
str
uctio
n
fo
r
robo
t
p
r
ogrammi
ng b
y
d
e
m
on
str
atio
n
(
Re
da Ha
nifi
Elhachemi
Am
ar
)
3069
=
′′
(
−
1
)
−
−
1
′′
(
)
2
(
−
−
1
)
(11)
=
(
−
−
1
)
−
(
2
−
−
1
2
)
−
(
3
−
−
1
3
)
2
(
−
−
1
)
(12)
=
−
1
−
−
1
3
−
−
1
2
−
−
1
(13)
4.
THE
R
OB
OT’S
MO
DEL O
F MOT
IO
N
In
our
w
ork,
we
us
e
a
m
anipu
la
to
r
r
obot
m
od
el
,
the
P
um
a
60
0.
T
he
par
am
et
ers
of
this
rob
ot
are
s
how
n
in
T
able
1
[
2
4
,
2
5
]
.
Moti
on
m
od
el
of
s
uch
a
m
echan
ism
is
us
ua
ll
y
descr
i
bed
by
th
e
f
ol
lowi
ng
m
at
rix
eq
uatio
n:
Γ
=
M
(
q
)
q
̈
+
C
(
q
,
q
̇
)
q
̇
+
G
(
q
)
+
F
(
q
̇
)
(14)
w
he
re
:
Γ
: Vecto
r of
act
uato
r jo
i
nt to
rque
M
(
q
)
: In
e
rtia
m
at
rix
C
(
q
,
q
̇
)
q
̇
: Vecto
r of
ce
nt
rifugal an
d
C
ori
olis to
rque
G
(
q
)
: Vecto
r of
gra
vitat
ion
al
to
rqu
es
F
(
q
̇
)
: Vecto
r of
act
uato
r jo
i
nt frict
ion
f
or
ces
q
,
q
̇
,
q
̈
: Ar
e
r
es
pecti
ve
ly
, th
e joi
nt a
ng
le
,
v
el
ocity
, a
nd accel
erati
on
vecto
rs
To
e
ns
ure
the
li
near
iz
at
ion
of
the
no
nlinear
syst
e
m
descr
ibed
by
(
1
4
)
in
cl
os
ed
lo
op,
we
intr
oduc
e
a
li
near
iz
at
ion
con
tr
ol
syst
em
based
on
e
xacti
ng
knowl
edg
e
of
the
r
obot
m
od
el
and
it
s
i
m
ple
m
entat
ion
.
In
this
co
ntr
ol
syst
e
m
,
the
lo
op
of
the
li
nea
rizat
ion
is
ac
hi
eved
by
c
hoos
i
ng
a
to
rque
Γ
app
li
ed
to
th
e
r
obot,
as foll
ow
:
=
(
)
0
+
(
,
̇
)
̇
+
(
)
+
(
̇
)
(15)
Γ
0
is
an
a
uxil
ia
ry
input
of
t
he
se
le
ct
con
tr
oller.
A
pro
portio
na
l
der
iv
at
ive
c
ontr
ol
(PD)
is
a
ty
pical
cho
ic
e
and
it
is g
ive
n by the e
qu
at
io
n:
0
=
̈
+
(
̇
−
̇
)
+
(
−
)
(16)
Table
1.
Param
et
ers
of
t
he
P
um
a 6
00 m
anip
ulator
robot
Para
m
eters
Valu
es
Mass o
f
the f
irst bo
d
y
10
.
5
2
1
Kg
Mass o
f
the seco
n
d
bo
d
y
10.
2
3
6
Kg
Mass o
f
the th
ird b
o
d
y
8.
7
6
7
Kg
Co
ef
f
icien
t o
f
vis
c
o
u
s f
riction
2.
5
2
N.
m
.s/
rd
Co
ef
f
icien
t o
f
vis
c
o
u
s f
riction
7
N.
m
.
s/rd
Co
ef
f
icien
t o
f
vis
c
o
u
s f
riction
1.
7
5
N.
m
.s/
rd
Co
ef
f
icien
t o
f
dry
f
riction
3.
6
N.
m
.
s/rd
Co
ef
f
icien
t o
f
dry
f
riction
10
N.
m
.s/
rd
Co
ef
f
icien
t o
f
dry
f
riction
2
.
5
N.
m
.
s/rd
Leng
th
of
the f
irst bo
d
y
0.
149
m
Leng
th
of
the seco
n
d
bo
d
y
0.
432
m
Leng
th
of
the th
ird
bo
d
y
0.
431
m
By
r
eplaci
ng
q
̈
=
Γ
0
in the
(1
6)
,
w
e
get:
0
=
̈
+
̇
+
(17)
e
=
q
d
−
q
: Vecto
r of
t
he po
sit
io
n
e
rror
e
̇
=
q
̇
d
−
q
̇
: Vecto
r of t
he
v
el
ocity
er
r
or
e
̈
=
q
̈
d
−
q
̈
: Vecto
r of t
he
accel
erati
on er
ror
q
d
,
q
̇
d
,
q
̈
d
:
A
re r
e
sp
ect
ively
v
ect
or
s
of
desire
d po
sit
ion,
velocit
y an
d
acce
le
rati
on.
k
p
,
k
v
: Gai
n
m
at
ri
ces o
f
the
P
D
c
on
t
ro
ll
er
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, N
o.
3
,
J
une
2020
:
30
66
-
3073
3070
The
er
r
or
(
17)
is
a
li
near
di
ff
e
ren
ti
al
eq
ua
ti
on
of
sec
ond
order.
Wh
ere
K
p
and
K
v
are
de
fine
d
po
sit
ive
dia
gonal
m
at
rices,
so
the
cl
os
e
d
lo
op
s
yst
em
beco
m
es
li
near
de
coupled
.
T
he
va
lues
of
the
a
nd
in the si
m
ulati
on are:
=
(
350
0
0
0
350
0
0
0
350
)
=
(
35
0
0
0
30
0
0
0
35
)
In the
F
ig
ur
e
3, we
can
see t
he
i
m
ple
m
entat
i
on of t
he
c
om
pu
te
d To
rque c
ontr
oller on t
he Pum
a robo
t.
Figure
3
.
Im
ple
m
ented
c
om
pu
te
d
t
orq
ue
c
ontr
oller
5.
SIMULATI
O
N AND
RES
U
LT
S
In
this
sect
io
n,
Since
the
tra
j
ect
ori
es
obta
ined
from
the
trackin
g
syst
em
sh
ou
l
d
be
inter
po
la
te
d,
we
sta
rte
d
by
te
sti
ng
the
co
nst
rained
cu
bic
sp
li
ne
i
nterpol
at
ion
on
di
ff
e
r
ent
data
set
s.
The
Fig
ur
es
4
and
5
sh
ow
a
com
pa
r
ison
b
et
wee
n
the
co
ns
trai
ne
d
cub
ic
in b
lue
li
ne
an
d
the
co
nventio
nal
cu
bi
c
sp
li
ne
inter
pola
ti
on
in
red
li
ne,
the
red
ci
rcles
are
the
interpo
la
te
d
data
points.
I
n
the
first
te
sts,
we
us
e
d
both
of
the
interp
ol
at
ion
m
et
ho
ds
t
o
rec
on
st
ru
ct
ha
ndwr
it
in
g
tra
j
ect
or
y
m
ade
at
al
m
os
t
the
sa
m
e
sp
ee
d.
We
have
ch
os
en
to
rec
on
st
ru
c
t
the
le
tt
er
M
usi
ng
both
of
t
he
cu
bic
sp
i
ne
i
nter
po
la
ti
on
a
nd
the
c
onstrai
ne
d
c
ub
ic
s
plin
e
beca
us
e
this
le
tt
er
con
ta
in
s
sud
de
n
directi
onal
c
hanges w
hic
h
will
help
us
to see
the
be
hav
i
our
of
th
e u
se
d
interp
olati
on
m
et
ho
d.
In
the
F
i
gure
4
(a
)
an
d
(
b)
,
we
can
see
res
pecti
vely
the
r
esults
of
i
nter
po
la
ti
on
f
or
th
e
X
-
a
xis
an
d
Y
-
a
xis
coor
din
at
es
an
d
in
the
F
ig
ur
e
4
(c
)
we
ca
n
s
ee
the
rec
on
st
r
ucted
tra
j
ect
or
y.
Fo
r
t
he
res
ul
ts
in
the
F
igur
e
4
(
b)
we
see
that
both
of
the
m
eth
ods
ga
ve
t
he
sam
e
resu
lt
s
this
due
t
o
th
e
fact
that
the
re
was
no
la
r
ge
data
var
ia
ti
ons
b
ut
in
the
F
ig
ure
4
(a
)
we
ca
n
s
ee
an
osc
il
la
tio
n
a
nd
an
ove
rsho
oting
of
th
e
reconstr
ucte
d
data.
The
F
i
gure
4
(
c)
we
ha
ve
the
resu
lt
of
the
r
econs
tructe
d
tr
ajecto
ry
w
her
e
we
can
cl
ea
rly
see
that
con
s
trai
ne
d
cub
ic
sp
li
ne
ga
ve
t
he
bette
r
resu
lt
s
a
nd
t
he
os
ci
ll
at
ion
of
the
c
ub
ic
sp
i
ne
inter
po
la
ti
on
af
fected
t
he
obta
ined
trajecto
ry.
F
or
the F
ig
ure
5,
w
e h
a
ve use
d
the
foll
ow
i
ng d
at
a
:
(a)
:
x=[0,
1,2,3,
4,5
,
6,7,8,
9,10
]
; y
=[0
,
0,0,0,
0,1,1,
1,1,1,
1]
(b): x
=
[
0,1,2,
3,4,5,
6,7,8,
9,10,
11,12,1
3,14
]
;
y=
[0
,
0,0,0,
0,1,1,
0,0,0,
-
1,0
,0,0,0]
I
n
Fig
ur
e
5,
it
is
note
d
t
hat
the
co
ns
trai
ne
d
c
ub
ic
s
pline
does
no
t
os
ci
ll
at
e
after
la
rge
am
pli
tud
e
var
ia
ti
on
an
d
r
edu
ce
s
the
ove
rsho
oting
c
ompari
ng
to
the
c
ub
ic
sp
li
ne.
Th
is
fact
m
akes
t
he
co
ns
trai
ne
d
cub
ic
sp
li
ne
m
or
e
st
able
an
d
h
as
l
ess
os
ci
ll
at
ion
s
wh
ic
h
m
akes
it
bette
r
f
or
t
he
reconstr
uctio
n
of
t
he
tra
j
ect
or
ie
s
.
The
inter
pola
ti
on
pro
gr
am
s
wer
e
m
ade
with
Ma
tl
ab
2017
a
on
a
com
pu
t
er
with
the
f
ol
lowing
co
nfi
gurati
on:
I7
3770
3.4
G
hz
a
nd
8
G
b
R
a
m
.
Table
2
shows
inter
pola
ti
on
exec
uti
on
ti
m
e
accor
ding
to
inter
pola
te
d
po
i
nt
nu
m
ber
.
O
ne
c
an
noti
ce
two
beh
a
viou
rs
.
T
he
first
is
fo
r
num
ber
of
point
s
below
200:
the
co
ns
trai
ne
d
cub
ic
sp
li
ne
is
faster
than
the
cu
bic
sp
li
ne
since
it
does
not
s
olve
a
syst
em
of
equ
at
io
ns.
T
he
seco
nd
case
is
w
he
n
the
num
ber
of
interp
olate
d
po
i
nts
excee
ds
200
t
he
c
ub
i
c
sp
li
ne
overc
om
es
the
con
s
trai
ned
c
ubic
sp
li
ne
.
In
t
he
F
ig
ur
e
6
one
ca
n
s
ee
a
hand
wr
it
ing
tra
j
ect
ory
reconstr
ucted
with
the
c
on
strai
ned
c
ubic
sp
li
ne
interp
olati
on.
Figure
7
sho
w
s
a
sim
ulati
on
of
P
um
a
60
0
r
obot
e
xec
uting
the
sam
e
traje
ct
or
y.
The
ad
va
ntag
e
of
the
sim
ulatio
n
process
is
r
econst
ru
ct
e
d
pa
th
visu
al
iz
at
ion.
The
pro
posed
syst
em
dep
en
ds
on
the
t
eacher
,
and do
not ge
ne
rate coll
isi
on
-
fr
ee t
raj
ect
or
ie
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Tra
je
ct
or
y
rec
on
str
uctio
n
fo
r
robo
t
p
r
ogrammi
ng b
y
d
e
m
on
str
atio
n
(
Re
da Ha
nifi
Elhachemi
Am
ar
)
3071
Figure
4
.
Tra
j
e
ct
or
y rec
onstr
uc
ti
on
us
in
g
c
ub
ic
sp
li
ne
a
nd c
on
st
raine
d
c
ub
i
c sp
li
ne
inter
pola
ti
on
s
Figure
5
.
I
nter
po
la
ti
on
of
di
fferent
data set
s
Table
2
.
Param
et
ers
of t
he
P
um
a 6
00 m
anip
ulator
robot
Nu
m
b
e
r
o
f
po
in
ts
Co
n
strain
ed
sp
lin
e
(
m
s
)
Cu
b
ic sp
lin
e (
m
s)
25
1
.38
3
2
.33
2
50
1
.90
0
2
.48
1
100
2
.50
2
3
.12
1
200
4
.08
4
3
.39
0
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, N
o.
3
,
J
une
2020
:
30
66
-
3073
3072
Figure
6
.
3D T
raj
ect
or
y
recon
structed
b
y t
he c
onstrai
ne
d
c
ubic
s
pline
Figure
7
.
Sim
ulati
on
of
rob
ot
trajec
to
ry r
e
pr
oductio
n
6.
CONCL
US
I
O
N
This
pap
e
r
pr
esented
a
rob
ot
pr
ogram
m
i
ng
by
dem
onstrat
ion
syst
em
in
a
tra
j
e
ct
or
y
le
vel.
The
syst
em
i
s
base
d
on
hu
m
an
m
ov
em
ent
reprod
uc
ti
on
by
f
ollow
i
ng
a
sin
gl
e
po
i
nt
of
i
nterest
(the
hand/
to
ol).
We
use
d
ARToolkit
as
a
visu
al
trac
king
syst
em
.
The
slo
w
fr
a
m
e
rate
and
t
he
m
iss
identific
at
ions
of
the
m
ark
ers
was
creati
ng
ga
pes
in
the
tra
je
ct
or
ie
s.
T
he
na
tural
cu
bic
spl
ine
and
co
ns
tr
ai
ned
cub
ic
s
pline
i
nter
po
la
ti
ons
wer
e
us
e
d
to
correct
these
weaknesse
s
and
t
o
rec
ons
truct
the
tr
a
j
e
ct
or
y.
A
c
om
par
ison
was
m
ade
be
tween
t
hose
bo
t
h
inte
rpolat
ion
m
et
ho
ds
f
or
3D
tra
j
ect
ori
es
rec
onstr
uc
ti
on,
1D
data
set
s
and
as
per
e
xe
cution
ti
m
e.
T
est
s
sh
ow
that
con
strai
ne
d
cub
ic
pli
ne
interp
olati
on
ov
e
r
com
es
the
cu
bic
s
plin
e
inter
po
la
ti
on.
The
co
ns
trai
ne
d
c
ub
ic
sp
li
ne
inter
po
la
ti
on
s
hows
good
res
ults
an
d
t
he
fa
ct
that
it
gen
erates
le
ss
os
ci
ll
at
ion
and
pr
e
ve
nts
ov
e
rs
hootin
g
m
akes
it
su
it
a
ble
for
the
tra
j
ect
ory
reconst
ru
ct
io
n.
In the
fu
t
ur
e
works t
his it wil
l be
us
e
d
i
n
a
re
al
tim
e trajecto
ry r
ec
onst
ru
ct
i
on.
REFERE
NCE
S
[1]
Acosta
-
calde
ron
,
C.
A.,
and
H.
Hu,
"
Robot
imi
ta
ti
on
:
Bod
y
sc
hema
and
bod
y
per
c
ept
,
"
Appli
ed
Bi
on
ic
s
and
Bi
omec
han
ic
s
,
v
ol.
2
,
no
.
3
,
pp
.
1
31
-
148,
2005
.
[2]
Acosta
-
calde
ron
,
C.
A.,
and
H.
Hu,
"
Robot
imi
ta
ti
on
from
hum
an
body
m
ovements
,
"
The
3
rd
Inte
rnational
Symposium on
I
mitat
ion in Anim
als an
d
Art
if
ac
ts
,
pp
.
1
-
9
,
UK
,
20
05.
[3]
Le
i
,
J.,
Song,
M.,
Li,
Z.
N.
and
C
hen,
C.
,
"
W
hole
-
bod
y
hum
anoi
d
robot
imita
ti
on
with
pose
sim
il
ari
t
y
eva
lu
ation,
"
Signal
Proce
ss
in
g
,
v
ol
.
108
,
pp
.
1
36
-
146,
2015
.
[4]
Benc
hikh
,
L.,
“
Method
for
tra
i
ning
a
robot
or
the
li
k
e,
and
de
vic
e
for
implem
ent
ing
said
m
ethod
,
”
U.S.
Patent
Applic
a
ti
on
No
12/812
792,
201
1.
[5]
Biggs,
G.,
and M
ac
Donald,
B
.
,
"
A survey
of
rob
ot
progra
m
m
ing
s
y
stems
,
"
Proceedi
ngs of
Australasian
Confe
ren
ce
on
Robotics
and
Aut
omation
,
Bri
sbane
,
Aus
tr
al
i
a
,
pp.
1
-
10
2
,
003.
[6]
Kuni
y
oshi,
M.
Ina
ba
,
and
H
.
Inoue
,
"
Lear
ning
b
y
wat
ch
ing:
Ex
tra
c
ti
ng
reu
sable
ta
sk
knowledge
fr
om
visual
observa
tion
of
hum
an
per
form
anc
e
,
"
IE
EE
transact
ions
on
robotic
s
a
nd
automati
on
,
v
ol
10,
no
.
2
,
pp.
799
-
822
,
19
94.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Tra
je
ct
or
y
rec
on
str
uctio
n
fo
r
robo
t
p
r
ogrammi
ng b
y
d
e
m
on
str
atio
n
(
Re
da Ha
nifi
Elhachemi
Am
ar
)
3073
[7]
M.
Cal
inon,
S.
,
Bil
la
rd
,
A.,
"
A
proba
bil
ist
ic
pr
ogra
m
m
ing
by
demons
tra
ti
on
fr
amework
handl
i
ng
constra
int
s
i
n
joi
nt
spa
ce
and
ta
sk
spa
ce
,
"
In
te
rnational
Con
fe
renc
e
on
In
te
l
li
gent
Robot
s
a
nd
Syste
ms
IRO
S
,
Nic
e,
Franc
e
,
pp.
367
-
372
,
20
08.
[8]
Aleot
ti,
J.
,
Cas
el
li,
S.
,
and
Re
ggia
ni
,
M.
,
"
Levera
ging
on
a
virt
ual
env
iron
m
ent
for
robot
progra
m
m
ing
b
y
demons
tra
ti
on,
"
Robot
ic
s
and
Au
tonomous Sy
ste
ms
,
vol
.
47,
pp.
153
-
161,
2004
.
[9]
Herre
ro,
Héc
tor
,
e
t
a
l
.
,
"S
kil
l
base
d
robot
pro
gra
m
m
ing:
As
s
embl
y
,
visio
n
a
nd
W
orkspa
ce
Monitori
ng
skil
l
int
er
ac
t
ion,
"
N
eu
rocomputing
,
pp
.
61
-
70
,
2017
.
[10]
Shim
iz
u,
M.
,
Y
oon,
W
.
,
Kita
g
a
ki,
K.
,
"
Exp
eri
m
ent
a
l
va
li
da
ti
on
of
ta
sk
skil
l
tra
n
sfer
appr
oa
ch
us
ing
a
hum
anoi
d
robot
,
"
In
te
rnati
onal
Symposium
on
Assembl
y
an
d
Manufacturing
ISA
M'07
,
pp
.
14
1
-
146,
2007
.
[11]
Kurie
n,
M.
,
Ki
m
,
M.
K.,
Kops
i
da,
M.
,
and
Br
il
aki
s
,
I.,
"
Re
al
-
ti
m
e
sim
ula
t
io
n
of
constru
ct
io
n
workers
usin
g
combined
hum
a
n
bod
y
and
h
an
d
tracki
ng
for
r
oboti
c
construct
ion
worker
s
y
s
t
em
,
"
Aut
omatio
n
in
Constr
uctio
n
,
vol.
86
,
pp
.
125
-
1
37,
2018
.
[12]
Vale
nt
ini
,
P.
P.,
"
Natur
al
interf
ac
e
for
in
te
ra
ctive
vir
tua
l
asse
m
bl
y
in
augme
nte
d
re
alit
y
usi
ng
Le
a
p
Motio
n
Control
le
r
,
"
Int
ernati
onal
Jour
nal
on
Inte
racti
v
e
Design
and
Manufac
turin
g
(
IJI
DeM
)
,
v
ol.
22,
no.
4,
pp.
1
-
9
,
2018
.
[13]
Khair
udin,
M
.
,
e
t
al
.
,
"Control
of
a
m
ovabl
e
robo
t
hea
d
using
v
isi
on
-
base
d
object
tra
ck
ing
,
"
Int
ernati
onal
Journal
of
E
le
c
tric
al
and
Computer
Eng
i
nee
ring (
IJE
C
E)
,
vol
.
9
,
no
.
4
,
pp
.
2503
-
2512
,
20
19.
[14]
Shi,
Qing,
e
t
a
l
.
,
"D
esign
and
implementatio
n
of
an
om
nidi
recti
ona
l
vis
ion
s
y
s
te
m
for
ro
bot
per
ce
pt
ion,
"
Me
chat
roni
cs
,
v
ol.
41
,
pp
.
58
-
66
,
2017
.
[15]
Cal
i
non
,
S
y
lv
ai
n
,
and
Aude
Bi
ll
a
rd,
"S
toc
hastic
g
esture
produc
ti
o
n
and
rec
ogn
it
i
o
n
m
odel
for
a
h
um
anoi
d
robot,
"
IEE
E
/R
SJ
Int
ernati
onal
Con
fe
ren
ce
on
Int
el
l
ige
nt
Robot
s and
S
yst
ems (
IROS
)
,
v
ol. 3
,
pp.
2769
-
277
4,
2004
.
[16]
Aleot
ti,
J. and
C
ase
ll
i
,
S.
,
"
Robust t
rajec
tor
y
lear
ning
and
appr
ox
i
m
at
ion
for
robot programm
ing
b
y
demons
tra
ti
on
,
"
Robot
ic
s
and
Au
tonomous Sy
ste
ms
,
vol
.
54,
no.
5,
pp
.
409
-
413
,
2006.
[17]
La
ver
y
,
J.E
.
,
"
Univar
i
at
e
cubic
Lp
spline
s
a
nd
shape
-
pre
servin
g,
m
ult
iscale
interpolation
b
y
un
iva
ri
at
e
cubic
L
1
spline
s
,
"
Compu
te
r A
ide
d
Geome
tric
Design
, v
ol
.
17,
no
.
4
,
pp
.
31
9
-
336,
2000
.
[18]
La
ver
y
,
J.E
.
,
"
Shape
-
pre
serv
ing,
first
-
der
iv
at
iv
e
-
base
d
par
ametr
i
c
and
nonp
aram
et
ric
cub
ic
L
1
spline
cu
rve
s
,
"
Computer
Ai
d
ed
Geomet
ric
Desi
gn
, v
ol
.
23
,
no
.
3
,
pp
.
276
-
296
,
2
006
.
[19]
Hernoux,
F.
,
B
e
are
e
,
R
.
,
Gajn
y
,
L.
,
e
t
al
.
,
"
Leap
Motion
pour
la
c
apt
ure
de
m
ouvement
3D
par
spli
ne
L1
,
"
Journ
ées
du
Gr
oupe
de
Tr
ava
i
l
en
Modé
li
s
ati
on
Géomé
triq
ue
,
Ma
rseille, Fr
anc
e
,
2013
.
[20]
Khan,
D.,
Ul
la
h
,
S.,
R
abbi
,
I.,
"
Fact
ors
aff
e
ct
i
ng
the
design
a
nd
tra
ck
ing
of
ARToolKit
m
ar
ker
s
,
"
Computer
Standards
&
Int
erfac
es
,
v
ol
.
41
,
pp.
56
-
66
,
2015
.
[21]
Rabbi
,
Ihsan
,
an
d
Sehat
Ull
ah
,
"
Ext
endi
ng
the
tr
ac
king
d
ista
nc
e
of
fiduc
i
al
m
ark
ers
for
la
rge
ind
oor
augme
nte
d
rea
l
ity
appl
i
catio
ns,"
Adv
an
ce
s
in
Elec
tri
cal and Com
pute
r E
ngin
ee
ring
,
v
ol
.
15
,
no.
2
,
pp
.
59
-
64
,
2015.
[22]
C.
J.
Kruger,
"
C
onstrai
ned
cubic
spline i
nt
erp
ol
ation,
"
Chemi
cal
Engi
ne
ering
App
li
cations
,
2003
.
[23]
Kokes,
J.
and
N
ghie
n,
N.B
.
,
"U
sing
const
rai
n
ed
cubi
c
splin
e
inst
ea
d
of
nat
ura
l
c
ubic
spline
to
e
lim
ina
te
over
shoo
t
and
under
sh
oot
i
n
Hilbe
rt
Huang
Tra
nsform
,
"
The
13th
Int
ernati
on
al
Carpathian
C
ontrol
Confe
ren
ce
,
High
T
at
r
as,
Slovaki
a
,
pp
.
30
0
-
306
,
2012
.
[24]
Ouam
ri,
B.
,
an
d
Ahm
ed
-
Foiti
h
Z.
,
"
Adapti
v
e
neur
o
-
fuz
z
y
infe
ren
ce
s
y
st
e
m
base
d
cont
rol
of
pum
a
6
00
robot
m
ani
pulat
or
,
"
Inte
rnat
ion
al
Journal
of
El
e
ct
rica
l
and
Computer
Enginee
ring
(
IJE
CE
)
,
vol
.
2,
no
.
1,
pp.
90
-
97
,
201
1
.
[25]
Ouam
ri,
B.
,
and
Ahm
ed
-
Foiti
h
Z.
,
"
Com
pute
d
Torque
Contro
l
of
a
Pum
a
600
Rob
ot
b
y
usi
ng
Fuzz
y
Logic
,
"
Inte
rnational
R
e
vi
ew
of Aut
omat
i
c
Control
,
vo
l
.
4
,
no
.
2
,
pp
.
248
-
252,
2011
.
Evaluation Warning : The document was created with Spire.PDF for Python.