Int
ern
at
i
onal
Journ
al of
R
obot
ic
s
and
Autom
ati
on (I
JRA)
Vo
l.
8
,
N
o.
4
,
D
ece
m
ber
201
9
,
pp.
277
~
292
IS
S
N:
20
89
-
4856
,
DOI: 10
.11
591/
i
jra
.
v
8
i
4
.
pp
277
-
292
277
Journ
al h
om
e
page
:
http
://i
ae
score.
c
om/j
ourn
als/i
ndex.
ph
p/IJRA
Envir
onment
m
ap
pin
g
,
m
ap
c
onstructin
g
,
and
p
ath
p
lanni
ng
for
u
nd
erw
ater
n
av
i
gation
of
a
l
ow
-
cost
µAUV
in
a
c
lutte
red
nu
clear storag
e p
ond
Yibin Pe
ng
,
P
eter
N
Gree
n
School
of El
ec
tr
i
ca
l
and
Elec
t
roni
cs
Engi
n
ee
rin
g
,
t
he
Univer
si
t
y
of
Manc
heste
r
,
UK
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
pr
26
,
201
9
Re
vised
A
ug
27
,
2019
Accepte
d
Oct
6
,
201
9
Thi
s
pap
er
pre
sents
a
nove
l
appr
oac
h
th
at
ena
bl
es
a
low
-
cost
µA
UV
(m
ic
ro
au
tonomous
under
wat
er
vehicl
e)
n
avi
g
at
e
and
work
s
afe
l
y
in
an
enc
losed
c
lut
t
er
e
d
und
erwa
t
er
e
nvironment.
In
orde
r
to
ac
h
ie
ve
a
au
tonom
y
col
li
sion
-
f
ree
na
viga
ti
on
in
unde
rwate
r
,
i
t
r
equi
r
es
a
reliab
l
e
ap
proa
ch
that
would
a
ll
ow
a
µA
UV
which
e
quippe
d
with
sparse
and
inacc
ura
te
sonar
sensors
to
m
ap
the
und
erwa
t
er
envi
ronm
ent
,
a
m
ap
const
ruc
ti
n
g
al
gor
it
hm
tha
t
conve
r
ts
the
obta
in
ed
d
at
a
to
useful
information
to
esta
bl
ish
a
m
ap
,
and a
pat
h
pl
anni
ng
a
lgori
thm
tha
t
p
l
ans
a
collision
-
fre
e
pa
th
from
the
star
t
to
the
d
esire
d
go
al.
The
proposed
a
pproa
ch
wi
ll
integra
t
e
env
ironment
surve
y
,
envi
ron
m
ent
re
c
onstruct
ion
,
and
pat
h
pl
anni
ng
as
a
comple
te
d
ta
sk
,
which
invol
ves
m
ap
a
cqui
siti
on
and
pat
h
pla
nn
ing.
A
complete
sim
ula
ti
on
o
f
the
envi
ronm
en
t
topo
log
y
ac
q
uisit
ion
and
r
oute
p
la
nn
ing
proc
ess
is
documente
d
in t
his pa
per
.
Ke
yw
or
d
s
:
A* p
at
h plan
nin
g
al
gorithm
Aer
ia
l m
app
in
g gr
i
d
m
ap
reconstr
uctio
n and i
nter
pola
ti
on
Au
t
onom
ou
s
unde
r
water
veh
ic
le
Ma
p gr
id
ding
Po
int cl
oud
m
ap
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Yib
in
Pe
ng
,
Sch
oo
l
of Elec
tric
al
an
d El
ect
ronics E
nginee
rin
g
,
The U
niv
er
sit
y o
f
Ma
nc
hester
,
Oxfor
d
Rd
,
Ma
nch
e
ste
r
M
13
9P
L
,
G
reat M
a
nch
e
ste
r
,
UK
.
Em
a
il
:
yib
in.peng@m
anch
est
er.ac.
uk
1.
INTROD
U
CTION
Investi
gatio
ns
on
the
a
utonom
ou
s
ve
hicle
hav
e
been
m
ade
sig
nificant
pro
gr
e
ss
ov
e
r
the
past
few
ye
ars.
Ba
se
d
on
it
s
w
orki
ng
env
i
ronm
ent
,
auto
no
m
ou
s
ve
hicle
s
can
be
c
at
egorised
i
nto
three
ty
pes
,
t
he
one
work
i
ng
i
n
t
he
ai
r
is
cal
le
d
unm
ann
ed
ae
rial
ve
hicle
s
(
UAVs
)
s
uc
h
as
dron
e
s
,
w
hi
ch
is
us
e
d
to
film
,
env
i
ronm
ental
s
urveil
la
nce
,
g
eo
grap
hy
st
udy
,
a
nd
eve
n
par
cel
delive
ry
.
T
he
one
w
or
king
i
n
the
gr
ound
is
cal
le
d
un
m
anned
gro
und
vehi
cl
e
(UG
V)
s
uc
h
as
sel
f
-
dr
i
vi
ng
car
s
,
w
hic
h
is
us
e
d
t
o
r
eplace
hum
an
pilot.
More
ov
e
r
,
t
he
on
e
w
orkin
g
unde
r
water
is
ca
ll
ed
auto
no
m
ous
un
derwate
r
veh
ic
le
(
A
UV)
.
This
pap
e
r
int
erests
at
A
U
V u
nd
e
r
water a
ppli
cat
i
on.
A
di
ver
sit
y
of
AUVs
has
al
r
eady
bee
n
devel
op
e
d
in
dif
fe
ren
t
pur
po
ses
and
de
plo
ye
d
in
dif
fer
e
nt
unde
rw
at
er
sit
uations
,
w
hich
inclu
ding
in
dustria
l
process
es
,
m
arine
res
earch
,
a
nd
ai
r
cras
h
in
vestig
at
io
ns
.
Th
os
e
in
dustri
al
processes
in
cl
ud
e
but
are
not
lim
it
ed
to
unde
rsea
oil
pi
pe
li
ne
le
aka
ge
de
te
ct
ion
[
1]
a
nd
s
pe
nt
nu
cl
ea
r
f
uel
st
or
a
ge
pond
m
on
it
ori
ng
[2]
.
The
ocea
n
res
earches
inclu
de
unde
rsea
oil
ex
plo
it
at
ion
,
m
arine
stud
y
an
d
in
form
ation
c
ollec
ti
on
,
o
cea
n
sea
rch
a
nd
r
escue
[
3]
.
As
f
or
ai
r
c
ras
h
i
nv
e
sti
gations
,
A
U
Vs
ha
ve
been u
sed
to find w
rec
kag
es
of m
issi
ng
aer
op
la
nes
,
su
c
h
a
s Mal
ay
sia
A
ir
li
nes
Fli
ght 3
70 acci
de
nt
[4]
.
This
work
is
m
ot
ivate
d
by
t
he
nu
cl
ea
r
sto
r
age
pond
de
co
m
m
iss
ion
in
g
pro
blem
.
Stor
ag
e
ponds
a
re
us
e
d
t
o
kee
p
sp
e
nt
fu
el
ass
e
m
blies
an
d
ot
her
ra
dio
act
iv
e
m
at
erial
unde
rw
at
er
unti
l
they
a
re
suffici
ently
ben
i
gn
to
be
s
ubj
ect
ed
to
processin
g
f
or
lo
ng
te
rm
safe
st
or
a
ge.
I
n
te
rm
s
of
siz
e
,
Olym
pic
swi
m
m
ing
pools
pro
vid
e
a
c
onve
nient b
e
nch
m
ark
,
al
th
ough
s
tora
ge
ponds
a
re
ty
pical
ly
m
u
ch
dee
per
.
W
a
ste
obj
ect
s
are
p
la
ce
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
277
–
292
278
on
t
he
bott
om
of
the
pond
wi
th
a
si
gn
ific
a
nt
de
pt
h
of
cl
ea
r
wate
r
ab
ove
f
or
s
hieldin
g
pu
rposes.
T
he
m
at
erial
rem
ai
ns
within
a
pond
for
a
num
ber
of
ye
a
rs
.
T
he
s
pe
nt
nucl
ear
f
uel
st
or
a
g
e
pond
c
onta
ins
highly
ra
dioa
ct
ive
m
et
al
s
su
ch
as
U
ra
niu
m
an
d
Th
or
ium
[
5
]
,
wh
ic
h
are
haz
ardo
us
to
hum
an
healt
h.
Plac
es
s
uc
h
as
t
he
se
a
re
near
ly
im
po
ssible
for
hu
m
ans
to
wor
k
in
e
ve
n
wea
rin
g
a
protect
ive
s
uit.
I
n
su
c
h
ci
rc
um
s
ta
nces
,
A
U
Vs
cou
l
d
po
te
ntial
ly
b
e
us
e
d
t
o
re
place
hum
ans
t
o
m
on
it
or
for
le
akin
g
co
ntainer
s
,
to
fin
d
the
disposit
ion
of
m
at
e
rial
i
n
old
e
r pon
ds
,
a
nd to
explo
re t
he
unkn
own u
nder
water e
nvir
on
m
ent.
Ba
sed
on
t
he
e
ra
of
c
onstructi
on
,
e
xisti
ng
st
orage
ponds
ca
n
be
cl
assifi
e
d
a
s
le
gacy
po
nd
s
or
m
od
e
rn
ponds.
The
la
tt
er
te
nd
to
be
equ
i
pp
e
d
with
regular
sto
rage
rac
ks
desig
ne
d
to
hold
can
ist
ers
co
ntaini
ng
the
rad
i
oacti
ve
m
a
te
rial
.
He
nce
t
he
distrib
ution
of
cl
utter
(the
so
li
d
obj
ect
s
w
it
hin
the
pond
)
is
re
gu
la
r
or
a
t
le
ast
well
-
orga
nise
d
[
2
]
.
Fi
gure
1
is
an
e
xam
ple
of
a
m
odern
pond.
Le
gacy
ponds
a
r
e
a
diff
e
ren
t
m
at
te
r.
Or
i
gin
at
in
g
f
rom
tim
es
wh
en
the
haza
rds
of
storing
nu
cl
ea
r
m
at
erial
wer
e
no
t
as
cl
ea
rly
unde
rstood
a
s
they
are
now
,
an
d
ha
nd
li
ng
te
ch
no
log
y
was
le
ss
well
dev
el
op
e
d
,
le
gacy
ponds
con
ta
i
n
m
at
erial
that
was
or
i
gin
al
ly
store
d
in
a
ha
ph
aza
r
d
m
anner
in
s
kips
(
du
m
ps
te
rs)
.
S
ome
m
a
te
rial
in
le
gacy
po
nd
s
ha
s
bee
n
im
m
ersed
for
m
any ye
ars
an
d has
par
ti
al
ly
d
isi
nteg
rated
[
2
]
. F
i
gure
2
is
an
e
xam
ple o
f t
he
le
gacy
pond.
Figure
1. Mo
de
rn nucle
ar
s
to
rag
e
po
nd [6]
Figure
2. Lega
cy
n
uclea
r
sto
r
age
pond
[7
]
It
is
e
vid
e
ntly
desi
rab
le
that
co
ndit
ion
s
in
both
cl
asses
of
po
nds
be
m
on
it
or
ed
,
al
thou
gh
this
is
a
pote
ntial
ly
hazardo
us
e
nd
e
avou
r.
In
t
he
case
of
m
od
er
n
ponds
,
it
is
help
fu
l
to
che
ck
f
or
l
eaka
ge
from
canist
ers
s
o
th
at
rem
edial
work
can
be
car
ri
ed
ou
t
in
a
ti
m
el
y
m
ann
er.
Fo
r
le
gacy
po
nd
s
,
it
is
neces
sary
to
determ
ine
the
disp
e
rsal
of
ob
j
ect
s
withi
n
a
pond
an
d
the
di
stribu
ti
ons
of
rad
i
oacti
vity
,
tem
per
at
ure
an
d
ot
her
par
am
et
ers
thr
oughout t
he
po
nd
,
s
o
t
hat p
la
ns f
or
m
at
erial
r
e
m
ov
al
and l
ong
-
te
rm
stor
ag
e can be
f
or
m
ulate
d.
The
resea
rch
r
eported
in
this
pap
e
r
buil
ds
on
w
ork
on
sm
a
ll
-
scal
e
,
low
-
c
os
t
AUVs
(
µ
A
UV
s
)
that
has
been
unde
rw
ay
in
Ma
nc
hester
f
or
se
ver
al
ye
ars
[
8
,
9
]
,
res
ul
ti
ng
i
n
the
A
V
EX
IS
ve
hicle
[
8
,
9
].
AV
E
X
IS
ha
s
4
degrees
of
fr
e
edo
m
.
Sup
port
ing
c
omm
un
ic
at
ion
a
nd
local
isa
ti
on
syst
em
s
are
cu
rr
e
ntly
unde
r
de
velo
pm
ent.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Enviro
nme
nt
m
apping
m
ap constructi
ng
,
an
d p
ath
plan
ning
for... (Yi
bin Pe
ng
)
279
On
e
ai
m
of
th
e
A
VE
XIS
w
ork
is
t
o
de
velo
p
a
s
war
m
(‘s
ho
al
’)
of
A
U
V
s
a
ble
to
e
xp
l
ore
,
m
ap
an
d
m
on
it
or
stora
ge
ponds.
Su
c
h
a
syst
em
requires
m
app
ing
a
nd
path
pl
ann
i
ng
ca
pab
il
it
ie
s
and
this
pa
per
disc
us
ses
init
ia
l
effor
ts t
o pro
vi
de
this
k
i
nd of
functi
onal
it
y.
The
pro
blem
of
the
µ
AUV
na
vig
at
io
n
i
n
a
stora
ge
pond
i
s
ho
w
t
o
e
nab
l
es
a
µ
A
UV
t
ra
ver
se
f
r
om
the
sta
rt
posit
ion
t
o
the
g
oa
l
po
sit
io
n
(t
he
destinat
io
n
w
her
e
t
he
r
obot
hea
ding
to
)
wh
il
e
a
vo
i
ding
any
ob
sta
cl
es
in
it
s
env
i
ronm
ent
without
the
help
of
GP
S
(G
P
S
w
orks
ba
dly
in
unde
r
water)
[
10
].
[
11
-
13
]
introd
uce
se
ve
ral
na
vig
at
io
n
strat
e
gies
f
or
m
ob
il
e
r
obots
.
3D
path
plan
ning
in
a
cl
uttered
pond
is
a
par
ti
cula
rly
interest
in
g
c
halle
ng
e
si
nce
it
r
el
ie
s
on
3D
m
a
ps
w
hose
acc
ur
acy
m
ay
no
t
be
high
,
a
nd
c
olli
sion
s
betwee
n
r
obots
an
d
cl
utter
m
ay
ha
ve
se
rio
us
conseq
ue
nces
f
or
the
r
obot
,
th
e
en
vir
onm
ent
,
bysta
nde
rs
,
or
al
l
three.
M
or
e
ove
r
,
t
he
possi
ble
appr
oach
es
are
co
ns
trai
ne
d
by
siz
e
an
d
co
st
issues.
Cl
early
,
the
µ
A
UVs
ne
ed
to
be
sm
all
en
ough
t
o
na
vig
at
e
t
hro
ugh
co
nfi
ne
d
sp
ace
s
,
an
d
a
ke
y
obj
ect
ive
of
this
wor
k
ha
s
been
t
o
kee
p
costs
low
t
o
e
nab
le
the syste
m
to
be
u
se
d
i
n
a
var
i
et
y of
wat
er
-
ba
sed processe
s.
Howe
ver
,
so
fa
r
,
m
os
t
appro
a
ches
s
uch
a
s
F
orward
-
Lo
okin
g
S
on
a
r
[
14]
,
Mult
i
-
Be
a
m
So
na
r
[
15]
has
been
c
onduct
e
d
in
e
xp
e
ns
i
ve
big
-
siz
ed
AUV
for
m
arine
res
earch
.
This p
a
per
p
r
opose
s
a
n
ap
proac
h
to ad
dress
the
pond
m
appi
ng
an
d
pat
h
pl
ann
i
n
g
pro
blem
fo
r
a
se
nsor
-
lim
it
ed
low
-
c
ost
µA
UV
,
w
hich
c
onsist
s
of
3
ste
ps
.
Step
1
,
a
lo
w
-
cost
AUV
wit
h
s
pa
rse
a
nd
i
naccurat
e
s
ona
r
se
nsors
to
m
ap
t
he
unde
rwat
er
e
nv
ir
onm
e
nt
a
nd
gathe
r
the
dept
h
m
easur
em
e
nt
data.
Ste
p
2
,
a
m
ap
co
nst
ru
ct
io
n
that
c
onve
rts
the
ob
t
ai
ned
data
t
o
us
ef
ul
inf
or
m
at
ion
to
est
ablish
a
n
oc
cup
a
ncy
gr
id
.
Step
3
,
A*
pa
th
-
plan
ning
a
lgorit
hm
that
plans
a
c
olli
sion
-
f
ree
path fr
om
the s
ta
rt to
the
d
e
sir
ed go
al
.
The
e
nv
i
ronm
ent
m
app
ing
s
trat
egy
is
bas
ed
on
t
he
ide
a
of
a
n
‘aeria
l
su
r
vey’.
I
n
this
c
onte
xt
,
an
aerial
s
urve
y
is
si
m
ply
a
set
of
de
pth
m
easur
em
ents
ta
ke
n
on
a
re
gu
la
r
m
easur
e
m
ent
gr
i
d
at
a
fi
xed
dep
t
h
within
the
cl
ear
water
re
gion
of
the
po
nds.
See
Fi
g
ure
3
.
The
res
ult
of
a
sur
vey
is
a
m
at
rix
of
tu
ples
,
w
hos
e
com
po
ne
nts
a
r
e
m
easur
em
en
t
po
sit
ion
a
nd
corres
pondin
g
dep
t
h.
This
da
ta
is
the
2D
pro
j
ect
io
n
of
th
e
3D
cl
utter
on
to
th
e
m
easur
em
ent
plane
a
nd
ca
n
be
us
e
d
to
c
on
struct
a
hei
gh
t
m
ap
of
t
he
ob
je
ct
s
at
the
bo
tt
om
of
the
pond.
T
hi
s
in
it
sel
f
pro
vid
es
us
e
f
ul
da
ta
ab
ou
t
t
he
disposit
ion
of
m
a
te
rial
within
a
po
nd.
Howev
e
r
,
a
sig
nifica
nt
a
m
ou
nt
of
inf
or
m
at
ion
is
l
os
t
and
heig
ht
m
aps
do
not
pro
vid
e
ade
quat
e
su
pp
or
t
f
or
th
e
pat
h
plan
ning
nee
de
d for
detai
le
d
e
xp
l
or
at
io
n.
What
is reall
y nee
ded is a
3D r
e
presentat
io
n of t
he pon
d
cl
utter
.
Fig
ure
3.
A
com
p
l
e
te
aerial
m
app
in
g
a
nd pat
h
pla
nn
i
ng in
a clutt
ere
d
e
nc
losure
In
this
pa
per
,
the
ad
van
ta
ge
is
ta
ken
of
the
fact
that
m
os
t
cl
utter
in
stora
ge
po
nd
s
is
pri
sm
a
ti
c
in
sh
a
pe
i.e.
obj
e
ct
s
hav
e
a
pol
ygonal
base
a
nd
a
cr
os
s
-
sec
ti
on
that
do
e
s
no
t
var
y
with
heig
ht
(
c
uboi
d
an
d
cy
li
nd
rical
co
nt
ai
ner
s
that
ar
e
ty
pical
ly
us
ed
in
m
os
t
ponds
)
.
T
he
a
ppr
oa
ch
ou
tl
ine
d
i
n
this
pa
per
locat
es
cl
utter
f
r
om
a
heig
htm
ap
and
us
e
s
t
his
in
f
orm
at
ion
to
ge
ne
rate
3D
po
i
nt
c
loud
m
ap
of
pri
s
m
at
ic
ob
j
ect
s
f
ro
m
wh
ic
h
a
f
ull 3
D
g
ri
d
m
ap
of
the po
nd can
be co
ns
tr
ucted
t
hat is s
uitable
for
A
*
path
p
la
nn
i
ng alg
or
it
hm
s.
In
orde
r
to
ens
ur
e
t
hat
the
A
*
path
plan
ning
al
gorit
hm
wo
r
ks
with
th
e
ob
ta
ine
d
3D
gri
d
m
ap
,
both
processes
m
us
t
be
perf
or
m
e
d
in
the
sam
e
coord
i
nate
s
yst
e
m
and
the
dep
t
h
m
easur
em
ent
at
ref
le
ct
ion
po
sit
io
ns
m
us
t
li
e
on
the
use
r
-
s
pecified
s
urvey
gr
i
d.
The
survey
gri
d
i
s
the
l
ocati
on
of
m
esh
poin
ts
i
n
the
m
easur
em
e
nt
plane
w
her
e
de
pth
m
easur
e
m
ents
are
ta
ke
n
(as
i
nd
ic
at
e
d
in
Fig
ur
e
1)
a
nd
the
A
*
m
eth
od
is
app
li
ed
to
a
w
orks
pace
base
d
on
a
discreti
z
ed
re
pr
ese
ntati
o
n
of
the
e
nv
ir
on
m
ent’s
ge
om
et
ry
.
In
this
syst
e
m
,
the
A
*
sea
rc
hin
g
gri
d
is
a
3D
gr
id
ba
sed
on
the
2D
s
urvey
gri
d
but
ha
s
the
sam
e
interval
in
the
z
-
axis.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
277
–
292
280
Be
cause
the
re
is
of
te
n
unli
kely
that
the
sa
m
pled
locat
io
ns
wil
l
coincide
with
the
s
urve
y
gr
id
,
su
c
h
as
the
case
that
incli
ne
d
obj
ect
s
cau
se
th
e
de
pth
m
easur
em
ent
at
po
si
ti
on
s
do
no
t
li
e
on
t
he
s
urve
y
gr
i
d.
The
refor
e
,
it
requires
f
urt
he
r
data
processi
ng
to
m
ake
the
m
con
sist
ency.
T
his
pap
e
r
e
xpla
ins
how
to
est
ablish
a
3D
dig
it
al
gr
i
d
m
ap
base
d o
n raw
de
pth
m
easur
em
ent
da
ta
an
d p
r
ov
i
de
s so
m
e si
m
ulati
on
resu
lt
s.
2.
A
ERI
AL
MAPPIN
G
The
te
rm
of
“
Aer
ia
l
m
app
in
g”
or
“aeria
l
s
urvey”
ori
gin
a
l
ly
co
m
e
fr
om
an
ai
rbo
rn
e
l
and
s
urvey
,
wh
ic
h
is
a
n
ef
f
ic
ie
nt
m
et
ho
d
of
m
app
in
g
ge
ogra
ph
ic
featu
r
e
of
la
rg
e
areas
.
It
is
al
so
kn
own
a
s
Ec
ho
-
s
oundin
g
so
na
r
in
the
under
sea
im
age
and
i
nvest
igati
on
resea
rc
h.
Si
de
-
scan
us
es
sonar
de
vices
tha
t
produce
a
c
onic
al
or
fan
-
s
hap
e
d
beam
do
w
n
to
the
seaflo
or
with
a
wi
de
be
a
m
that
per
pe
nd
ic
ular
to
t
he
path
of
t
he
sens
or
thr
ough
the
wa
te
r
to
gat
her
th
e
ge
ogra
ph
ic
inf
or
m
at
ion
of
the
sea
be
d
[
16
].
Ech
o
-
sou
nd
i
ng
us
es
s
on
a
r
de
vices
that
pro
du
ce
a
con
ic
al
beam
do
w
nwar
d
to
t
he
pond
bo
tt
om
with
a
wide
be
a
m
ang
le
pe
r
pe
nd
ic
ular
to
th
e
path
of the se
nsor
th
rou
gh the
wate
r
to
m
easur
e th
e d
e
pth
of the
bo
tt
om
[17]
.
This
m
app
i
ng
al
gorithm
is
sim
pl
y
a
set
of
de
pth
m
easur
em
ents
ta
ken
on
a
re
gula
r
m
easur
em
ent
gr
i
d
at
a
fixed
de
pth
withi
n
the
cl
ear
water
re
gio
n
of
the
pond
s.
I
n
pract
ic
e
,
µAUV
will
proj
ect
a
ver
ti
cal
beam
from
a
pl
ane
with
a
s
pecifie
d
heig
ht
t
ow
a
r
d
a
n
area
by
hi
gh
-
f
reque
ncy
acoust
ic
pulse
e
m
issi
on
in
order
to
captu
re
de
pth
i
nfor
m
at
ion
ab
ou
t
the
c
orres
pondin
g
bo
tt
om
su
r
face.
T
he
m
ov
e
m
ent
plane
is
set
ab
ov
e
cl
utters
and
sam
pled
in
a
unifo
rm
ly
sp
ace
d
2d
gri
d
(see
Fig
ur
e
1)
.
Th
e
µA
U
V
will
m
ov
e
to
e
ach
gr
i
d
a
nd
m
easur
e
the
de
pth
of
t
he
current
gr
i
d
by
the
recei
ve
d
pulse
that
re
f
le
ct
ed
off
f
ro
m
the
obj
ect
’s
to
p
surfa
ce.
A
r
outi
ne
of
t
his
s
urvey
is
show
n
i
n
Fi
g
ure
4
.
The
res
ul
t
of
a
n
aerial
s
urvey
is
an
ar
r
ay
of
t
up
le
s
,
wh
ose
com
pone
nts
are
m
easur
em
ent posit
ion
s
and its
corres
pondin
g heig
ht. See
Fi
g
ure
5
for
t
he
i
nput a
nd outp
ut
o
f
this m
et
hod
.
Fig
ure
4.
Sca
n i
ti
ner
ary
Figure
5
.
Dia
gra
m
o
f
i
nput a
nd
ou
t
pu
t
The
Ec
ho
-
s
ound
i
ng
is
ca
rr
ie
d
out
by
ac
ousti
c
sensors
w
hi
ch
rely
on
m
od
ulati
ng
s
urface
ac
ousti
c
wav
e
s
to
se
nse
a
ph
ysi
cal
obj
ect
.
Th
ese
dev
ic
es
tra
ns
duce
el
ect
rical
sign
al
s
to
ac
ousti
c
wa
ves
,
r
ecei
ve
ref
le
ct
ed
wa
ve
s
,
an
d
c
onver
t
receive
d
wa
ve
s
bac
k
to
el
e
ct
rical
sign
al
[
18
]
.
D
ur
in
g
th
e
wav
e
pro
pa
ga
ti
on
,
acoust
ic
wa
ve
s
will
be
ref
l
ect
ed
,
refract
e
d
,
or
at
te
nuat
ed
by
th
e
m
e
diu
m
.
For
t
he
de
pth
m
easurem
ent
,
the
r
eflect
ion
i
s
the
m
ai
n
co
ncern.
I
n
te
rm
s
of
analy
sis
,
the
f
orm
of
a
wav
e
in
-
dep
t
h
m
easur
em
ent
is
not
conve
nient
,
bu
t
the
cha
racteri
sti
c
of
t
he
ac
ousti
c
wa
ve
c
a
n
be
a
ppr
oxim
a
t
ed
to
a
ray
an
d
sim
ply
neg
le
ct
ed
it
s
wav
e
nat
ur
e
si
nce
the
wa
veleng
t
h
of
the
ty
pical
aco
us
ti
c
wav
e
is
sm
al
ler
com
par
e
d
to
the
siz
e
of
ph
ysi
cal
obj
ect
s und
e
r
t
he
water [
19
]
.
A
s
a
co
ns
e
que
nce
,
t
he
ray
tra
ci
ng
alg
or
it
hm
is
intr
oduce
d
t
o
m
od
el
a
nd
si
m
ula
t
e
the
s
ound
pro
pa
gation [
20
]
.
T
he
pro
pa
gation
of
the
aco
us
ti
c
wa
ve
is re
gard
ed
a
s r
ay
s
,
an
d
the
c
on
ic
al
b
e
a
m
is
const
it
uted
of
a
finite
num
ber
of
rays
.
The
dep
t
h
ca
n
be
c
al
culat
ed
by
knowin
g
th
e
ti
m
e
interval
be
tween
the tra
ns
m
itted sig
nal and
rec
ei
ved
sig
nal and the
pr
op
a
gation sp
eed
of the
aco
us
ti
c ray.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Enviro
nme
nt
m
apping
m
ap constructi
ng
,
an
d p
ath
plan
ning
for... (Yi
bin Pe
ng
)
281
The
m
ai
n
pr
oble
m
is
find
in
g
the
c
orrect
sign
al
(
ray)
f
r
om
the
receive
d
sig
nal
sp
ect
r
um
.
Be
cause
there
are
unw
anted
sig
nals
and
backg
rou
nd
noise
s
i
n
the
recei
ved
si
gnal
sp
ect
ru
m
.
T
he
bac
kgr
ound
noise
include
s
ref
le
c
ti
on
s
from
sm
a
ll
obj
ect
s
,
re
fle
ct
ion
s
of
si
delo
bes
or
par
t
of
r
everberati
on
(e
sp
eci
al
ly
in
s
ha
ll
ow
waters
)
,
the
unwa
nted
si
gn
a
l
includes
la
te
r
ref
le
ct
ions
(
a
sign
al
that
ha
s
m
ulti
ple
tim
e
s
of
ref
le
ct
io
ns)
an
d
diffuse
re
flect
ion
s
(
a
si
g
nal
t
hat
scat
te
rs
in
al
l
directi
on
w
hen
hits
an
obje
ct
).
T
he
refor
e
,
the
use
f
ul
si
gnal
is
the
no
rm
al
ref
le
ct
ion
of
t
he
em
it
te
d
ray
,
because
a
rec
ei
ver
a
nd
a
tr
ansm
itter
com
bin
e
d
in
one
dev
ic
e
(transce
iver
)
,
so
t
he
on
ly
first
-
order
s
pe
cular
ref
le
ct
io
n
ca
n
be
rec
ei
ved
is
the
norm
al
ref
le
ct
ion
.
The
or
et
ic
al
ly
,
it
is ref
le
ct
ed
of
f
the nea
rest
obsta
cl
es an
d
ha
s the h
ig
hest re
m
ai
nin
g
ene
r
gy
[
18
]
.
Ba
se
d o
n
this
pro
per
ty
,
2
ke
y
factor
s
ca
n
be
us
e
d
to
est
i
m
at
e
the
nor
m
al
ref
le
ct
ion
:
tim
e
of
flig
ht
(TOF)
a
nd
i
m
pu
lse
re
spo
ns
e
e
nerg
y
le
vel.
In
t
he
i
m
pu
lse
res
pons
e
in
t
he
tim
e
do
m
ai
n
,
the
norm
al
ref
le
ct
ion
sig
nal
is
the
pu
lse
with
t
he
hi
gh
e
st
ene
r
gy
re
spon
s
e
a
nd
t
he
s
hortest
T
OF.
On
ce
t
he
TOF
of
the
desire
d
sig
nal
is
known
,
t
he
dep
t
h
is
giv
e
n by the e
quat
io
n:
ℎ
=
∗
2
(1)
w
he
re
v
is t
he spee
d of
sou
nd in wat
er a
nd t
is t
he
ti
m
e o
f
fi
gh
t
of the
norm
al
r
eflect
ion
sign
al
.
Th
e
f
l
owcha
rt
of the im
ple
m
e
nted ray
-
traci
ng
base
d
aerial
m
app
in
g
M
AT
LAB
sim
ulator
is sh
own
in Fi
gure
6.
Figure
6
.
Ra
y t
racin
g
fl
ow cha
rt
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
277
–
292
282
A
ray
-
tr
aci
ng
si
m
ulator
is
bu
il
t
on
MATL
AB.
A
ss
um
e
the
po
nd
is
a
50
m
*25m
swi
m
m
ing
pool
-
siz
e
d
e
nc
losure
with
a
dep
t
h
of
10m
.
Assum
e
the
side
-
sca
n
sam
pling
res
olu
ti
on
is
0.5m
*0
.5m
,
the
ob
ta
ine
d
de
pth
m
easur
em
ent
data
an
d
it
s’
co
rr
es
pondin
g
surf
ace
pl
ot
are
s
ho
wn
i
n
Fig
ure
7
a
nd
8
resp
ect
ively
.
Figure
7. Ge
om
et
ry
m
od
el
of a s
t
or
a
ge p
ond
Figure
8.
Ra
w
dep
t
h
m
easur
e
m
ent p
oi
nts
obta
ined by aeria
l
m
app
in
g
3.
OCCUP
A
NCY
G
RI
D MA
P
C
O
NS
T
RUC
TING
A
ND P
ATH PL
A
NNING
The
ra
w
data
obta
ined
by
the
aerial
m
app
ing
is
a
to
po
l
og
ic
al
de
pth
m
easur
em
ent
data.
It
is
require
d
to
be
processe
d
with
bo
unda
ry
traci
ng
(t
o
e
xtract
obj
ect
s
’
bounda
ry
po
i
nt
s
,
est
im
at
e
ob
je
ct
s’
occupied
reg
i
on
and
ve
rtic
es
,
a
nd
cal
culat
e
th
e
top
pla
ne
e
quat
ion
)
,
Pla
nar
dep
t
h
mea
s
ur
e
men
t
data
re
ge
ner
atin
g
(to
ge
ner
at
e
dep
t
h
m
easur
e
m
ent
points
th
at
li
e
on
t
he
s
urvey
gri
d
base
d
on
plane
e
quat
ion
e
stim
at
ed
by
extra
ct
ed
ve
rtic
es
po
i
nt)
,
3D
stru
ct
ur
e
co
ns
tr
ucting
(to
c
onstruc
t
the
rest
bo
dy
by
interp
olati
ng
po
i
nts
un
de
r
the
ass
um
ption
of
the
ca
nister
sha
pe)
,
a
nd
gr
id
meth
od
(to
init
ia
li
ze
the
e
nv
ir
on
m
ent’s
point
cl
oud
m
at
rix
by
bin
a
ry
var
ia
bl
e)
in
orders
to
est
ab
li
sh
the
3D
digi
ta
l
gr
id
m
ap
f
or
A
*
path
pla
nn
i
ng
al
gorith
m
.
The
flo
w
c
har
t
of
the
pro
cessi
ng
proce
dure is s
how
n
i
n
Fi
gure
9.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Enviro
nme
nt
m
apping
m
ap constructi
ng
,
an
d p
ath
plan
ning
for... (Yi
bin Pe
ng
)
283
Figure
9. The
proces
si
ng f
lo
w
c
har
t
3.1.
Bo
und
ar
y
t
racin
g
Boun
dar
y
traci
ng
al
gorithm
is
a
te
ch
nique
that
is
a
pp
li
e
d
t
o
dig
it
al
im
ag
es
in
orde
r
t
o
extract
thei
r
bounda
ry
[21].
A
di
gital
im
ag
e
is
a
gro
up
of
pix
el
s
on
a
s
quare
te
ssell
at
ion
each
hav
i
ng
a
certai
n
value
.
Sinc
e
poi
nt
cl
oud
da
ta
has
sim
i
la
r
featur
e
s
so
bo
unda
ry
traci
ng
al
gorithm
can
be
us
ed
he
re.
Ther
e
are
4
c
om
m
on
ways
to
trace
t
he
bo
unda
ry.
Th
ey
are
Squa
re
Traci
ng
Algorithm
,
Moore
-
Nei
ghbor
T
ra
ci
ng
,
Ra
dial
S
weep
,
and
T
he
o
Pavli
dis'
Algorithm
[22]
se
parat
el
y.
T
hey
t
r
acki
ng
bo
unda
ry
in
a
diff
e
re
nt
m
anner.
This
pap
e
r
us
es
the
ra
dial
swe
ep
al
gorithm
t
o
extr
act
boun
dar
y
points
fro
m
po
int
cl
oud
data
in
ord
er
t
o
fin
d
ver
ti
ces
po
i
nts
because
this al
gorithm
is e
asy
to
im
ple
m
ent.
The b
oundary t
racin
g
incl
ud
e
s
2
ste
ps
:
‐
Extract
boun
da
ry points.
‐
Rem
ov
al
point
s of the
trace
d object.
Ther
e
is
a
dis
ti
ng
uis
he
d
hei
gh
t
differe
nce
in
the
to
polo
gical
point
cl
oud
data
,
wh
i
ch
se
gm
ents
gro
und
point
s
et
s
an
d
ob
j
ect
s
’
point set
s.
He
nce
,
he
ig
ht
c
he
ckin
g
will
p
r
oc
eed
i
n
seq
ue
nc
e
in
o
r
der
to
f
ind
the
first
ob
j
ect
point
w
hich
is
cl
assifi
ed
as
the
or
i
gin
of
the
boun
dar
y.
Af
te
r
the
first
ob
j
ec
t
po
int
is
f
ound
,
the
bounda
ry traci
ng
proce
dure i
s invok
e
d.
Thi
s u
ses the c
onc
ept of ch
ai
n
co
de
[
23
,
24]
,
whose
ch
ecki
ng
e
ncode
r
m
ov
es
al
ong
the
bo
unda
ry
of
t
h
e
reg
i
on
to
e
xtract
bounda
ry
points
(in
a
cl
oc
kwise
or
antic
lo
ckw
is
e
directi
on). T
his conti
nues
un
t
il
the ch
ec
king
encode
r
ret
urn
s to
t
he
sta
rtin
g p
os
it
ion.
In
orde
r
t
o
fin
d
the
nex
t
bounda
ry
po
i
nt
,
th
e
heig
ht
of
8
di
recti
on
s
point
s
that
s
urrou
nd
ing
t
he
st
a
rt
po
i
nt
will
be
c
heck
e
d
i
n
cl
oc
kw
ise
orde
r
f
r
om
the
so
ut
h
di
recti
on
unti
l
find
i
ng
the
first
nex
t
boun
dar
y
po
i
nt.
Diag
ram
of
t
he
8
directi
ons
re
sp
ect
t
o
t
he
sta
rt
po
i
nt
a
nd
c
he
ck
directi
on
is
sho
wn
in
Fi
gure
10
(i
,
j)
is
use
d
t
o
represe
nt 8 d
ir
ect
ion
s
of a
dja
cent points
r
el
a
ti
ve
to th
e start
point.
Figure
10.
8 Di
recti
on
s
res
pec
t t
o
the c
urre
nt
po
i
nt
On
ce
t
he
ne
xt
bounda
ry
point
has
bee
n
f
ound
,
it
is
save
d
and
reg
a
r
ded
as
the
ne
w
sta
rt
po
i
nt
an
d
a
si
m
il
ar
procedure
to
th
e
a
bo
ve
is
rep
eat
e
d
,
with
one
im
po
rta
nt
diff
e
re
nc
e.
T
he
diff
e
re
nce
is
rathe
r
t
ha
n
sta
rt
check
i
ng
at
t
he
sout
h
directi
on
it
c
omm
ences
f
ro
m
the
dir
ect
ion
t
hat
t
he
pr
e
vious
sta
rt
po
i
nt
with
res
pe
ct
t
o
the curre
nt sta
r
t po
i
nt. A c
omplet
e boun
dar
y
traci
ng is
dem
on
st
rated i
n
Fi
gure
11.
As
s
how
n
in
F
igure
11
,
red
dots
re
pr
ese
nt
the
f
ound
boun
dar
y
po
i
nts
an
d
gree
n
dots
r
epr
ese
nt
th
e
nex
t
s
uccessi
ve
boun
dar
y
points.
T
he
sam
ple
data
is
scan
ned
at
blac
k
dot
f
ro
m
bo
tt
om
to
top
searc
hing
f
o
r
the
sta
rt
point.
A
fter
fin
ding
the
sta
rtin
g
poi
nt
,
sta
rt
the
ini
ti
al
bounda
ry
c
heck
in
the
cl
oc
kw
ise
di
recti
on
in
the
s
outh
di
rec
ti
on
of
t
he
sta
rt
ing
point
un
ti
l
the
nex
t
bounda
ry
po
i
nt
is
f
ound.
A
fter
t
hat
,
reg
a
rd
i
ng
t
he
f
ound
nex
t
boun
dar
y
po
i
nt
as
the
ne
w
sta
rt
,
a
nd
r
epeat
the
boun
dar
y
c
heck
at
the
s
ou
t
h
-
west
directi
on
of
th
e
ne
w
sta
rting
po
i
nt. The
bounda
ry check
i
ng
will
term
inate
w
hen the initi
al
starti
ng
po
i
nt (
the
f
irst red
dot) is
fou
nd.
Figure
11
de
m
on
strat
es
an
exam
ple
of
t
he
boun
dar
y
t
racin
g
process
on
a
sam
ple
point
set
.
T
he
m
os
t
i
m
po
rtant
t
hing
in
t
h
e
bounda
ry
traci
ng
pr
oc
edure
is
the
"
sense
of
direct
ion
".
The
rig
ht
directi
ons
a
re
with
resp
ect
t
o
t
he
pr
e
vious
posit
ion
i
ng.
T
he
refo
re
,
it
'
s
i
m
po
rta
nt
to
kee
p
t
rac
king
of
the
cu
rr
e
nt
an
d
prev
iou
s
or
ie
ntati
on
i
n o
rd
e
r
to
m
ake the
r
ig
ht m
ov
es.
Le
gend
H
:
Hei
gh
t
NN
:
N
or
th
NE
:
N
or
th
-
E
ast
NW
:
N
or
th
-
Wes
t
WW
:
Wes
t
EE
:
Eas
t
SW
:
Sout
h
-
We
st
SS
:
S
outh
SE
:
S
outh
-
West
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
277
–
292
284
Figure
11.
B
ound
a
ry T
racin
g
In
pract
ic
e
,
the
re
will
be
m
ore
than
one
obje
ct
in
the
e
nv
i
ronm
ent
so
the
obta
ined
poin
t
cl
oud
will
con
ta
in
a
di
ff
e
ren
t
set
of
poin
ts
(one
set
of
points
c
orres
pond
s
to
on
e
ob
je
ct
).
H
oweve
r
,
the
pr
ob
le
m
is
it
will
rep
eat
e
dly
trac
e
the
fi
rst
ob
j
e
ct
an
d
ca
nnot
br
ea
k
out
t
he
l
oop
.
I
n
or
der
to
esc
ape
f
ro
m
the
lo
op
,
it
is
r
equ
i
red
to
rem
ov
e
eac
h
trace
d
ob
j
ec
t
fr
om
the
po
i
nt
cl
oud
s
o
th
at
it
do
es
not
pr
e
ve
nt
tracki
ng
of
t
he
rem
ai
nin
g
obj
ect
s.
On
ce
t
he
rem
ov
al
pro
cess
has
bee
n
done
,
restart
t
o
f
i
nd
the
nex
t
obj
ect
poi
nt
a
nd
trace
the
s
ucc
essive
bounda
ry
po
i
nts point a
gain
. See
the e
xam
ple in F
i
gure
12.
Figure
12. Re
m
ov
al
o
f
a
trac
ed object
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Enviro
nme
nt
m
apping
m
ap constructi
ng
,
an
d p
ath
plan
ning
for... (Yi
bin Pe
ng
)
285
Ther
e
are
2
se
ts
of
po
i
nts
in
the
point
cl
oud
,
so
the
re
a
re
2
ob
j
ect
s.
T
he
re
d
po
i
nts
a
r
e
boun
dar
y
po
i
n
ts
,
t
he
blue
points
are
re
m
ind
er
obj
e
ct
po
i
nts
a
nd
the
heig
hts
of
the
black
points
w
her
e
sca
nn
i
ng
sta
rts.
Figure
9
dem
on
strat
es
how
t
o
e
xtract
bo
undar
y
po
i
nts
w
hen
there
are
m
ul
ti
ple
obj
ec
ts
in
the
po
i
nt
cl
oud.
Howe
ver
,
it
a
ppare
ntly
le
ft
on
e
pr
ob
le
m
:
When
2
ob
j
ect
s
ov
erlap
or
t
ou
c
h
each
oth
e
r
,
the
point
cl
oud
of
their
tou
c
hed
e
dge
will
be
m
erg
e
d
t
oget
her.
As
a
res
ult
,
t
his
al
gorithm
will
treat
them
as
one
obj
ect
.
F
or
this
par
ti
cula
r
case
,
segm
entat
ion
of point cl
oud wil
l be
dep
l
oy
ed
to
classi
fy t
his
2 obje
ct
.
3.2.
En
vironm
ent
rec
onstr
uctio
n by p
oin
t
cl
oud
Usu
al
ly
,
t
he
ra
w
de
pth
m
easu
rem
ent
po
i
nts
do
no
t
c
oinc
id
e
with
the
use
r
’s
s
pecifie
d
s
urvey
gr
id
,
s
o
it
is
necessary
to
cal
culat
e
the
plane
e
quat
io
n
to
c
reate
ne
w
m
easur
em
ent
po
ints
t
hat
co
incide
wit
h
the
us
er
’s
sp
eci
fied
s
urve
y gr
i
d.
A
plane
in
3D
sp
a
ce
is
de
fin
ed
by
t
hr
ee
points
,
wh
ic
h
do
no
t
al
l
li
e
on
the
s
am
e
li
ne
or
by
a
poi
nt
and
a
norm
al
vecto
r
t
o
the
pl
ane
,
w
hich
is
cal
le
d
the
vect
or
e
qu
at
io
n
of
a
pla
ne.
In
othe
r
w
ords
,
in
order
to
cal
culat
e
the
v
ect
or
e
quat
io
n
of
a
plane
,
it
is
re
qu
i
red
to
kn
ow
the
co
ordi
na
te
of
at
le
ast
3
points
t
hat
lo
cat
ed
in
the
pla
ne
an
d
not
in
the
sa
m
e
strai
gh
t
li
ne.
H
ow
e
ver
,
th
e
sel
ect
ion
of
po
i
nts
nee
ds
to
be
care
fu
l
,
be
cause
it
cannot
be
ass
um
ed
that any
3 arbitra
ry point
s taken f
ro
m
the p
la
ne
are
a
bl
e to r
e
present t
he plane.
Th
e
c
ho
s
en
po
i
nts
s
hould
be
a
ble
to
best
descr
i
be
t
he
plane
par
am
et
er.
Ve
rtic
es
point
s
are
t
he
best
opti
on
to
cal
cul
at
e
the
plane.
Since
th
e
bo
unda
ry
points
i
n
t
he
point
cl
ou
d
a
re
ext
r
act
ed
by
boun
dar
y
traci
ng
,
ve
rtic
es
ca
n
be
fou
nd
by
com
par
in
g
their
x
value
a
nd
y
va
lue.
Once
the
co
ordin
at
e
of
3
ve
rtic
esp
oin
ts
is
known
,
the
vector
pla
ne
equ
at
io
n
c
an
be cal
culat
ed. T
he gene
ral f
or
m
o
f
the
vecto
r
equati
on of a
plane
is:
⃗
∙
(
−
1
)
=
0
(2)
⃗
is
the
norm
al
vecto
r
of
the
pl
ane.
T
he
nor
m
al
vector
⃗
=
(a
,
b
,
c)
c
an
be
ca
lc
ulate
d
by
Cr
os
s
Pr
oduct
of
any
2
vect
or
s
t
hat
c
onsti
tuted
by
the
known
3
ver
ti
ces.
P1=
(x
1
,
y
1
,
z
1
)
is
one
of
the
known
ve
rtic
es
points.
P=(x
,
y
,
z
)
is
any
po
i
nt
th
at
locat
ed
on
t
his
pla
ne.
T
his
e
quat
ion
ca
n
be
t
ran
s
f
or
m
ed
int
o
t
he
ge
ner
al
plana
r
equ
at
io
n
,
an
d t
he heig
ht of a
ny
p
oi
nt on
t
he plane is
g
i
ven
by:
[
,
,
]
∙
[
−
1
,
−
1
,
−
1
]
=
0
+
+
−
(
1
+
1
+
1
)
=
0
=
(
1
+
1
+
1
)
−
−
(3)
Fo
r
t
he
case
th
at
hei
gh
t
m
eas
ur
em
ents
at
re
f
le
ct
ion
posit
io
ns
that
do
not
li
e
on
the
s
urve
y
gri
d
s
uch
as ti
lt
ed
obj
ect
s
,
this
vecto
r plane e
qu
at
io
n (
3
)
ca
n be
us
e
d t
o
cal
culat
e the
h
ei
ght
of
t
he p
oin
ts
who
se
po
sit
ion
li
e
on
t
he
s
urv
ey
gri
d
an
d
the
ob
j
ect
’s
to
p
pl
ane
(
x
a
nd
y
va
lue
is
the
co
ordi
nate
of
the
c
orres
pondin
g
s
urve
y
gr
i
d)
.
T
hus
,
ac
qu
i
red
a
ne
w
point
cl
oud
that
denotes
the
co
r
respo
nd
i
ng
obje
ct
’
to
p
pla
ne
by
po
i
nts
t
hat
l
ie
on
the s
urvey g
rid
. S
ince
the
re a
r
e num
ber
s of
obj
ect
s
,
the
abo
ve pr
ocesses
wi
ll
b
e und
e
rtak
en fo
r
eac
h obj
ect
.
Af
te
r
re
gen
e
rati
ng
the
ne
w
m
easur
em
ent
po
i
nts
of
the
obj
e
ct
’
uppe
r
pla
ne
,
the
ne
xt
ste
p
is
to
furthe
r
process
it
to
create
a
com
plete
d
3D
m
odel
of
the
e
nviro
nm
ent.
This
proce
dure
is
cal
le
d
en
vironm
ent
reconstr
uctio
n.
As
ob
j
ect
s
ar
e
assum
ed
to
be
pr
ism
at
ic
,
so
thei
r
side
s
urfaces
are
pe
rpend
ic
ular
t
o
it
s
uppe
r
su
r
face.
This
a
cru
ci
al
ass
umpti
on
f
or
reconstr
ucting
the
obj
ect
becau
se
the
key
idea
of
reconstr
ucti
ng
t
he
env
i
ro
nm
ent
by
Po
int
Cl
ou
d
is
to
ad
d
ne
w
po
i
nts
to
t
he
m
od
ifie
d
m
eas
ur
em
ent
points
in
the
pe
rp
e
ndic
ular
directi
on o
f
t
he
uppe
r
plane
t
o
create
the r
em
ai
nin
g
pa
rt
bet
ween
the
to
p
pl
ane
de
pth
m
ea
su
rem
ent
data o
f
t
he
ob
sta
cl
e
a
nd
t
he
gr
ound
m
e
asur
em
ent
data
.
A
dd
e
d
po
i
nts
m
us
t
sat
isfy
the
co
ndit
ion
t
hat
li
es
on
t
he
us
er
’s
sp
eci
fied
s
urve
y
gr
i
d.
Th
us
,
it
ge
ner
at
es
the
3D
point
cl
oud
m
od
el
of
obst
acl
es.
T
his
te
c
hn
i
qu
e
is
known
as
li
near
heig
ht
in
te
rpolat
ion
[25
]
.
The
sc
hem
atic
diag
ram
of
the
perpe
nd
ic
ular
he
ig
ht
inter
p
olati
on
is
s
ho
wn
i
n
Figure
13.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
277
–
292
286
Figure
13. A s
chem
at
ic
o
f
th
e li
near
pe
rp
e
ndic
ular
h
ei
gh
t i
nter
po
la
ti
on
An
exam
ple
of
us
i
ng
li
near
pe
rp
e
ndic
ular
he
igh
t
i
nterpola
ti
on
to
gen
e
rate
a
poi
nt
cl
ou
d
of
the
3D
m
od
el
o
f dete
c
te
d
ob
j
ect
s is s
how
n belo
w.
It
i
s explai
ne
d w
it
h
the ai
d of
F
igure
14.
Figure
14. A s
chem
at
ic
o
f
th
e li
near
pe
rp
e
ndic
ular
h
ei
gh
t i
nter
po
la
ti
on
Assum
e
the
reso
luti
on
of
the
su
r
vey
gri
d
i
s
α
*α
,
an
d
the
po
sit
io
n
of
m
easur
em
ent
res
ult
li
es
on
the
s
urvey
gri
d.
Po
i
nt
A
,
B
,
C
,
D
are
4
v
erti
ces
points
of
t
he
m
easur
em
ent
po
i
nt
cl
ou
d.
Assum
e
the
co
ordinate
of
A
is
(x
a
,
y
a
,
z
a
)
,
B
is
(x
b
,
y
b
,
z
b
)
,
C
is
(x
c
,
y
c
,
z
c
)
,
D
is
(x
d
,
y
d
,
z
d
)
res
pecti
vely
.
It
ne
eds
to
cal
culat
e
the
x
,
y
c
oor
din
at
e of
a
dded
p
oi
nts e
,
f
,
g
a
nd
m
ake
s
ur
e
they
a
re
ly
ing
on
the
s
urvey
gri
d.
P
oint
e
,
f
,
g
exist
on
t
he
vecto
r
that
pe
r
pendicula
r
to
t
he
pla
ne
a
nd
pa
ss
thr
ough
B
,
so
it
nee
ds
to
c
al
culat
e
the
eq
uation
of the
vecto
r.
The
e
quat
ion o
f
the
v
ect
or
an
d
the
coo
rd
i
nate of the
point i
n
the
v
ect
or is
giv
e
n by:
⃗
⃗
⃗
⃗
⃗
=
[
−
,
−
,
−
]
⃗
⃗
⃗
⃗
⃗
=
[
−
,
−
,
−
]
⃗
=
⃗
⃗
⃗
⃗
⃗
×
⃗
⃗
⃗
⃗
⃗
=
[
,
,
]
(4)
=
⃗
+
∙
⃗
=
[
,
,
]
+
∙
[
,
,
]
(5
)
[
,
,
]
=
[
+
∙
,
+
∙
,
+
∙
]
(6)
=
+
∙
(7)
=
+
∙
(8
)
=
+
∙
(9
)
Gr
oup poin
t se
t
P
oin
ts in
de
t
ec
t
ed
objec
t's
t
op pla
ne
Gr
oup poin
t se
t
P
oin
ts in
de
t
ec
t
ed
objec
t's
t
op pla
ne
P
oin
ts ad
ded
b
y
perpe
ndicu
lar h
eig
h
t
in
t
erpola
tion
A
B
C
D
x
z
y
e
f
g
Gr
oun
d
Remain
in
g
p
art
Evaluation Warning : The document was created with Spire.PDF for Python.