TELKOM
NIKA
, Vol. 11, No. 5, May 2013, pp. 2402 ~
2408
ISSN: 2302-4
046
2402
Re
cei
v
ed
Jan
uary 13, 201
3
;
Revi
sed Ma
rch 6, 2
013;
Acce
pted Ma
rch 1
5
, 2013
Disaster Rescue Simulation based on Complex
Adaptive Theory
Feng Jiang*
1,
2
, Liangtu Song
1, 2
, Yida
Fan
3
,Wen
bo Li
2
and Jie Zhang
1, 2
1
Universit
y
of S
c
ienc
e an
d T
e
chno
log
y
of Chi
na, Dep
a
rtmen
t
of Auto
mation, Hefei, China
2
Institute of Intelli
ge
nt Machi
n
es, Chin
ese Ac
adem
y of Scie
nces, Hefei, C
h
in
a
3
Nationa
l Disa
s
ter Reducti
on
Center, Ch
ina
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
jfjf@mail.u
s
tc.edu.cn
A
b
st
r
a
ct
Disaster resc
u
e
is on
e of th
e key
meas
ur
es of
disast
er reductio
n
. T
h
e rescue
proc
ess is a
compl
e
x proc
e
ss w
i
th the ch
aracteri
stics
of large scale,
complic
ate str
u
cture, no
n-l
i
n
ear. It is har
d
to
descri
be a
nd
ana
ly
z
e
th
em
w
i
th tradition
al
meth
ods. Ba
sed o
n
co
mp
l
e
x ad
aptiv
e th
eory, this p
a
p
e
r
ana
ly
z
e
s th
e compl
e
x ad
apta
t
ion of the r
e
s
c
ue pr
ocess
fr
om s
e
ve
n feat
ures: ag
greg
at
ion, n
onl
in
earit
y,
mo
bil
i
ty, divers
ity, taggin
g
, in
ternal
mo
de
l a
nd b
u
il
din
g
bl
ock. W
i
th the supp
ort of Re
past pl
atform,
a
n
age
nt-bas
ed mode
l
i
n
clu
d
i
ng rescue ag
ents and
victi
m
age
nts w
a
s pro
p
o
s
ed. Mor
eover,
tw
o simulati
on
s
w
i
th different p
a
ra
meters
are
empl
oyed
to e
x
amin
e t
he fe
a
s
ibil
ity of the
mo
de
l.
As a r
e
sult, the pr
op
o
s
e
d
mo
de
l h
a
s
be
e
n
sh
ow
n that
it
is effici
ent i
n
d
eali
n
g
w
i
th the
disaster
rescu
e
si
mul
a
tio
n
and
can
pr
ovid
e th
e
referenc
e for makin
g
decis
io
n
s
.
Ke
y
w
ords
:
di
saster rescue,
compl
e
x ada
pti
v
e theory, ag
e
n
t
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Chin
a is one
of the countri
es whi
c
h a
r
e
most a
ffected
by natural disa
sters. Acco
rding to
the
International Dis
a
s
t
er Database s
t
atis
tic
s
,
th
ere
are
24 time
s and
17 tim
e
s of
disaste
r
s
,
respe
c
tively, occurre
d
in
Chin
a amo
n
g
the t
op 1
0
0
natural di
sasters
whi
c
h
cau
s
e
d
larg
es
t
numbe
r of d
eaths
and di
rect e
c
ono
mic losse
s
in
th
e 20th centu
r
y. In rece
nt years, alth
ou
gh
disaster reducti
on
capacit
y
is st
rengthening, natural di
sasters are still the major i
s
sues
which
impact econo
mic con
s
tructi
on
a
nd so
cial
develo
p
ment
. The
situatio
n is extrem
el
y severe in
fa
ce
of natural
disasters. Th
e p
o
sitive
an
d ef
fective re
scu
e
is a
n
impo
rt
ant mean
s to
redu
ce
disaster
los
s
e
s
.
Di
sa
st
er
re
sc
ue i
n
v
o
lv
es
so
cial f
a
ct
or
s
su
ch a
s
gove
r
nment, so
cia
l
orga
nizatio
n
s,
v
i
ct
ims,
t
i
me
f
a
ct
or
s s
u
c
h
as o
c
cu
rr
enc
e t
i
me
, disa
ster
dura
t
ion, spatial
factors such
as
occurre
n
ce pl
ace, weathe
r con
d
ition
s
, and also re
sou
r
ce fa
ctors su
ch a
s
ba
sic rescue fa
cilitie
s,
relief
sup
p
lie
s. Th
ese
fa
ctors form
a
h
e
terog
enei
ty, differen
c
e,
d
i
versity, su
dd
en, ra
ndo
mn
ess
system. The
r
efore, disaste
r
re
scue
sy
stem is a comp
lex adaptive system.
The theo
ry of complex ad
aptive syste
m
has
be
en applie
d in variou
s fields
such a
s
manufa
c
turi
n
g
, medical scien
c
e, en
gin
eerin
g, et
c. a
nd one
of the importa
nt area
s lie
s in
the
multi-ag
ent system (MAS) simulatio
n
. Qinglin G
uo
and Ming Z
hang [1] prop
ose
d
archite
c
ture
con
s
i
s
ts of v
a
riou
s
autono
mous
ag
ents that
are ca
pa
ble
of comm
u
n
icatin
g with
each othe
r a
n
d
makin
g
d
e
ci
sion
s b
a
sed
on thei
r
kn
owle
dge to
form a
archit
ecture of m
u
lti-age
nt-b
ased
Intelligent M
anufa
c
turin
g
System. Arend Li
gtenbe
rg et al. [2] explore
d
the
use
of MAS to
simulate
spa
t
ial sce
nari
o
s ba
sed o
n
modelin
g mu
lti-acto
r de
ci
sion
-ma
k
in
g within a spa
t
ial
planni
ng pro
c
e
ss. Bonni
e
Ruben
stein
Montano et
al. [3] presented a Bayesia
n
learnin
g
approa
ch
for a m
u
lti-ag
en
t system
whi
c
h l
e
a
r
ned
to
identify a
n
approp
riate
a
gent to
an
swer
free-text que
ries an
d key
w
ord sea
r
che
s
for defen
se
contra
cting. Kevin F.R. Liu [4] introduced
an
agent-ba
s
ed
environ
menta
l
eme
r
ge
ncy
manag
eme
n
t
frame
w
o
r
k a
s
a
lo
osely couple
d
colle
ction
of agents th
at can coope
rate to achi
e
v
e a
commo
n goal prepa
redn
ess for
and re
sp
on
se to
environ
menta
l
eme
r
gen
cy
situation
s
. K
a
iYi
ng
Che
n
and
Chu
n
Jay Ch
en [5] u
s
ed multi
-
age
nt
techn
o
logy to
con
s
tru
c
t a multi-sectio
n flexib
le manu
facturin
g syst
em (FMS) m
odel, and utili
zed
simulatio
n
to
build a
man
u
f
acturin
g
e
n
vironm
ent ba
sed on
JA
DE
frame
w
ork fo
r multi-agent
to
combi
ne
with
dispatching
rule
s.
Jie
Lin
and
Qin
gqi
Long
[6] p
r
e
s
ente
d
a
mu
lti-agent
-ba
s
e
d
distrib
u
ted si
mulation platform
to su
ppo
rt
t
he extre
m
ely com
p
lex
semi
con
d
u
c
tor ma
nufa
c
tu
ring
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 230
2-40
46
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2402 – 240
8
2403
analysi
s
. Th
e platform d
e
velopme
n
t
and th
e de
si
gn of g
r
a
phi
cal
user i
n
te
rface
were a
l
so
exploited in that pape
r.
As a
re
se
arch of di
sa
ste
r
re
scue
process, this pa
p
e
r
uses com
p
lex ad
aptive
syste
m
theory to e
s
ta
blish
a mod
e
l
contai
ns
re
scue
age
nt
s a
nd victim a
g
e
n
ts. Based o
n
the full analy
s
is
of the behavi
o
r an
d influe
ncin
g facto
r
s in the re
scu
e
pro
c
e
s
s, a model in
cludi
ng judg
ment
s of
the re
scue
m
e
mbe
r
s,
beh
aviors amo
n
g
victims
and
t
he imp
a
ct
of relief
sup
p
lie
s i
s
e
s
tabli
s
h
ed.
In this article,
the disa
ster rescue mo
del
has the follo
wing feature
s
:
(1)
Simul
a
tio
n
of disa
ster environm
ent: Usin
g the
agent si
mula
tion
techn
o
lo
gy to
create
a
virtual disa
ste
r
enviro
n
ment
. Victims and
rescu
e
memb
ers a
r
e ma
ppi
ng as a
gent
s.
(2) Simulati
on of
relief supplies:
Reli
ef supplies
play a very
importa
nt
rol
e
in the
disa
ster
rescu
e
.
In the re
scue p
r
o
c
e
ss,
not o
n
l
y
to c
onsi
der
the intera
ctio
n of
sup
p
lies
betwe
en the
rescu
e
age
nt and victim age
nt, but also
to include
the
intera
ctio
n model
betwe
en the
victims.
(3)
Simulati
on
of the
search a
c
tion: The se
arch
operation
is
t
he first ste
p
in
the rescue
pro
c
e
ss. Rescue
age
nts
need to dete
r
mine
the
ne
xt course
of
action a
c
cord
ing
to
the
curre
n
t
ci
rcum
stan
ce
s. This
req
u
ires t
he si
mula
tion model
can
provide accu
rat
e
judgme
n
t and
rapid respon
se capa
bility of
rescue a
g
e
n
ts und
er different situ
atio
ns.
(4)
Simulation of the re
scue proc
e
s
s: The re
scue p
r
o
c
ess is the co
re of
disa
ster relie
f, but also
the most co
mplex step.
The re
scue p
r
ocess invo
lv
es the ability
to deal with
a variety o
f
different situ
ations,
the match of
different
m
a
teri
als
requ
est
s
and the
col
l
aboration of
different age
n
t
group
s.
The re
st of
this pap
er i
s
orga
nized a
s
:
Section
2
provide
s
th
e theory of
compl
e
x
adaptive
syst
em and
its relation
wi
th the di
sa
ster
rescue
pro
c
e
s
s. In Sectio
n
3, we a
pply
the
prop
osed mu
lti-agent mod
e
l to simulate the resc
ue
process on
the Repa
st p
l
atform. Final
ly,
Section 4 is t
he co
ncl
u
si
on
of this paper.
2. Rese
arch
Metho
d
2.1. Comple
x Adap
tiv
e
Analy
s
is of the Disas
ter
Rescue Pro
c
e
s
s
The theo
ry of complex a
daptive syste
m
has
seve
n relevant concepts: ag
g
r
egatio
n,
nonlin
earity,
mobility, diversity, taggi
ng
, intern
al
mo
del a
nd
build
ing bl
ock. T
h
e first fou
r
a
r
e
certai
n cha
r
a
c
teri
stics of
the
in
dividual whi
c
h will
pla
y
a role in
the
ada
ptation
a
nd evol
ution;
the
latter three
a
r
e th
e m
e
ch
anism
of i
n
tera
ction
bet
wee
n
the
in
dividual
and
the envi
r
onm
ent.
Disaste
r
rescue sy
stem i
s
a
compl
e
x adaptive
system be
ca
use the sy
ste
m
also ha
s the
f
o
llowin
g
cha
r
act
e
ri
st
ic
s of
compl
e
x ada
ptive system
s:
(1)
Aggre
gation:
In the di
sa
ster rescu
e
p
r
oce
s
s, we
re
gard
victims
as victim
age
nts, re
scue
teams
as re
scue
age
nts,
whi
c
h h
a
s m
ade the
mod
e
l ha
s the
a
ggre
gation
chara
c
te
risti
c
.
The victim
a
gent do
esn’t
mean j
u
st
on
e of vi
ctim
s i
n
the g
ene
ra
l se
nse,
but
a group
of
victims
whi
c
h
can
se
ek re
lief, help ea
ch othe
r, as
well a
s
m
a
ke de
cisi
on
s. The rescu
e
agent i
s
also
an a
g
g
r
egate
in
a
certai
n
sen
s
e. It
doe
sn’t m
ean
ju
st the rescu
e
t
eam
whi
c
h
can
offer
reli
ef sup
p
lie
s, but a g
r
ou
p
whi
c
h can re
scue
b
u
ri
ed person, provi
de
me
dical
assista
n
ce in the broad sense.
Theref
ore, in the disa
ster re
scue
process, the
interactio
n
betwe
en ag
e
n
ts pro
m
ote
s
the
evolution
of the system
.
(2)
Nonli
nea
rity: In the di
sast
e
r
re
scue
proc
ess, the inte
raction
betwe
en the va
riou
s ag
ents of
the system
d
oes
not meet
the homo
g
e
neity pr
in
cipl
e. In addition
, the intera
ction bet
ween
agent
s is n
o
t
simply a p
a
ssive on
e-way ca
u
s
al
re
lationship, b
u
t a com
p
le
x adaptive
relation.
(3)
Mobility: In the di
saster
rescue process,
high
resource mobilit
y ex
ists bet
ween
rescue
agent
s a
nd v
i
ctim ag
ents.
These
re
sou
r
ce
s
cont
ai
n h
u
man
re
so
urce
s, mate
rial
re
sou
r
ce
s
and informati
on re
sou
r
ce
s. Human
re
so
urces
refe
r to the colle
ctio
n of worke
r
s
involved in
the re
scue
p
r
oce
s
s, such
as m
edi
cal
rescue,
distri
b
u
te aid
an
d
comman
d
staffs; Materi
al
resou
r
ces in
clud
e re
scu
e
funds an
d relief su
p
p
lies; Inform
ation re
sou
r
ce
s provid
e
informatio
n service
s
, incl
u
d
ing all types
of co
mmu
nication, distre
ss sig
nal for re
scue, etc.
(4)
Diversity: Each
re
scue
team ha
s
its uniqu
e
techn
o
logy, sup
p
lies
a
nd service
cha
r
a
c
teri
stics. Mean
while
there are differe
n
c
e
s
bet
wee
n
the different team
s
whi
c
h have
the sa
me
kin
d
of re
scue t
e
ch
niqu
es. T
herefo
r
e, n
o
t only types o
f
agents
are
diverse, but
also a
gent
s in the same
class are dive
rse.
With the advance of t
he re
scue proce
s
s, the
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Disaste
r
Re
scue Sim
u
latio
n
based on
Com
p
lex Adapt
ive The
o
ry (F
eng Ji
ang
)
2404
variou
s a
gent
s
cha
nge th
ei
r be
havior to
adapt th
e mu
tative situatio
n, whi
c
h
ope
ns
up the
possibility for further diversi
t
y.
(5)
Taggin
g
: Ta
g
g
ing
ca
n b
e
tangible
or in
tangi
ble,
su
ch a
s
te
am
b
anne
r, volu
nteer’
s
dre
s
s
and Sloga
ns.
(6)
Internal mo
de
l: The re
scue
agent will
ch
o
o
se th
e way t
o
impleme
n
t the re
scue
according to
the informati
on of rescu
e
objective
s a
nd ope
ration
al mech
ani
sms and
rule
s. The rescu
e
agent will
cha
nge its intern
al stru
cture to
adapt the ne
w situatio
n when ne
ce
ssary.
(7)
Building blo
c
k: isast
e
r re
scue pe
rson
nel, funds, relief suppli
e
s and equipm
ent, rescue
kno
w
le
dge, o
n
-site i
n
form
a
t
ion, the rescue
team man
ageme
n
t and
operational
mech
ani
sm
are b
u
ildin
g b
l
ocks i
n
the d
i
sa
ster
re
scu
e
pro
c
e
s
s. Di
fferent co
mbi
nation
s
of the
s
e b
u
ilding
blocks esta
bli
s
h differe
nt internal mo
del
s and pro
d
u
c
e
different beh
a
v
ior re
sults.
In summa
ry, the disaste
r
rescue sy
st
em has
the
seven ge
ne
ral cha
r
a
c
teri
stics and
mech
ani
sm o
f
the complex
adaptive syst
em, and so
b
e
see
n
as a
complex ada
ptive system.
2.2. Formal Des
c
ription
of Victim Ag
ents
Disaste
r
re
scue m
odel
virt
ualizes di
sa
ster
env
iro
n
me
nt to a
flat
sp
ace
which i
s
mesh
ed
uniformly an
d
each g
r
id
co
rre
sp
ond
s to 0.5 * 0.5m
space. Accord
ing to the act
ual situatio
n of
environ
menta
l
disa
ste
r
s structu
r
e, the
attribute
val
ue of e
a
ch
grid i
s
assig
ned a
s
foll
o
w
s:
v
i
ct
ims,
re
sc
u
e
membe
r
s o
r
null.
This mo
del wi
ll define ea
ch
victim as an
agent:
Victim Agent = < St, E, V >
St is the state colle
ction o
f
agents at time t;
E is th
e colle
ction o
f
external events that
can
stimulate
the state and
behavio
r of agents; V is a
colle
ction of
agent be
havi
o
rs, in
clu
d
ing
all
the acts ta
ke
n spo
n
tane
ou
sly or
inspire
d
by external events.
S
t
= < loc,
c,
q >
Loc i
s
grid co
ordin
a
tes of the agent at time
t; c is the
curre
n
t health
state of the agent; m
is cu
rrent que
sts of the age
nt.
2.3. Formal Des
c
ription
of Re
scue
Agents
In this model,
every rescu
e
member
(o
r
rescue team
) i
s
define
d
as
an age
nt.
Re
scue Agen
t = < St, E, V
>
St is the state colle
ction o
f
agents at time t;
E is th
e colle
ction o
f
external events that
can
stimulate
the state and
behavio
r of agents; V is a
colle
ction of
agent be
havi
o
rs, in
clu
d
ing
all
the acts ta
ke
n spo
n
tane
ou
sly or
inspire
d
by external events.
S
t
= < loc,
v
,
dir,
m >
Loc is g
r
id
co
ordin
a
tes of t
he a
gent
at ti
me
t; v i
s
the
instanta
neo
u
s
velo
city of t
he a
gent
at time t; dir records the d
i
rectio
n of the agent
mov
e
ment from time t-1 to time t; m is curren
t
sup
p
lie
s of the agent.
Standards of
rescu
e
age
nts’ beh
aviors are defin
ed b
e
low:
(1)
If there is a victim agent in
the visible ra
nge, then mo
ve to it.
(2)
If there i
s
n
o
victim ag
ent
in the vi
sibl
e
ra
nge, th
e
n
choo
se th
e
dire
ction
wh
ich
has the
highe
st den
si
ty of
victim agents to move.
(3)
Material
exch
ange
bet
wee
n
victim
age
n
t
s an
d
re
scue
age
nts. If m
e
et all th
e d
e
m
and
s, then
the victim is rescue
d.
3. Results a
nd Analy
s
is
3.1. Repas
t
Platform
Rep
a
st (Re
c
ursive Po
rou
s
Agent Simulation Tool
kit) is a sim
u
lation-ba
sed
Agent
softwa
r
e a
r
chitecture of
Social S
c
ien
c
e
s
Com
puti
ng Rese
arch
Cente
r
of the University of
Chi
c
ag
o. Thi
s
a
r
chitectu
re
provid
es
a serie
s
of
lib
ra
ries u
s
e
d
to g
enerate, ru
n
and di
spl
a
y the
model a
nd
collect the
rel
a
ted data. Repa
st ha
s the ab
stra
ction
of t
he unde
rlying st
ru
ctu
r
e,
stron
g
scala
b
ility and good performan
ce ability, and has bee
n applied to all asp
e
ct
s of social
simulatio
n
.
Now
Re
pa
st h
a
s
develo
ped
into
a ve
rsa
t
ile, multi-ag
e
n
t sim
u
lation
platform.
Th
e
Rep
a
st Simu
lation progra
m
contai
ns t
he followi
ng
parts: a
gent
cla
ss, mo
del
cla
ss, b
eha
vior
cla
ss,
dat
a
so
urc
e
cla
s
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 230
2-40
46
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2402 – 240
8
2405
Table 1. Co
re
module
s
of the Re
pa
st Simulation p
r
og
ram
Module Illustration
Agent class
Define the pr
ope
rties and behaviors of the agent
Model class
Use templates to build simulation
model
Behavior class
The decouple be
tw
e
en
the simulation scheduler and the agent
Data source class
Anal
y
z
e the simulation program
3. 2 Parameters Initializ
ation
Acco
rdi
ng to
the ab
ove, th
e Repa
st mo
del
sets the
para
m
eters t
o
be
ru
n at
a
50 *
50
environ
ment whi
c
h
in
clud
es
th
ree re
scue age
nt
s.
Paramete
rs
of victim ag
ents a
nd rescue
agent
s are
sh
own in Ta
ble
2
and Tabl
e 3
.
Table 2. Para
meters of victim agents
Parameters
Il
l
u
s
t
rati
on
State
T
w
o states of the
victim agent are included: 0- not r
e
scued and 1- h
a
s been
rescued. At the b
eginning of the simulation, the
y
a
r
e all marked as 0
Service ID
Three se
rvices are included: 1-
w
a
ter, 2-te
nts and 3
-
food. At the
beginning of
the simulation, the
y
ar
e rand
oml
y
assigned.
Request numbe
r
The amount
of service request th
at vict
im agent asked. At the beginning of the
simulation, they
are ran
doml
y
ass
i
gned
Health degre
e
health conditions of victims
Coordinates
Coordinates of th
e victim agent. The
y
a
r
e ran
domly assigned at the
start
Color
The color of the
victim agent is d
e
signed as follows: w
h
ite fo
r the a
gents that
are not
rescued.
All the colors are initially
white
Table 3. Para
meters of re
scue a
gent
s
Parameters
Il
l
u
s
t
rati
on
Storage
Material reserves of each rescue a
gent.
This model
sets three kinds: drinking
w
a
t
e
r, ten
t
s and f
ood
Coordinates
Coordinates of th
e victim agent. The
y
a
r
e ran
domly assigned at the
start
There are ma
ny factors affecti
ng
the re
scue route
s
:
types of
re
sc
ue
serv
i
c
e
s
the
res
c
u
e
agent
ca
n of
fer; the q
uan
tity of equip
m
ents
and
material
s
th
e
re
scue age
nt
ca
n carry;
the
distan
ce b
e
twee
n the re
scue a
gent a
nd the
victim
agent; healt
h
con
d
ition
s
and ag
gregat
ion
den
sity of victim agents.
3.3 Model Simulation Process
(1)
As every cycl
e begin
s
, the victim agent gene
rate
s a reque
st.
(2)
The rescu
e
A
gent sea
r
che
s
whethe
r the
r
e i
s
a victim
agent in th
e a
d
jacent g
r
ids.
If there i
s
,
rescu
e
the victim agent. If it meets all the dema
n
d
s
, the victim a
gent is re
scu
ed; if it only
sat
i
sf
ie
s
k
ki
nds
of
t
h
e
d
e
mand
s,
a
d
d
s
k t
o
th
e h
ealth d
egree
of the vi
ctim age
nt an
d
dedu
cts th
e relief su
pplie
s of the rescu
e
age
nt. If not, the re
scu
e
age
nt dete
r
mine
s the
movement di
rectio
n a
c
cording to the
weight
value o
f
victim agent
s’ num
ber i
n
the visible
rang
e and th
eir health d
e
g
r
ee.
(3)
The victim
a
gents’
a
c
tivity: adjacent vi
ctim
ag
ents sha
r
e su
pplie
s.
If
meet k kind
s of
the
deman
ds, the
healthy degree plu
s
k.
(4)
If not rescued
, the health degre
e
of
the victim agent subtra
cts on
e.
(5)
Whe
n
the re
scue n
u
mbe
r
reached 9
5
%, the simulatio
n
end
s.
3.4 Experimental Analy
s
is
Acco
rdi
ng to
para
m
eters
o
f
the simul
a
tion mo
del a
s
well a
s
m
u
tative con
d
ition
s
duri
ng
the a
c
tual
re
scue
process, this
pap
er
set two
ex
pe
ri
ments with
the diffe
rent
n
u
mbe
r
of
vict
im
agent
s (50
0
a
nd 250 victim
agents).
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Disaste
r
Re
scue Sim
u
latio
n
based on
Com
p
lex Adapt
ive The
o
ry (F
eng Ji
ang
)
2406
Figure 1. The
Repa
st simul
a
tion display
In Figu
re 1, t
he
simulatio
n
run
s
at a 5
0
* 50
enviro
n
m
ent in
cludi
n
g
thre
e rescu
e
age
nts.
The left expe
riment sets 5
00 victim age
nts whil
e
the right sets 25
0
victim agent
s. Grey g
r
ids
are
origin
al locations of three
rescu
e
ag
ent
s. Red,
g
r
ee
n and blu
e
colors re
sp
ecti
vely repre
s
e
n
t
victims
whi
c
h
have b
een
rescue
d by th
ree
re
sc
ue a
gents, while
white rep
r
e
s
ents
th
e
victi
m
whi
c
h h
a
s no
t been
re
scu
ed. To
a
cert
ain extent, vi
ctims of
the same colo
r su
bstantially clu
s
ter
together.
Thi
s
i
s
d
ue to
th
e jud
g
ment
o
f
the di
stan
ce
between
the
re
scue
age
n
t
and th
e victi
m
agent in
de
cision
-ma
k
in
g
crite
r
ia. Mea
n
whil
e othe
r
crite
r
ion i
s
al
so in
clu
ded
(su
c
h a
s
h
eal
th
con
d
ition
s
, etc.), thus
cau
s
es the di
stan
ce bet
we
en some vic
t
ims
of the s
a
me color is
so far.
Figure 2. Nu
mbers of re
scued victims
Figure 2
sho
w
s th
e flu
c
tua
t
ions of
re
scu
ed vict
im
s’ nu
mber
over tim
e
. The
re
sults of two
experim
ents
are almo
st the same. Fro
m
the graph
we can se
e the numb
e
r of
rescue
d victims is
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 230
2-40
46
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2402 – 240
8
2407
risin
g
fast in the begin
n
ing.
But as time
goe
s on,
the curve
s
of re
scue
d victims’
numbe
r sh
ow a
relatively sta
b
le tre
nd till
the num
ber
of re
sc
ued vi
ctims
rea
c
h
e
s
90%
of the
numb
e
r of t
o
tal
victims. This i
s
be
cau
s
e e
a
c
h ag
ent material
s
re
se
rve
is enou
gh in
the early re
scue ope
ration
s,
gene
rally me
et the cu
rren
t reque
st
of
the victims.
Ho
wever, in
the later p
e
ri
od, due to t
he
redu
ction
of
t
he
m
a
terial
s reserve, the rescue ag
ent
s can't satisfy the requ
est
s
of victims. So in
each cycle
the cha
n
ce
of
the re
scue agent
s whi
c
h
do
not
parti
cipate
in th
e
re
scue
process
increa
se
s, that cau
s
e
s
the cu
rve a
p
p
r
oaching o
n
a
stable. In
a cycle of this
model, a re
scue
agent
doe
sn’t
take
the
re
scue
op
eratio
ns
ca
n b
e
su
mmari
zed
a
s
the follo
wing
two
situatio
n
s
:
The mate
rial
stora
ge
can
'
t satisfy the
re
que
st of
the
curre
n
t victim agent; the di
stan
ce b
e
twe
e
n
the rescu
e
ag
ent and the n
eare
s
t victim agent is too f
a
r.
Figure 3. Storage
s of No.1
rescu
e
age
nt
Figure 3 sho
w
s the mate
ri
al stora
ge ch
ange of No.1
rescu
e
age
nt. This model h
a
s thre
e
kind
s of relie
f supplie
s: drinkin
g
wate
r, tents
and food, respe
c
tively, marke
d
with a differe
nt
notation i
n
th
e figure. Th
e
left cu
rve i
s
decrea
s
in
g fa
st while th
e right curve
re
mains relatively
stable. Si
nce
the l
e
ft expe
riment
ha
s
5
00 victim
ag
e
n
ts
while
the
right
ha
s 2
5
0
,
the d
e
n
s
ity o
f
victim ag
ents of the
left i
s
mu
ch
more
than the
ri
ght
one.
Thi
s
causes rescu
e
age
nts re
scue
more
victim
s while consum
e
more relief
material
s.
4. Conclusio
n
This
pap
er u
s
e
s
compl
e
x adaptive a
n
d
multi-
ag
ent t
heory
as th
e
basi
c
id
ea to
analyze
the di
sa
ster rescue
process. A
multi
-
ag
ent mod
e
l
co
ntains re
scue
age
nts
and
victim ag
ents
ha
s
been p
r
op
ose
d
on the Rep
a
st platform a
nd has b
een
sh
o
w
n that it is efficient in dealin
g with the
disa
ster re
scue si
mulation
. Therefo
r
e,
this
ap
proa
ch may give
experts a m
o
re
credibl
e
and
informative sugge
stion to
make b
e
tter d
e
ci
sion
s.
Future
re
se
arch m
a
y focu
s on ho
w to m
a
ke it
clo
s
e
r
to the hu
man
way of thin
ki
ng an
d
how to
ma
ke
it more
suitab
le to de
scribe
the hum
an d
e
ci
sion
-ma
k
in
g pro
c
e
s
s in t
he interactio
n
of so
cial g
r
o
ups. Fu
rthe
rmore, e
n
viro
nment c
han
g
e
s may b
e
consi
dered in
the future m
odel,
su
ch a
s
addi
ng we
ather o
r
terrain p
a
ra
meters.
Ackn
o
w
l
e
dg
ements
This pa
per is
financi
a
lly su
pporte
d by
th
e National
Na
tural S
c
ien
c
e
Found
ation
of Chi
n
a
(No: 91
024
00
8, 41101
516
and 61
203
37
3).
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Disaste
r
Re
scue Sim
u
latio
n
based on
Com
p
lex Adapt
ive The
o
ry (F
eng Ji
ang
)
2408
Referen
ces
[1]
Qingli
n
Guo, Ming Z
h
a
ng. A novel a
ppro
a
c
h
fo
r multi-ag
e
n
t-base
d
Intelli
gent Man
u
facturin
g S
y
stem
.
Information Sci
ences
. 20
09; 1
79: 307
9–
30
90
.
[2]
Arend L
i
gte
n
b
e
r
ga, Monic
a
W
a
cho
w
i
cza, Arn
o
ld
K Bre
g
ta, Adrie Be
ule
n
sb,
Dirk L Kettenis
b
. A desig
n
and a
p
p
licati
o
n of a multi-a
gent
s
y
stem for simulati
on
of multi-actor
spatia
l pla
n
n
i
n
g
.
Journa
l of
Enviro
n
m
enta
l
Mana
ge
me
nt
. 200
4; 72:43
–5
5.
[3]
Bonn
ie R
u
b
e
n
s
tein Mo
ntan
o, Victoria Y
oon
, Ke
vin Dr
um
me
y
,
Ja
y
Li
eb
o
w
itz. Age
n
t l
earn
i
ng
in th
e
multi-ag
ent co
ntracting s
y
ste
m
[MACS].
De
cision S
upp
ort Systems
. 20
08
; 45: 140–
149.
[4]
Kevin F
R
Li
u. Age
n
t-bas
ed res
ourc
e
discov
e
r
y
a
r
chitecture f
o
r
envir
onm
ent
al em
erg
enc
y
mana
geme
n
t.
Expert Systems w
i
th Applicati
ons
. 200
4; 27: 77–
95
[5]
Kai Yin
g
Ch
en,
Chun Ja
y C
h
e
n
. Appl
yi
ng mu
lti-ag
ent
techn
i
que i
n
multi-se
ction fle
x
i
b
le m
anufactur
i
n
g
sy
s
t
e
m
.
Expert
Systems w
i
th Appl
icatio
ns
. 2
010; 37: 7
310
–
731
8.
[6]
Jie L
i
n
and
Qing
qi L
o
n
g
. D
e
vel
opme
n
t of
a multi-
ag
ent
-base
d
distri
bu
ted simu
latio
n
platform for
semico
nductor manufactur
i
ng.
Expert Systems w
i
th Applicati
ons
. 201
1; 38: 523
1–
523
9.
[7]
Xi
ao
le
Xu, W
e
i Hu
ang, S
hen
g
y
on
g C
hen,
L
i
xin Ga
o. Co
n
s
ensus
of Mult
i-ag
ent S
y
stem
s
w
i
t
h
T
i
me
Dela
ys
an
d Me
asurem
ent No
i
s
es. T
ELKOMNIKA Indon
esi
an Jo
urna
l
of Electrical Eng
i
neer
ing
. 20
12
;
10(6): 13
70-
13
80.
[8]
Jing
an Y
a
n
g
,
Yanb
in Z
h
ua
n
g
. T
o
w
a
r
d
s Be
havi
o
r C
ontrol
for Evol
ution
a
r
y
Rob
o
t Bas
e
d o
n
R
L
w
i
t
h
ENN.
IAES Internati
ona
l Jour
nal of Ro
botics
and Auto
matio
n
. 2012; 1(
1): 33-48.
[9]
Siti Maimunah, Husni S Sa
stramih
a
rdj
a
, D
w
i H W
i
d
y
antoro,
NFN Kuspri
ya
nto. CT
-FC: more
Compre
he
nsiv
e T
r
aversal F
o
cused Cr
a
w
l
e
r.
T
E
LKOMNIKA Indon
esia
n Jour
nal
of Electrical
Engi
neer
in
g
. 2012; 10(
1): 189
-198.
[10]
Che
n
, X. Ag
e
n
t-base
d
mod
e
lli
ng
and s
i
m
u
lati
on
of ur
b
an ev
acuati
o
n
:
Relativ
e
effe
ctiveness
of
simulta
neo
us and
sta
ged ev
acuati
on
strate
gies.
Jour
na
l o
f
the Operatio
nal R
e
se
arch
Society
. 20
08;
59(1): 25-
33.
[11]
Aipi
ng T
,
Aihua
W
.
An intelli
ge
nt simu
latio
n
system for earth
quak
e disaster
assessme
n
t.
Computers &
Geosciences
. 200
9; 35: 871-
879.
[12]
Keisuk
e Un
o, Kazuo K
a
sh
i
y
a
m
a.
Deve
lopm
ent of simu
latio
n
s
y
st
em for th
e dis
a
ster evac
uatio
n b
a
s
e
d
on multi-
age
nt mode
l usin
g
GIS. 2008; 13(S
1
): 348-35
3.
Evaluation Warning : The document was created with Spire.PDF for Python.