on i at l mu .uniroma2
TRANSCRIPT
An
Intr
oduc
tion
to A
gent
-bas
ed
Md
lid
Sil
iM
odel
ing
and
Sim
ulat
ion
Dr.
Emili
ano
Casa
licch
io
casa
licch
io@
ing.
uniro
ma2
.itg
Dow
nloa
d @
ww
w.e
mili
anoc
asal
icch
io.e
u
(tal
ks &
sem
inar
s se
ctio
n)
Otli
Out
line
•Pa
rt1:
An
intr
oduc
tion
to A
gent
-bas
ed M
odel
ing
and
Sim
ulat
ion
(ABM
S)–
Mot
ivat
ion
–W
hat i
s an
age
ntTh
df
ABM
S–
The
need
for
ABM
S–
Back
grou
nd o
n A
BMS
–W
hyan
dw
hen
ABM
SW
hy a
nd w
hen
ABM
S•
Part
2:
–A
BMS
appl
icat
ions
ABM
S ap
plic
atio
ns–
How
to d
o A
BMS
•Pa
rt 3
:–
Elec
tric
ity m
arke
t, s
uppl
y ch
ain
exam
ple
–A
BMS
in W
orkf
low
s an
d BP
re-e
ngin
eeri
ng
Bibl
ih
Bibl
iogr
aphy
•Ch
arle
s M
. Mac
al, M
icha
el J.
Nor
th, T
UTO
RIA
L O
N A
GEN
T-BA
SED
M
OD
ELIN
G A
ND
SIM
ULA
TIO
N, P
roce
edin
gs o
f the
200
6 W
inte
r Sim
ulat
ion
Conf
eren
ce
•Ch
arle
s M
. Mac
al, M
icha
el J.
Nor
th, T
UTO
RIA
L O
N A
GEN
T-BA
SED
M
OD
ELIN
GA
ND
SIM
ULA
TIO
NPA
RT2
HO
WTO
MO
DEL
WIT
HA
GEN
TSM
OD
ELIN
G A
ND
SIM
ULA
TIO
N P
ART
2: H
OW
TO
MO
DEL
WIT
H A
GEN
TS,
Proc
eedi
ngs
of th
e 20
06 W
inte
r Sim
ulat
ion
Conf
eren
ce
•Ch
arle
sM
.Mac
al,M
icha
elJ.
Nor
th,M
anag
ing
Busi
ness
Com
plex
ity:
Char
les
M. M
acal
, Mic
hael
J. N
orth
, Man
agin
g Bu
sine
ss C
ompl
exit
y:
disc
over
y st
rate
gic
solu
tion
wit
h ag
ent-
base
d m
odel
ing
and
sim
ulat
ion,
O
xfor
d U
nive
rsity
Pre
ss, 2
007
Mti
tiM
otiv
atio
n
•Sy
stem
s ar
e ev
en m
ore
com
plex
and
in
terd
epen
dent
inte
rdep
ende
nt–
Fina
ncia
l and
gov
ernm
ent p
roce
sses
and
ser
vice
s de
pend
on
IT s
ervi
ces;
IT s
ervi
ces
depe
nds
on
p;
pel
ectr
icity
–G
oods
sup
ply
chai
ns d
epen
d on
tran
spor
tatio
n sy
stem
s, IT
ser
vice
s, e
lect
rici
ty, o
il di
stri
butio
n, w
ater
di
stri
butio
n et
c…Th
ih
df
dli
d•
Ther
e is
the
need
for
new
mod
elin
g an
d si
mul
atio
n pa
radi
gms
that
allo
w to
thin
k sy
stem
s i
diff
tin
a d
iffer
ent w
ay:
–Pa
rts
mak
e th
e w
hole
Pt
kth
hl
Part
s m
ake
the
who
le•
Age
ntba
sed
mod
elin
gan
dsi
mul
atio
nis
foun
ded
on•
Age
nt-b
ased
mod
elin
g an
d si
mul
atio
n is
foun
ded
on
the
notio
n th
at:
–Th
ew
hole
ofm
any
syst
ems
oror
gani
zatio
nis
grea
ter
then
–Th
e w
hole
of m
any
syst
ems
or o
rgan
izat
ion
is g
reat
er th
en
the
sum
of i
ts c
onst
ituen
t par
ts•
Tom
anag
esu
chsy
stem
sth
esy
stem
sor
orga
niza
tions
To m
anag
e su
ch s
yste
ms
the
syst
ems
or o
rgan
izat
ions
m
ust b
e un
ders
tood
as
colle
ctio
ns o
f int
erac
ting
com
pone
nts
–Ea
ch o
f the
se c
ompo
nent
s ha
s its
ow
n ru
les
and
resp
onsi
bilit
ies
Nft
ht
lt
lt
lth
bh
i–
Non
e of
the
com
pone
nts
com
plet
ely
cont
rols
the
beha
vior
of
the
syst
em–
All
the
com
pone
nts
cont
ribu
teto
the
resu
ltsin
ala
rge
or–
All
the
com
pone
nts
cont
ribu
te to
the
resu
lts in
a la
rge
or
smal
l way
Cl
Ad
tiS
tCo
mpl
ex A
dapt
ive
Syst
ems
•A
col
lect
ion
of c
ompo
nent
s w
ith th
e ab
ove
char
acte
rist
ics
issa
idto
bea
char
acte
rist
ics
is s
aid
to b
e a
–Co
mpl
ex A
dapt
ive
Syst
em (C
AS)
•CA
S ar
e ch
arac
teri
zed
by e
mer
gent
beh
avio
r:S
tti
hth
lt
lt–
Syst
em re
actio
n w
here
the
com
plet
e re
sults
are
mor
e th
en th
e su
m o
f the
indi
vidu
al c
ompo
nent
s ou
tcom
es
•M
anag
ing
a CA
S re
quire
a g
ood
unde
rsta
ndin
g of
em
erge
ntbe
havi
orem
erge
nt b
ehav
ior
•A
BMS
help
s in
und
erst
andi
ng e
mer
gent
beh
avio
r
Wha
tis
anag
ent?
Wha
t is
an a
gent
?
Wha
t is
an a
gent
•N
o un
iver
sal a
gree
men
t on
the
prec
ise
defin
ition
of t
he te
rm
“t”
“age
nt”
–(B
onab
eau
2001
) any
type
of i
ndep
ende
nt c
ompo
nent
(sof
twar
e, m
odel
, in
divi
dual
etc
);an
inde
pend
entc
ompo
nent
’sbe
havi
orca
nra
nge
from
indi
vidu
al, e
tc.);
an
inde
pend
ent c
ompo
nent
s be
havi
or c
an ra
nge
from
pr
imiti
ve re
activ
e de
cisi
on r
ules
to c
ompl
ex a
dapt
ive
inte
llige
nce.
–(M
ello
ulie
tal.
2003
)com
pone
nts
with
anad
aptiv
ebe
havi
or;c
ompo
nent
s(M
ello
ulie
t al.
2003
) com
pone
nts
with
an
adap
tive
beha
vior
; com
pone
nts
that
can
in s
ome
sens
e le
arn
from
thei
r en
viro
nmen
ts a
nd c
hang
e th
eir
beha
vior
s in
resp
onse
.
–(C
asti
1997
) age
nts
shou
ld c
onta
in b
oth
base
-leve
l rul
es fo
r beh
avio
r as
w
ell a
s a
high
er-le
vel s
et o
f “ru
les
to c
hang
e th
e ru
les.”
The
bas
e le
vel r
ules
pr
ovid
ere
spon
ses
toth
een
viro
nmen
twhi
leth
e“r
ules
toch
ange
the
rule
s”pr
ovid
e re
spon
ses
to th
e en
viro
nmen
t whi
le th
e ru
les
to c
hang
e th
e ru
les
pr
ovid
e ad
apta
tion.
•Th
efu
ndam
enta
lfea
ture
ofan
agen
tis
the
capa
bilit
yof
the
The
fund
amen
tal f
eatu
re o
f an
agen
t is
the
capa
bilit
y of
the
com
pone
nt to
mak
e in
depe
nden
t dec
isio
ns.
•Th
isre
quire
sag
ents
tobe
acti
vera
ther
than
pure
ly•
This
requ
ires
agen
ts to
be
acti
ve ra
ther
tha
n pu
rely
pa
ssiv
e.
Age
nts
char
acte
rist
ics
g
•A
n ag
ent i
s id
entif
iabl
e (A
gent
s ar
e se
lf-co
ntai
ned)
–a
disc
rete
indi
vidu
al w
ith a
set
of c
hara
cter
istic
s an
d ru
les
gove
rnin
g its
beh
avio
rs a
nd d
ecis
ion-
mak
ing
capa
bilit
y
–Th
edi
scre
tene
ssre
quire
men
tim
plie
sth
atan
agen
thas
aTh
e di
scre
tene
ss re
quire
men
t im
plie
s th
at a
n ag
ent h
as a
bo
unda
ry a
nd o
ne c
an e
asily
det
erm
ine
•w
heth
erso
met
hing
ispa
rtof
anag
ent,
whe
ther
som
ethi
ng is
par
t of a
n ag
ent,
•is
not
par
t of a
n ag
ent,
•or
is a
sha
red
char
acte
rist
ic.
Age
ntch
arac
teri
stic
s(c
ont
)A
gent
cha
ract
eris
tics
(con
t.)
Ati
itt
dli
ii
it
ith•
An
agen
t is
situ
ated
, liv
ing
in a
n en
viro
nmen
t with
w
hich
it in
tera
cts
with
oth
er a
gent
s.
–A
gent
s ha
ve p
roto
cols
for i
nter
actio
n w
ith o
ther
age
nts,
su
ch a
s co
mm
unic
atio
n pr
otoc
ols,
and
the
capa
bilit
y to
re
spon
d to
the
envi
ronm
ent.
–A
gent
s ha
ve th
e ab
ility
to re
cogn
ize
and
dist
ingu
ish
the
gy
gg
trai
ts o
f oth
er a
gent
s.
At
ht
iti
(t
)A
gent
cha
ract
eris
tics
(con
t.)
•A
n ag
ent i
s go
al-d
irect
ed, h
avin
g go
als
to
achi
eve
(not
nece
ssar
ilyob
ject
ives
toac
hiev
e (n
ot n
eces
sari
ly o
bjec
tives
to
max
imiz
e) w
ith re
spec
t to
its b
ehav
iors
•A
nag
enti
sau
tono
mou
san
dse
lf-di
rect
edA
n ag
ent i
s au
tono
mou
s an
d se
lfdi
rect
ed
–A
n ag
ent c
an fu
nctio
n in
depe
nden
tly in
its
envi
ronm
ent a
nd in
its
deal
ings
with
oth
er
gag
ents
, at l
east
ove
r a
limite
d ra
nge
of s
ituat
ions
•A
n ag
ent i
s fle
xibl
e, a
nd h
as th
e ab
ility
to
gy
lear
n an
d ad
apt i
ts b
ehav
iors
ove
r tim
e ba
sed
on e
xper
ienc
e–
This
requ
ires
som
e fo
rm o
f mem
ory
–A
n ag
ent m
ay h
ave
rule
s th
at m
odify
its
rule
s of
be
havi
or
Mt
ht
iti
ft
Met
a-ch
arac
teri
stic
s of
age
nts
•A
gent
s ar
e di
vers
e, h
eter
ogen
eous
, and
dyn
amic
in
thei
r att
ribu
tes
and
beha
vior
al ru
les.
•
Beha
vior
al ru
les
vary
–
in th
eir s
ophi
stic
atio
n,
p,
–ho
w m
uch
info
rmat
ion
is c
onsi
dere
d–
in th
e ag
ent d
ecis
ions
(cog
nitiv
e “l
oad”
), –
the
agen
t’s in
tern
al m
odel
s of
the
exte
rnal
wor
ld in
clud
ing
othe
r age
nts,
dh
ff
hi
–an
d th
e ex
tent
of m
emor
y of
pas
t eve
nts
the
agen
t ret
ains
an
d us
es in
its
deci
sion
s.•
Age
nts
also
vary
byth
eira
ttri
bute
san
dac
cum
ulat
ed•
Age
nts
also
var
y by
thei
r att
ribu
tes
and
accu
mul
ated
re
sour
ces.
Thd
fA
BM4
The
need
for
ABM
: 4 re
ason
s
•Th
e an
swer
is b
ecau
se w
e liv
e in
an
ii
ll
ldin
crea
sing
ly c
ompl
ex w
orld
–Sy
stem
inte
rdep
ende
ncie
sy
p
–To
o m
uch
com
plex
ity
–Ev
en m
ore
finer
leve
l of g
ranu
lari
ty o
f dat
a
–Ev
enm
ore
high
erco
mpu
tatio
nalp
ower
Even
mor
e hi
gher
com
puta
tiona
l pow
er
Wh
ABM
ti
td
di
Why
ABM
: sys
tem
inte
rdep
ende
ncie
s
•Th
e sy
stem
s th
at w
e ne
ed to
ana
lyze
and
mod
el
are
beco
min
gm
ore
com
plex
inte
rms
ofth
eir
are
beco
min
g m
ore
com
plex
in te
rms
of th
eir
inte
rdep
ende
ncie
s.•
The
trad
ition
alm
odel
ing
tool
sar
eno
tas
The
trad
ition
al m
odel
ing
tool
s ar
e no
t as
appl
icab
le a
s th
ey o
nce
wer
e.
–A
nex
ampl
eap
plic
atio
nar
eais
the
dere
gula
tion
ofth
e–
An
exam
ple
appl
icat
ion
area
is th
e de
regu
latio
n of
the
elec
tric
pow
er in
dust
ry.
–In
terd
epen
denc
ies
amon
gin
fras
truc
ture
s(e
lect
ric
Inte
rdep
ende
ncie
s am
ong
infr
astr
uctu
res
(ele
ctri
c po
wer
, nat
ural
gas
, tra
nspo
rtat
ion,
pet
role
um, w
ater
, te
leco
mm
unic
atio
ns, e
tc.)
are
beco
min
g th
e fo
cus
blh
hh
publ
ic a
tten
tion
as th
ese
syst
ems
appr
oach
thei
r de
sign
lim
its a
nd s
uffe
r reg
ular
bre
akdo
wns
.
Wh
AB
MT
hl
itW
hy A
BM
: Too
muc
h co
mpl
exity
•So
me
syst
ems
have
alw
ays
been
too
com
plex
for
us
to a
dequ
atel
y m
odel
.
•Fo
rex
ampl
em
odel
ing
econ
omic
mar
kets
has
For
exam
ple,
mod
elin
g ec
onom
ic m
arke
ts h
as
trad
ition
ally
relie
d on
the
notio
ns o
f per
fect
mar
kets
, ho
mog
eneo
usag
ents
and
long
run
equi
libri
umho
mog
eneo
us a
gent
s, a
nd lo
ng-r
un e
quili
briu
m
beca
use
thes
e as
sum
ptio
ns m
ade
the
prob
lem
s l
llll
blan
alyt
ical
ly a
nd c
ompu
tatio
nally
trac
tabl
e.
•W
e ar
e be
ginn
ing
to b
e ab
le to
take
a m
ore
real
istic
g
gvi
ew o
f the
se s
yste
ms
thro
ugh
ABM
S.
Why
ABM
: fin
er le
vel o
f dat
a y
gran
ular
ity•
Dat
a ar
e be
com
ing
orga
nize
d in
to d
atab
ases
at f
iner
le
vels
ofgr
anul
arity
leve
ls o
f gra
nula
rity
•M
icro
-dat
a ca
n no
w s
uppo
rt m
icro
-sim
ulat
ions
Why
ABM
: hig
her c
ompu
tatio
nal
yg
ppo
wer
•Co
mpu
tatio
nal p
ower
is a
dvan
cing
rapi
dly
•W
e ca
n no
w c
ompu
te la
rge-
scal
e m
icro
-sim
ulat
ion
mod
els
that
wou
ld n
ot h
ave
been
pla
usib
le ju
st a
p
jco
uple
of y
ears
ago
(200
5!!!
).
Otli
Out
line
•Pa
rt1:
An
intr
oduc
tion
toA
BMS
•Pa
rt 1
: An
intr
oduc
tion
to A
BMS
–M
otiv
atio
n–
Wha
tis
anag
ent
Wha
t is
an a
gent
–Th
e ne
ed fo
r A
BMS
–W
hy a
nd w
hen
ABM
S–
Back
grou
nd o
n A
BMS
•Pa
rt 2
:–
ABM
S ap
plic
atio
ns–
How
to d
o A
BMS
Pt3
•Pa
rt 3
:–
Elec
tric
ity m
arke
t, s
uppl
y ch
ain
exam
ple
ABM
Sin
Wor
kflo
ws
and
BPre
engi
neer
ing
–A
BMS
in W
orkf
low
s an
d BP
re-e
ngin
eeri
ng
Bk
dA
BMS
Back
grou
nds
on A
BMS
•A
BMS
has
conn
ectio
ns to
man
y ot
her f
ield
s in
clud
ing
–co
mpl
exity
sci
ence
, p
y,
–sy
stem
s sc
ienc
e,
–Sy
stem
s D
ynam
ics,
Sear
ch “
syst
em s
cien
ce”
or “
syst
em
dyna
mic
s” o
n w
ikip
edia
yy
–co
mpu
ter s
cien
ce,
–m
anag
emen
t sci
ence
, –
the
soci
al s
cien
ces
in g
ener
al, a
nd
–tr
aditi
onal
mod
elin
g an
d si
mul
atio
n•
ABM
S dr
aws
on th
ese
field
s fo
r –
its th
eore
tical
foun
datio
ns,
–its
con
cept
ual w
orld
vie
w a
nd p
hilo
soph
y, a
nd
–fo
r app
licab
le m
odel
ing
tech
niqu
es.
Bk
dA
BMS
(t
)Ba
ckgr
ound
s on
ABM
S (c
ont.
)
•A
BMS
has
its d
irect
his
tori
cal r
oots
in c
ompl
ex a
dapt
ive
syst
ems
(CA
S)
–“s
yste
ms
are
built
from
the
grou
nd-u
p,”
in c
ontr
ast t
o th
e to
p-do
wn
syst
ems
view
take
n by
Sys
tem
s D
ynam
ics.
•CA
S –
conc
erns
itse
lf w
ith th
e qu
estio
n of
how
com
plex
b
hi
ii
ti
tbe
havi
ors
aris
e in
nat
ure
amon
g m
yopi
c, a
uton
omou
s ag
ents
.–
was
orig
inal
lym
otiv
ated
byin
vest
igat
ions
into
adap
tatio
nw
as o
rigi
nally
mot
ivat
ed b
y in
vest
igat
ions
into
ada
ptat
ion
and
emer
genc
e of
bio
logi
cal s
yste
ms.
–ha
ve th
e ab
ility
to s
elf-
orga
nize
and
dyn
amic
ally
reor
gani
ze
thei
r com
pone
nts
in w
ays
bett
er s
uite
d to
sur
vive
and
ex
cel i
n th
eir e
nviro
nmen
ts, a
nd th
is a
dapt
ive
abili
ty
occu
rsre
mar
kabl
yov
eran
enor
mou
sra
nge
ofsc
ales
occu
rs, r
emar
kabl
y, o
ver
an e
norm
ous
rang
e of
sca
les
Back
grou
nds
on A
BMS
g•
Prop
ertie
s an
d m
echa
nism
s co
mm
on to
all
CAS
(Hol
land
19
95):
1995
):•
CAS
prop
ertie
s–
Agg
rega
tion:
allo
ws
grou
psto
form
,A
ggre
gatio
n: a
llow
s gr
oups
to fo
rm,
–N
onlin
eari
ty: i
nval
idat
es s
impl
e ex
trap
olat
ion,
–Fl
ows:
allo
w th
e tr
ansf
er a
nd tr
ansf
orm
atio
n of
reso
urce
s an
d in
form
atio
n,–
Div
ersi
ty: a
llow
s ag
ents
to b
ehav
e di
ffer
ently
from
one
ano
ther
and
of
ten
lead
sto
the
syst
empr
oper
tyof
robu
stne
ssof
ten
lead
s to
the
syst
em p
rope
rty
of ro
bust
ness
. •
CAS
mec
hani
sms:
–Ta
ggin
g: a
llow
s ag
ents
to b
e na
med
and
reco
gniz
ed,
ggg
gg
,–
Inte
rnal
mod
els:
allo
ws
agen
ts to
reas
on a
bout
thei
r wor
lds,
–
Build
ing
bloc
ks: a
llow
s co
mpo
nent
s an
d w
hole
sys
tem
s to
be
df
ll
fi
lco
mpo
sed
of m
any
leve
ls o
f sim
pler
com
pone
nts.
•
Thes
e CA
S pr
oper
ties
and
mec
hani
sms
prov
ide
a us
eful
fr
amew
ork
for
desi
gnin
gag
ent-
base
dm
odel
sfr
amew
ork
for
desi
gnin
g ag
ent-
base
d m
odel
s.
Otli
Out
line
•Pa
rt1:
An
intr
oduc
tion
toA
BMS
•Pa
rt 1
: An
intr
oduc
tion
to A
BMS
–M
otiv
atio
n–
Wha
tis
anag
ent
Wha
t is
an a
gent
–Th
e ne
ed fo
r A
BMS
–Ba
ckgr
ound
on
ABM
S–
Why
and
whe
n A
BMS
•Pa
rt 2
:–
ABM
S ap
plic
atio
ns–
How
to d
o A
BMS
Pt3
•Pa
rt 3
:–
Elec
tric
ity m
arke
t, s
uppl
y ch
ain
exam
ple
ABM
Sin
Wor
kflo
ws
and
BPre
engi
neer
ing
–A
BMS
in W
orkf
low
s an
d BP
re-e
ngin
eeri
ng
Whe
nA
BMS
Whe
n A
BMS
•W
hen
ther
eis
ana
tura
lrep
rese
ntat
ion
asag
ents
Whe
n th
ere
is a
nat
ural
repr
esen
tatio
n as
age
nts
•W
hen
ther
e ar
e de
cisi
ons
and
beha
vior
s th
at c
an b
e de
fined
di
scre
tely
(with
bou
ndar
ies)
hh
dd
hh
bh
•W
hen
it is
impo
rtan
t tha
t age
nts
adap
t and
cha
nge
thei
r beh
avio
rs•
Whe
n it
is im
port
ant t
hat a
gent
s le
arn
and
enga
ge in
dyn
amic
st
rate
gic
beha
vior
sst
rate
gic
beha
vior
s•
Whe
n it
is im
port
ant t
hat a
gent
s ha
ve a
dyn
amic
rela
tions
hips
with
ot
her
agen
ts, a
nd a
gent
rela
tions
hips
form
and
dis
solv
e•
Whe
n it
is im
port
ant t
hat a
gent
s fo
rm o
rgan
izat
ions
, and
ad
apta
tion
and
lear
ning
are
impo
rtan
t at t
he o
rgan
izat
ion
leve
l•
Whe
nit
isim
port
antt
hata
gent
sha
vea
spat
ialc
ompo
nent
toth
eir
Whe
n it
is im
port
ant t
hat a
gent
s ha
ve a
spa
tial c
ompo
nent
to th
eir
beha
vior
s an
d in
tera
ctio
ns•
Whe
n th
e pa
st is
no
pred
icto
r of t
he fu
ture
•W
hen
scal
ing-
up to
arb
itrar
y le
vels
is im
port
ant
•W
hen
proc
ess
stru
ctur
al c
hang
e ne
eds
to b
e a
resu
lt of
the
mod
el,
rath
erth
ana
mod
elin
put
rath
er th
an a
mod
el in
put.
Otli
Out
line
•Pa
rt1:
An
intr
oduc
tion
toA
BMS
•Pa
rt 1
: An
intr
oduc
tion
to A
BMS
–M
otiv
atio
n–
Wha
tis
anag
ent
Wha
t is
an a
gent
–Th
e ne
ed fo
r A
BMS
–W
hy a
nd w
hen
ABM
S–
Back
grou
nd o
n A
BMS
•Pa
rt 2
:–
ABM
S ap
plic
atio
ns–
How
to d
o A
BMS
Pt3
•Pa
rt 3
:–
Elec
tric
ity m
arke
t, s
uppl
y ch
ain
exam
ple
ABM
Sin
Wor
kflo
ws
and
BPre
engi
neer
ing
–A
BMS
in W
orkf
low
s an
d BP
re-e
ngin
eeri
ng
ABM
S A
pplic
atio
nsS
ppca
tos
Pti
lt
bd
dli
•Pr
actic
al a
gent
-bas
ed m
odel
ing
and
sim
ulat
ion
is a
ctiv
ely
bein
g ap
plie
d in
man
y ar
eas
dl
bh
h•
mod
elin
g ag
ent b
ehav
ior
in th
e st
ock
mar
ket (
LeBa
ron
2002
) an
d su
pply
cha
ins
(Fan
g et
al.
2002
)20
02)
•pr
edic
ting
the
spre
ad o
f ep
idem
ics
(Hua
ng e
t al.
2004
) d
hh
fbf
and
the
thre
at o
f bio
-war
fare
(C
arle
y20
06),
•m
odel
ing
the
grow
th a
nd
decl
ine
of a
ncie
nt c
ivili
zatio
ns
(Koh
ler
et a
. 200
5)•
mod
elin
g th
e co
mpl
exiti
es o
f g
pth
e hu
man
imm
une
syst
em
(Fol
cik
and
Oro
sz20
06),
•an
dm
any
othe
rar
eas
and
man
y ot
her
area
s
ABM
S A
pplic
atio
ns
•A
BMS
appl
icat
ions
rang
e fr
om
ppg
–sm
all,
eleg
ant,
min
imal
ist m
odel
s–
to la
rge-
scal
e de
cisi
on s
uppo
rt s
yste
ms.
gpp
y•
Min
imal
ist m
odel
s ar
e ba
sed
on a
set
of i
deal
ized
as
sum
ptio
ns, d
esig
ned
to c
aptu
re o
nly
the
mos
t sal
ient
fe
atur
es o
f a s
yste
m.
–a
wid
e ra
nge
of a
ssum
ptio
ns c
an b
e va
ried
ove
r a la
rge
bf
il
tinu
mbe
r of s
imul
atio
ns.
•D
ecis
ion
supp
ort m
odel
s te
nd to
be
larg
e-sc
ale
appl
icat
ions
desi
gned
toan
swer
abr
oad
rang
eof
real
appl
icat
ions
, des
igne
d to
ans
wer
a b
road
rang
e of
real
-w
orld
pol
icy
ques
tions
. –
incl
udes
real
data
–in
clud
es re
al d
ata
–ha
s pa
ssed
som
e de
gree
of v
alid
atio
n te
stin
g to
est
ablis
h cr
edib
ility
in th
eir r
esul
ts.
y
Ht
dA
BMS
How
to d
o A
BMS
•At
a g
ener
al le
vel,
one
goes
abo
ut b
uild
ing
an a
gent
-bas
ed
mod
el in
muc
h th
e sa
me
way
as
any
othe
r ty
pe o
f mod
el
1.id
entif
y th
e pu
rpos
e of
the
mod
el, t
he q
uest
ions
the
mod
el is
in
tend
edto
answ
eran
den
gage
the
pote
ntia
luse
rsin
the
inte
nded
to a
nsw
er a
nd e
ngag
e th
e po
tent
ial u
sers
in th
e pr
oces
s.
2.s y
stem
atic
ally
ana
lyze
the
syst
em u
nder
stu
dy, i
dent
ifyin
g y
yy
yy,
yg
com
pone
nts
and
com
pone
nt in
tera
ctio
ns, r
elev
ant d
ata
sour
ces,
and
so
on.
3ap
ply
the
mod
elan
dco
nduc
tase
ries
of“w
hat
if”
3.ap
ply
the
mod
el a
nd c
ondu
ct a
ser
ies
of
wha
t-if
ex
peri
men
ts b
y sy
stem
atic
ally
var
ying
par
amet
ers
and
assu
mpt
ions
. d
dh
bf
hd
ld
lb
4.un
ders
tand
the
robu
stne
ss o
f the
mod
el a
nd it
s re
sults
by
usin
g se
nsiti
vity
ana
lysi
s an
d ot
her
tech
niqu
es.
How
to d
o A
BMS:
age
nt p
rosp
ectiv
e vs
gp
ppr
oces
s-ba
sed
pros
pect
ive
•A
gent
-bas
ed m
odel
ing
brin
gs w
ith it
a fe
w u
niqu
e as
pect
s ow
ing
to
the
fact
that
ABM
S ta
kes
the
agen
t per
spec
tive,
firs
t and
fore
mos
t,
gp
p,
,in
con
tras
t to
the
proc
ess-
base
d pe
rspe
ctiv
e th
at is
the
trad
ition
al
hallm
ark
of s
imul
atio
n m
odel
ing.
•Pr
actic
alA
BMS
requ
ires
one
to:
•Pr
actic
al A
BMS
requ
ires
one
to:
1.id
entif
y th
e ag
ents
and
get
a th
eory
of a
gent
beh
avio
r, 2.
iden
tify
the
agen
t rel
atio
nshi
ps a
nd g
et a
theo
ry o
f age
nt
yg
pg
yg
inte
ract
ion,
3.
get t
he re
quis
ite a
gent
-rel
ated
dat
a,
4va
lidat
eth
eag
entb
ehav
ior
mod
els
inad
ditio
nto
the
mod
elas
a4.
valid
ate
the
agen
t beh
avio
r m
odel
s in
add
ition
to th
e m
odel
as
a w
hole
, and
5.
run
the
mod
el a
nd a
naly
ze th
e ou
tput
from
the
stan
dpoi
nt o
f lin
king
th
em
icro
scal
ebe
havi
ors
ofth
eag
ents
toth
em
acro
scal
ebe
havi
ors
the
mic
ro-s
cale
beh
avio
rs o
f the
age
nts
to th
e m
acro
scal
ebe
havi
ors
of th
e sy
stem
.
How
to d
o A
BMS:
not
yet
mat
ure
yfo
rmal
ism
s•
ABM
doe
s no
t as
of y
et h
ave
a m
atur
e se
t of s
tand
ard
form
alis
ms
or p
roce
dure
sfo
r m
odel
dev
elop
men
t and
i
hh
hfS
agen
t rep
rese
ntat
ion
such
as
thos
e th
at a
re p
art o
f Sys
tem
s D
ynam
ics
mod
elin
g.
•O
ther
than
the
impl
emen
ted
soft
war
eco
deth
ere
isno
•O
ther
than
the
impl
emen
ted
soft
war
e co
de, t
here
is n
o sc
hem
e fo
r un
ambi
guou
sly
repr
esen
ting
an a
gent
-bas
ed
mod
el.
•H
owev
er, a
gent
mod
elin
g do
cum
enta
tion
sche
mes
alo
ng
thes
e lin
es h
ave
rece
ntly
bee
n pr
opos
ed w
ith th
e in
tent
of
id
lf
bili
dd
ibili
prom
otin
g ag
ent m
odel
tran
sfer
abili
ty a
nd re
prod
ucib
ility
(G
rim
m e
t al.
2006
). •
Age
ntba
sed
mod
elin
gca
nbe
nefit
from
the
use
ofth
e•
Age
nt-b
ased
mod
elin
g ca
n be
nefit
from
the
use
of th
e U
nifie
d M
odel
ing
Lang
uage
(UM
L) fo
r re
pres
entin
g m
odel
s.
How
to d
o A
BMS:
UM
L ba
sed
visu
al
mod
elin
g•
UM
L is
a v
isua
l mod
elin
g la
ngua
ge fo
r rep
rese
ntin
g ob
ject
-ori
ente
d (O
-O) s
yste
ms
(Boo
ch, R
umba
ugh
et a
l. 19
98) t
hat i
s co
mm
only
ado
pted
to s
uppo
rt a
gent
-ba
sed
mod
els
in b
oth
the
desi
gn a
nd c
omm
unic
atio
n hph
ases
.•
UM
L co
nsis
ts o
f a n
umbe
r of
hig
h-st
ruct
ured
type
s of
di
agra
ms
and
grap
hica
lele
men
tsth
atar
eas
sem
bled
indi
agra
ms
and
grap
hica
l ele
men
ts th
at a
re a
ssem
bled
in
vari
ous
way
s to
repr
esen
t a m
odel
. •
The
UM
Lre
pres
enta
tion
isat
ahi
ghle
velo
f•
The
UM
L re
pres
enta
tion
is a
t a h
igh
leve
l of
abst
ract
ion,
inde
pend
ent o
f the
mod
el’s
im
plem
enta
tion
inth
epa
rtic
ular
O-O
prog
ram
min
gim
plem
enta
tion
in th
e pa
rtic
ular
OO
pro
gram
min
g la
ngua
ge u
sed.
How
to d
o A
BMS:
O-O
mod
elin
g pa
radi
gmg
pg
•M
ost l
arge
-sca
le a
gent
-bas
ed m
odel
ing
tool
kits
that
g
gg
prov
ide
basi
c ag
ent f
unct
iona
lity
are
base
d on
the
obje
ct
orie
nted
par
adig
m.
•A
gent
-bas
ed s
imul
atio
n is
not
the
sam
e as
obj
ect-
orie
nted
si
mul
atio
n, b
ut th
e O
-O m
odel
ing
para
digm
is a
use
ful
bi
ft
dli
it
bid
dba
sis
for
agen
t mod
elin
g, s
ince
an
agen
t can
be
cons
ider
ed
a se
lf-di
rect
ed o
bjec
t with
the
capa
bilit
y to
aut
onom
ousl
y ch
oose
actio
nsba
sed
onth
eag
ent’s
situ
atio
nch
oose
act
ions
bas
ed o
n th
e ag
ents
situ
atio
n.•
The
O-O
par
adig
m is
nat
ural
for
agen
t mod
elin
g, w
ith it
s us
eof
obje
ctcl
asse
sas
agen
ttem
plat
esan
dob
ject
use
of o
bjec
t cla
sses
as
agen
t tem
plat
es a
nd o
bjec
t m
etho
ds to
repr
esen
t age
nt b
ehav
iors
. O-O
mod
elin
g ta
kes
a da
ta-d
rive
n ra
ther
than
a p
roce
ss-d
rive
n p
pers
pect
ive.
•
One
way
to b
egin
the
mod
elin
g pr
oces
s is
to d
efin
e y
gg
pab
stra
ct d
ata
type
s an
d ob
ject
s.
Ht
dA
BMS
5l
tH
ow to
do
ABM
S: 5
gen
eral
ste
ps
1.A
gent
s: Id
entif
y th
e ag
ent t
ypes
and
oth
er o
bjec
ts
(cla
sses
) alo
ng w
ith th
eir
attr
ibut
es.
2.En
viro
nmen
t: D
efin
e th
e en
viro
nmen
t the
age
nts
will
live
in
and
inte
ract
with
.3
AM
hd
Sif
hh
db
hih
3.A
gent
Met
hods
: Spe
cify
the
met
hods
by
whi
ch a
gent
at
trib
utes
are
upd
ated
in re
spon
se to
eith
er a
gent
-to-
agen
tint
erac
tions
orag
enti
nter
actio
nsw
ithth
eag
ent i
nter
actio
ns o
r ag
ent i
nter
actio
ns w
ith th
e en
viro
nmen
t.4.
Age
nt In
tera
ctio
ns: A
dd th
e m
etho
ds th
at c
ontr
ol w
hich
g
agen
ts in
tera
ct, w
hen
they
inte
ract
, and
how
they
inte
ract
du
ring
the
sim
ulat
ion.
li
lh
dli
5.Im
plem
enta
tion:
Impl
emen
t the
age
nt m
odel
in
com
puta
tiona
l sof
twar
e.
How
to d
o A
BMS:
dis
cove
ring
age
nts
gg
•Id
entif
ying
age
nts,
acc
urat
ely
spec
ifyin
g th
eir
beha
vior
s, a
nd a
ppro
pria
tely
repr
esen
ting
agen
t in
tera
ctio
ns a
re th
e ke
ys to
dev
elop
ing
usef
ul a
gent
m
odel
sm
odel
s.
–A
gent
s ar
e ge
nera
lly th
e de
cisi
on-m
aker
s in
a s
yste
m.
Thes
ein
clud
etr
aditi
onal
deci
sion
-mak
ers
such
asTh
ese
incl
ude
trad
ition
al d
ecis
ion
mak
ers,
suc
h as
m
anag
ers,
as
wel
l as
nont
radi
tiona
l dec
isio
n-m
aker
s, s
uch
as c
ompl
ex c
ompu
ter s
yste
ms
that
hav
e th
eir o
wn
bh
ibe
havi
ors.
•O
ne n
eeds
a th
eory
of a
gent
beh
avio
r.ti
dli
hih
ttt
ttti
id
–no
rmat
ive
mod
el in
whi
ch a
gent
s at
tem
pt to
opt
imiz
e an
d us
e th
is m
odel
as
a st
artin
g po
int f
or d
evel
opin
g a
sim
pler
an
dm
ore
desc
ript
ive
heur
istic
mod
elof
beha
vior
.an
d m
ore
desc
ript
ive
heur
istic
mod
el o
f beh
avio
r. –
beha
vior
al m
odel
if a
pplic
able
beh
avio
ral t
heor
y is
av
aila
ble
(e.g
. con
sum
er s
hopp
ing
beha
vior
).
Ht
dA
BMS
How
to d
o A
BMS:
mor
e…
•D
isco
very
age
nts
–Id
entif
ying
age
nts,
acc
urat
ely
spec
ifyin
g th
eir
beha
vior
s, a
nd
appr
opri
atel
yre
pres
entin
gag
enti
nter
actio
nsap
prop
riat
ely
repr
esen
ting
agen
t int
erac
tions
–A
gent
s ar
e ge
nera
lly th
e de
cisi
on-m
aker
s in
a s
yste
m.
•tr
aditi
onal
dec
isio
n-m
aker
s, s
uch
as m
anag
ers
•no
ntra
ditio
nal d
ecis
ion-
mak
ers,
suc
h as
com
plex
com
pute
r sy
stem
s th
at h
ave
thei
r ow
n be
havi
ors
•A
BMS
life
cycl
ey
–D
evel
opin
g an
age
nt-b
ased
sim
ulat
ion
is p
art o
f the
mor
e ge
nera
l m
odel
sof
twar
e de
velo
pmen
t pro
cess
.–
Des
ktop
ABM
SD
eskt
op A
BMS
–La
rge-
scal
e A
BMS
•A
BMS
tool
kit
–Re
past
(Nor
th e
t al.
2006
), Sw
arm
(SD
G 2
006;
Min
aret
al.
1996
), N
etLo
go(N
etLo
go20
06) a
nd M
ASO
N (G
MU
200
6)
Otli
Out
line
•Pa
rt1:
An
intr
oduc
tion
toA
BMS
•Pa
rt 1
: An
intr
oduc
tion
to A
BMS
–M
otiv
atio
n–
Wha
tis
anag
ent
Wha
t is
an a
gent
–Th
e ne
ed fo
r A
BMS
–W
hy a
nd w
hen
ABM
S–
Back
grou
nd o
n A
BMS
•Pa
rt 2
:–
ABM
S ap
plic
atio
ns–
How
to d
o A
BMS
Pt3
•Pa
rt 3
:–
Elec
tric
ity m
arke
t, s
uppl
y ch
ain
exam
ple
ABM
Sin
Wor
kflo
ws
and
BPre
engi
neer
ing
–A
BMS
in W
orkf
low
s an
d BP
re-e
ngin
eeri
ng