Rep esen ing Complex Mul i–Agen
O ganisa ions in UML⋆
Joaquin Pe˜na, Ra ael Co chuelo and Miguel To o
Dp o. de Lenguajes y Sis emas In o m´a icos
A da. de la Reina Me cedes, s/n. Se illa 41.012 (Spain)
E–mail: [email p o ec ed], web page: www. dg-se ille.in o
Abs ac . In e ac ion has been p o ed one o he main sou ces o com-
plexi y in Mul i-Agen Sys ems (MAS) and many esea ches a e wo king
on echniques o pallia e i . Fu he mo e, O ganiza ion modelling ech-
niques lies on ep esen ing he g oups o agen s which a e ela ed by
some kind o in e ac ion. Cu en UML app oaches ep esen hese ela-
ionships as a se o bina y links usually ep esen ed as s e eo yped UML
associa ions no p o iding abs ac ion ools o manage he complexi y
de i ed om in e ac ions. In his pape , we a gue o mul ipa y links
in o de o inc ease he le el o abs ac ion o o ganiza ion models and
hus, hei abili y o manage complexi y.
keywo ds: Complex sys ems, o ganiza ion modelling, mul ipa y in e -
ac ions, agen p o ocol desc ip ions, UML.
1 In oduc ion
1.1 O ganiza ions and Complexi y
The o ganiza ional me apho has been p o ed one o he mos app op ia e ools
o enginee Mul i-Agen Sys ems (he ea e MAS) being used as he abs ac ion
which guides he analysis and design o MASs, e.g. [14]. O ganiza ion o Mul i-
Agen Sys ems is usually seen by many esea che s as a collec ion o in e ac ing
oles [8], e.g. GAIA [14] o AUML social s uc u es ep esen a ions [11]. An o -
ganiza ion shows he g oups o agen s o med in he sys em due o ge bene i s
om one o ano he in a collabo a i e o compe i i e manne . As a ma e o
ac , i shows ha an o ganiza ion eme ges when exis s some kind o in e ac-
ion be ween i s pa icipan s (ei he h ough di ec communica ion by means o
speech ac s o h ough he en i onmen ).
Mos esea che s ag ee on ha MASs a e a special kind o dis ibu ed sys ems
(objec s wi h hei own h eads o execu ion) wi h special ea u es whe e a highe
deg ee o complexi y exis s han in cu en Objec –Based So wa e Sys ems [7].
This complexi y o MASs is consequence o hei ea u es and mainly o hei
in e ac ing na u e: Complexi y is caused by he collec i e beha iou o many basic
in e ac ing agen s. James Odell [7].
⋆The wo k epo ed in his a icle was pa ially suppo ed by he Spanish Minis y
o Science and Technology unde g an s TIC97-0593-C05-05

✁
✂✂
✁✄ ☎✆ ✝✞✟✠
✝
✡☛✁✄ ☎✆ ✝✞✟✠
☞✌✍
✟
✎
✌✏ 
✞
✂
✁ ✆
✑
✡
✄✒
✓
✞✆✁
Pe o ms
cu ency ans e
✔
✆✁✄ ✕✖✁✟
✎

✁
✂✂
✁✄ ☎✆ ✝✞✟✠
✝
✡☛✁✄ ☎✆ ✝✞✟ ✠
☞✌✍
✟
✎
✌✏ 
✞
✂
✁ ✆
✔
✆✁✄ ✕✖✁✟
✎
O de s
om
O de
cu ency
ans e
Deli e s o
Fig. 1. Mul ipa y o ganiza ion ela ionships s. bipa y ela ionships
Thus, one o he main goals o Agen So wa e Enginee ing (AOSE) is ocused
on dealing wi h complexi y o MASs o which cu en so wa e enginee ing ech-
niques a e no specially ailo ed [5,6]. In his sense, Odell e al. [1], Woold idge
and Jennings in [14], and o he s [11] has iden i ied a se o echniques ha cope
wi h modelling such complex in e ac ion/social s uc u es.
1.2 The Need o Mo e Abs ac O ganiza ion Modelling A i ac s
In e ac ion modelling, and hus o ganiza ion modelling, encompasses wo as-
pec s: i) he s uc u al aspec : which models he ela ionships be ween a i-
ac s in he sys em om he in e ac ion poin o iew, e.g. as i is depic ed in
Figu e 1, a selle ’s bank agen s is linked wi h a buye ’s bank agen by means
o a ela ionship ha ep esen s ha hey mus in e ac o pe o m a money
ans e , and ii) he beha iou al aspec : which models he o de o appa i ion
o hese ela ionships o e ime, e.g. i s he i ems o be pu chased a e chosen
(o de s i ems) o la e pu chase hese i ems (o de s money ans e ), e ce e a.
No ice ha o he aspec s such as en i onmen modelling [11,14] and o ganiza-
ion ules [14] a e also c ucial aspec s which we do no cope wi h in his pape
since hey do no a ec di ec ly he complexi y o models.
In o de o be able o ace la ge MASs a se o abs ac ion mechanisms mus
be p oposed co e ing bo h aspec s. By he bes o ou knowledge wo echniques
can be used:
On he one hand, he use o ole modelling echniques p o ides a c ucial ool
o dealing wi h complexi y since i allows us o ”di ide and conque ” seeing
a complex o ganiza ion as a se o sepa a e sub–g oups which can be s udied
sepa a ely.
On he o he hand, mul ipa y ela ionships p o ides means o mo e abs ac
o ganiza ion s uc u e models han bipa y links such as associa ions. As i is
depic ed in Figu e 1, some ela ionships may in ol e mo e han wo agen s, e.g.
a pu chase whe e a buye ’s bank a selle ’s bank, a Use Agen and a Poin o
Sales Agen pa icipa es. I a concep ually a omic ela ionship a some le el o
abs ac ion is ep esen ed by means o bipa y ela ions we ha e o di ide i in o
a se o bipa y ela ions men ally hus dec easing he le el o abs ac ion. Fo
example, since in Figu e 1 he pu chase can be ep esen ed abs ac ly by a single
ou –pa y ela ionship i we limi o bipa y ela ionships we ha e o ep esen
i by ou ine g ain ela ionships. In la ge MAS his implies dec easing he le el
o abs ac ion o models om he beginning which in consequence dec ease ou
capaci y o dealing wi h complex sys ems.
Many esea che s has iden i ied his need p o iding in e ac ion abs ac ions
in o de o encapsula e he p o ocol pe o med be ween an a bi a y se o
agen s, e.g. AUML nes ed p o ocols, MESSAGE in e ac ions o GAIA p o o-
cols. Un o una ely, by he bes o ou knowledge he abs ac ions p oposed in
he li e a u e do no co e simul aneously bo h s uc u al and beha iou al as-
pec s using UML.
In his pape , we p esen pa o he no a ion o ou me hodology o model
complex Mul iagen Sys ems (h p://www. dg-se ille.in o/joaquinp/MaCMAS).
We p esen an in e ac ion abs ac ion based on p e ious wo k [1,2,12,13,14] o
g aphically ep esen bo h aspec s in UML 2.0 [9]. The main ad an age o ou
app oach is ha we p o ide he mechanisms needed o add a mo e abs ac
model o he o ganiza ion han o he p oposals ha use UML. Fu he mo e,
ou p oposal does no disable o he s bu i p o ides a mechanism o abs ac ion
which can be used o p oduce abs ac simple models which can be e ined o
each he le el o de ails applied in o he s.
This pape is o ganised as ollows: in Sec ion 2 we p esen he ela ed wo k;
in Sec ion 3 we p esen a mul ipa y abs ac ion called mul i-Role In e ac ion
o ep esen o ganiza ion ela ionships; in Sec ion 4 we p esen a case s udy; in
Sec ion 5 we discuss on he mos app op ia e UML no a ion o mRIs and we
illus a e i wi h he case s udy; in Sec ion 6 we p esen he UML models we
p opose. Finally, in Sec ion 7, we summa ise ou main con ibu ions.
2 Rela ed Wo k
The e exis s wo pa hs in he o ganiza ion modelling li e a u e [11]: i) he be-
ha iou is ic pe spec i e, and ii) he men alis ic pe spec i e. The o me sees
o ganiza ion as a collec ion o oles whose ea u es shows he ex e nal in e ace
ha agen s playing a ole o e o he g oup, hus ocusing on he mac o–le el
and equi ing o unde s and o ganiza ion (beha iou and s uc u e) in a p o-
g amma ic way. The la e desc ibes o ganiza ions in e ms o men al no ions
such us desi es, belie es, ac s, ules, e ce e a, hus ocusing on he agen mic o–
le el and equi ing o unde s and he o ganiza ion as a b anch o en i ies which
in conjunc ion o e s a join beha iou no desc ibed explici ly.
The on ie be ween bo h pe spec i es is placed be ween such sys ems ha
can be unde s ood p og amma ically and such whose complexi y makes impos-
sible o ea hem om a beha iou is ic pe spec i e1. This pape ocus on he
beha iou is ic app oach p o iding mechanisms o abs ac ion ha allows us o
1No ice ha a sha p sepa a ion can no be es ablished
mo e he on ie o beha iou is ic app oach o mo e complex sys ems using
UML.
i) S uc u al Aspec
In AUML [11] s uc u al ela ions be ween oles in an o ganiza ion a e ep e-
sen ed as bina y UML associa ions which o ce designe s o decompose men ally
mul ipa y ela ions. We hink ha UML associa ion is no he mos app op ia e
UML a i ac o ep esen in e ac ion ela ionships since hey ha e been adi-
ionally used o ep esen in o ma ion ela ionships which may lead designe s o
see hese ela ionships as ype ela ions. No ice ha n–a y associa ions may be
used allowing ep esen ing ela ionships mo e abs ac ly. Un o una ely, hey a e
no usually used in cu en app oaches and hey also p esen he same seman ic
d awbacks.
Al hough AUML ecognises he need o mul ipa y in e ac ions which hey
called nes ed p o ocols, hese mul ipa y in e ac ions be ween oles a e no used
o ep esen s uc u al aspec s in hei o ganiza ion models [11]. GAIA also
ecognises he need o mul ipa y in e ac ions which hey called p o ocols. Un-
o una ely, hey do no p o ide an UML no a ion. In MESSAGE O ganiza-
ion models acquain ance ela ions o ep esen o ganiza ion s uc u e. Un o -
una ely, hese ela ions a e based on s e eo yped associa ions igno ing o he
na i e UML cons uc ions ha ep esen he same concep s and which i be e
seman ically wi h in e ac ion ela ionships.
MESSAGE also p o ides he concep o in e ac ion which may be also used
o ep esen o ganiza ion ela ionships. Un o una ely, au ho s do no show he
ela ion be ween acquain ance ela ions and in e ac ions. They show nei he he
UML cons uc ion on which in e ac ions a e based. Finally, MESSAGE is based
on UML 1.3 which did no ep esen oles p ope ly [4].
ii) Beha io al Aspec
Using mul ipa y links o ep esen o ganiza ion s uc u e equi es o ai-
lo ed ools ha allows us o ep esen s he sequences o execu ion o such abs ac
join asks.
GAIA ep esen s such o de by means o egula exp essions assigned o each
ole in an o ganiza ion based on FUSION no a ion [3]. Fo example, i a Role
A pa icipa es in h ee join asks namely I1,I2and I3wi h oles B and C in
all o hem, he exp ession A=I1I2I3shows ha ole A pa icipa es i s in
he p o ocol (join ask) I1, o la e pa icipa e in I2, and o inalize wi h I3.
Un o una ely, hey do no p o ide an UML–based g aphical no a ions no a way
o ep esen ing he whole beha iou o an o ganiza ion using a single model.
Finally, MESSAGE does no ep esen he sequences o in e ac ions o ac-
quain ance ela ions. Al hough MESSAGES wo k lows ep esen he sequence
o asks au ho s do no desc ibe how in e ac ions, acquain ance ela ions and
wo k low ela e.
3 A Fi s Class Abs ac ion o Model O ganiza ions
An mRI is an ins i u ionalised pa e n o in e ac ion ha we p opose as abs ac-
ion ool o ep esen mul ipa y ela ionships be ween an a bi a y numbe o
oles. An mRI is he ma e ialisa ion o a ce ain o ganiza ion goal equi ed in
he sys em a he analysis s age, i.e. mRIs ep esen s mul ipa y in e ac ions
ha se e al agen s playing he oles de ined on i ha e o pe o m o achie e an
o ganiza ion goal.
The in o ma ion ep esen ed by an mRI ocuses on he na u e o he join
p ocess and no on how i is ca ied ou .
They a e he co ne s one o ou app oach since o ganiza ion is always de-
sc ibed by means o his abs ac ion. Using mRI as he minimum modelling
elemen , we do no ha e o ake in o accoun all he links equi ed by a com-
plex ask (s uc u al aspec ) no he messages ha a e exchanged o accomplish
i (beha iou al aspec ) a s ages whe e hese de ails ha e no been iden i ied
clea ly o a e no e en known. Ob iously, when he le el o de ail o mRIs has
been inc eased enough o clea ly unde s and he sys em o ganiza ion, bipa y
links o he s uc u al aspec and messages desc ip ions o he beha iou al as-
pec a e he mos adequa e app oach o de ine mRIs in e nally such as hose
p oposed in AUML.
4 Case S udy: The UN Secu i y Council’s P ocedu e o
Issue Resolu ions
The case s udy we use o illus a e ou app oach is a simpli ied e sion o he
Modelling TC UN Secu i y Council’s P ocedu e o Issue Resolu ions case s udy2.
In www. dg-se ille.in o/joaquinp/MaCMAS/example is a ailable he comple e
model o he case s udy using ou no a ion.
To pass a UN-SC esolu ion, he ollowing p ocedu e would be ollowed: 1)
A leas one membe o UN-SC submi s a p oposal o he cu en Chai ; 2)
The Chai dis ibu es he p oposal o all membe s o UN-SC and se a da e
o a o e on he p oposal; 3) A a gi en da e ha he Chai se , a o e om
he membe s is made; 4) Each membe o he secu i y council can o e ei he
FOR o AGAINST o SUSTAIN; 5) The p oposal becomes a UN-SC esolu ion,
i he majo i y o he membe s o ed FOR, and no pe manen membe o ed
AGAINST; 6) The membe s o e one a a ime; 7) The Chai calls he o de
o o e, and i is always he las one o o e; 8) The o e is open (in o he
wo ds, when one o es, all he o he membe s know he o e); 9) The p oposing
membe (s) can wi hd aw he p oposal be o e he o e s a s and in ha case no
o e on he p oposal will ake place; 10) All ep esen a i es o e on he same
day, one a e ano he ; 11) A o e is always inished in one day. The da e o he
o e is se by he chai .
2h p://www.auml.o g/auml/documen s/UN-Case-S udy-030322.doc

5 UML No a ion
When an ex ension mus be de ined, OMG s ongly ecommends o base i on
he mos seman ically nea cons uc ion in o de o a oid seman ic mis akes o
edundan language ex ensions [10, pag. 3-26]. In his sense, we ha e ca e ully
s udied UML o use he mos app op ia e modelling a i ac o ep esen mRIs.
In his wo k, we ha e based on he las e sion o UML: UML 2.0 [9]. The
new ea u es ha i p esen s i s be e wi h ou pu pose han p e ious e sions
whe e oles we e no p ope ly suppo ed [4].
We ha e de e mined ha om he dynamic modelling a i ac s p o ided by
UML 2.0, collabo a ions p esen s a qui e simila seman ic o mRIs. In [9] he
OMG p o ides he ollowing summa y o collabo a ions 3:
UML 2.0 [9, pag. 125]: A beha iou o a collabo a ion will e en ually
be exhibi ed by a se o coope a ing ins ances (speci ied by classi ie s)
ha communica e wi h each o he by sending signals o in oking ope a-
ions. Howe e , o unde s and he mechanisms used in a design, i may
be impo an o desc ibe only hose aspec s o hese classi ie s and hei
in e ac ions ha a e in ol ed in accomplishing a ask o a ela ed se
o asks, p ojec ed om hese classi ie s. Collabo a ions allow us o de-
sc ibe only he ele an aspec s o he coope a ion o a se o ins ances
by iden i ying he speci ic oles ha he ins ances will play. In e aces
allow he ex e nally obse able p ope ies o an ins ance o be speci ied
wi hou de e mining he classi ie ha will e en ually be used o speci y
his ins ance. Consequen ially, he oles in a collabo a ion will o en be
yped by in e aces and will hen p esc ibe p ope ies ha he pa icipa -
ing ins ances mus exhibi , bu will no de e mine wha class will speci y
he pa icipa ing ins ances.
As can be seen he seman ics o collabo a ions i p ope ly wi h ou pu pose
since hey ep esen mul ipa y in e ac ions using oles o abs ac ly desc ibe
he ea u es o agen s ha will play hem.
UML p o ides wo no a ions o collabo a ions: he in e nal s uc u e no a-
ion and he composi e s uc u e no a ion. The o me shows he in e nals o he
collabo a ion ep esen ing oles and hei communica ion pa hs using bipa y
links which is no ou pu pose (simila o [11]). The la e shows he collabo a-
ion using a collabo a ion icon: a dashed ellipse which he name o he collabo-
a ion inside wi hou de ailing how i is ca ied ou . Roles o he collabo a ion
a e shown as associa ions (Collabo a ionRole), e.g in Figu e 2 we ep esen he
o ganiza ion o med o o e a p oposal whe e h ee oles pa icipa es: Chai ,
Vo e and Obse e . A he end o Collabo a ionRoles we place he in e ace e-
qui ed by each ole showing he ex e nal ea u es ha an agen playing a ce ain
ole may expose o he o ganiza ion [9, pag. 131]. UML in e aces may con ain
se ices and a ibu es hus we place he e he knowledge p ocessed by each ole
and he se ices o e ed.
3No ice ha he collabo a ion no a ion and seman ics has change om p e ious UML
e sions
✗
✘
✙✚✛
✜
✚✛
✢
✣
✤
✥
✛
✦
✧
★
✚✛
✩✩
✪
✫
✬
✭✮
✯
✰✱✭
✲✲
✪
✳
✴
✵✭✮
✶
✭✮
✷ ✸
✥
✙
★
✗
✹
✦
✧
★
✚✙
✷
✩✩
✪
✫
✬
✭✮
✯
✰✱✭
✲✲
✪
✺
✻
✰
✼
✮
✷
✦
✧
★
✚
✷ ✸
✥
✙
★
✗
✹
✦
✧
★
✚✙
✷ ✸
✥
✙
★
✗
✹
✽
✚✾
✘
✚✛✙
✷
✩✩
✪
✫
✬
✭✮
✯
✰✱✭
✲✲
✪✿
❀
✬
✭✮
✷ ❁
✛✧❂✧✙✤
❃
✷
✽
✚✾
✘
✚✛
✷
✦
✧
★
✚
✷
❄
❅❆
❇
❈ ❉
❅
❇
❊❋
●
❉
❅
❇
❊ ❍■
❏
❑
▲❍❋
●
▼
❍❆
❇
◆
❖
❉
❅
❇
❊❆ ▲■
P
❉
❅
❇
❊❋
●
❉
❅
❇
❊ ❍■
◆
◗
❆❊❋
❘
❊❋
●
▼
❍❆
❇
◆
❖
❉
❅
❇
❊❆
❙
❚▲❋
P
❈ ❉
❅
❇
❊❋
●
❯❊❯
◗
❊❋
■❅
❇
❍■
❏
❑
▲❍❋
●
▼
❍❆
❇
❅
❖
❉
❅
❇
❊❆
❱
❲❲
❳
Vo e
❨
❀✰
❃
❩ ❬
★
❂✚✛
✹
✧✛✾✙ ✤
✙
✥
❳❭
❃
✚✜
✧
★
✚
✥
❳
✹
✧✛✾
✥
❳❭
✧
✘
✙✚✛
✜
✚✛✙
Fig. 2. mRI Vo e
❪
❫❴
❵
❛❜❝❫
❞
❡
❢
❣
❢
❤
❣
✐❥
❦
❧
♠♥
♦ ♣
q
s
✉
✈✇
♦ ①
②
③
✉
④
⑤
⑥
⑦q⑧q③⑨
⑩
❶❞❷❸
❹
✐❡
❣
❥❤❺
❣
❻
♥
❣
❼
✐❤
❽ ❾
⑧⑧q
②
❿
✉
✇✈❿
✉
③➀
➁
✈
s
➂
②
❿➃⑨❿
s
❿✈➃q
✉
②
⑨
✉
②
q❿⑨
➄
q➂
✉
③q✇✈➃qq
s
➅
➁
②
➀
➁
③
➁
q➂
s➄
✈
③q
s
❷
❤❥
❣
❢
❺
❢
♠❤❡
❣
❦
❧
♠♥
♦ ♣
q
s
✉
✈✇
♦
⑥
⑦q⑧q③⑨
⑩
❪
➆❫
➇
➈
❜
➈
➉➆➊
➇
❿
②
✉
②
⑨
✉
q⑦
➋
➌
⑧✈
➍
⑥
⑨⑦
✉
②
➀
②
⑧⑨❿
✉
➋
➌
⑧✈
➍
♣
q
s
✉
✈✇
➍
⑥
⑦q⑧q③⑨
➎
➏➏
❿
Fig. 3. Pa ame ised mRI o FIPA Con ac Ne P o ocol
Al hough UML 2.0 collabo a ions do no de ine any a ibu e we ha e added
he goal o he collabo a ion using a ex ual desc ip ion. In Figu e 2 his a ibu e
can be obse ed inside he collabo a ion icon in a compa men wi h he name
Goal. Fu he mo e, in o de o ep esen he ini ia o o he mRI we ep esen
hem wi h an a ow om he in e ace o he collabo a ion, in ou example he
ole Vo e .
Each ole ha pa icipa es on an mRI can be deco a ed wi h a gua d in o de
o indica e when i is in e es ed on pa icipa ing in i . Gua ds a e g aphically
ep esen ed as ex ual no es linked wi h he associa ion Collabo a ionRole. Fo
example, he Chai will only pa icipa e in he mRI Vo e i and only i he o e
has no o ed p e iously: V o e .V o e 6∈ Chai .Lis O V o es.
Finally, some in e ac ions pa e ns can be gene alised in o de o euse hem.
Pa ame e ised mRIs a e ep esen ed as pa ame e ized UML collabo a ions adding
a compa men o show he pa ame e s o he mRI. I con ains he conc e e oles
ha pa icipa e in he mRI and he conc e e knowledge ha is managed. Fo
example, in Figu e 3, we show a pa ame e ised mRI o he FIPA Con ac Ne
P o ocol whe e we can see ha he Type o he P oduce Role and Consume
ole a e open and also he knowledge ha is exchanged. As his pa e n should
be well known o be eused, i can be a ached wi h a FIPA pa ame e ised p o-
ocol desc ip ion and a code amewo k ha allows us o implemen i almos
di ec ly.
6 UML Rep esen a ion o O ganiza ions
In ou app oach we ep esen an o ganiza ion om wo pe spec i es: i) he O -
ganiza ion s uc u e model which shows s a ically all he ela ionships ha may
appea in o an o ganiza ion by means o UML collabo a ions and ii) he o ga-
niza ion beha iou model which ep esen he o de o appa i ions o hese links
o e ime. In ollowings sec ions we de ail bo h models.
No ice ha ou app oach mus be applied o complex sys ems a he ea ly
s ages o modelling in o de o cla i y he o ganiza ion o he sys ems hanks o
abs ac models we p o ide. AS a ma e o ac , we do no co e how oles map
in o agen s since his alls in he scope o ine g ain echniques.
6.1 O ganiza ion S uc u e Model: Role Models
A ole model ep esen s an o ganiza ion s uc u e as a se o oles ha ela es by
means o mRIs. They ep esen a pa ial iew o he whole o ganiza ion o he
sys em whe e a ce ain o ganiza ion goal o he MAS is ep esen ed o hogonally
o he es o hem.
As we can see i s de ini ion is simila o mRI bu he main di e ence is he
le el o de ail ha each concep ep esen . While an mRI ep esen a goal as a
whole, a ole model ep esen s he goal as a se o mRIs, hus gi ing a de ailed
de ini ion o he o ganiza ion s uc u e. Thus, when se e al mRIs a e used o
desc ibe he same goal, each o hem ep esen s sub–goals o he gene al one.
The e may exis a di ec mapping be ween o igina ion s uc u es ep esen ed
by means o a single mRI and ole models which de ail i . The e inemen ech-
niques p esen ed in [12] cons i u es a way o iden i ying he ole model which
ep esen s an mRI in e nally.
Figu e 4 shows he ole model o he issue esolu ion o ganisa ional goal o
ou case o s udy whe e we can iden i y se e al mRIs: Accep /Rejec p oposal,
Submi p oposal, Vo e, and wi hd aw p oposal which model abs ac ly he whole
case o s udy. No ice ha , since in his ole model each ole is used by se e al
mRIs se e al nes ed in e aces can be iden i ied: one o each mRI linked o he
main in e ace, e.g. he in e ace equi ed by he ole Chai o he Vo e mRI i
is nes ed in he in e ace IChai .
6.2 O ganiza ion Beha io Model: S a eMachines and
P o ocolS a eMachines
The beha iou al aspec o an o ganiza ion, ha i is o say, how he mRIs in
a ole model sequence, can be ep esen ed in wo ways: a single dynamic iew
based on UML 2.0 S a e Machines [9, pag. 446] which ep esen s he o de o
mRIs in he ole model and a se UML 2.0 P o ocolS a eMachines [9, pag. 422],
➐➑
➒➓➔
→
➓➔
➣↔
↕
➙
➔
➛
➜
➑
➝
➙
➞➞
➓➔
➟
➠
➡
➢
➤
➥
➦ ➧
➙
➒
➞
➐➨
➩
➔➫➭➫➒↕
➯
➦ ➲
↕
➞
➓
➦ ➳
➫
➞
➓
➦ ➧
➙
➒
➞
➐➨
➳
➫
➞
➓➒
➦ ➧
➙
➒
➞
➐➨
➵
➓➝
➑
➓➔➒
➦ ➧
➙
➒
➞
➐➨
➩
➔➫➭➫➒↕
➯
➒
➦
➟
➸
➺
➻
➼
➤
➽
➾➥
➦
➩
➔➫➭➫➒↕
➯
➦
➵
➓➝
➑
➓➔
➦ ➲
↕
➞
➓
➦
➸
➺
➻
➼
➤
➽
➚
➥➪➶➪➹➢
➘
➴
➪➢
➯
➷
➛
➜
➑
➝
➙
➞
↕ ➬➓
➮
➭➔➫➭➫➒↕
➯ ➨
➫➔ ↕
➔➓➒➫
➯
➜
➞
➙
➫➬
➐➑
➒➓➔
→
➓➔
➣↔
↕
➙
➔
➳
➫
➞
➓➔
➟
➱
➻
➹➾➥
✃
➾➥
➦ ➧
➙
➒
➞
➐➨
➳
➫
➞
➓➒
➦
➩
➔➫➭➫➒↕
➯
➦
➟❐
➪
➽
➾➥
➦
➩
➔➫➭➫➒↕
➯
➦
➵
➓➝
➑
➓➔
➦ ➳
➫
➞
➓
➦
❐
➪
➽
➾
➴
➪➢
➯
➷
➵
↕➬↕❒➓
➞
↔
➓
→
➫
➞
➙
➬❒ ➭➔➫❮➓➒➒ ➫
➨
↕
➭➔➫➭➫➒↕
➯
➐➑
➒➓➔
→
➓➔
➣↔
↕
➙
➔
➛
➜
➑
➝
➙
➞➞
➓➔
❰
➤
➽
➡Ï
➥➢Ð
➚
➥➪➶➪➹➢
➘
➴
➪➢
➯
➷ Ñ
➑
➫➔
➞
↕ ➭➔➫➭➫➒↕
➯
➒➜
➑
➝
➙
➒➒
➙
➫➬
Ò
Ó➾➶
➽
Ô
Õ
➾
Ö
➾Ó
➽
➚
➥➪➶➪➹➢
➘
➴
➪➢
➯
➷ Ñ
❮❮➓➭
➞
➫➔ ➔➓
×
➓❮
➞
↕
➭➔➫➭➫➒↕
➯
➣↔
↕
➙
➔
➐➑
➒➓➔
→
➓➔ Ø
ÙÙ
➬
Ø
ÙÙ
➬
Ø
ÙÙ
➬
Ø
ÙÙ
➬
Fig. 4. Role Model Issue Resolu ion
one o each ole, which ep esen s sepa a ely he beha iou o each ole in he
ole model. In bo h cases, ansi ions a e used o ep esen mRIs execu ion. UML
de ines hem as ollows:
S a eMachine: S a e machines can be used o exp ess he beha io o pa o
a sys em. Beha io is modeled as a a e sal o a g aph o s a e nodes in-
e connec ed by one o mo e joined ansi ion a cs ha a e igge ed by he
dispa ching o se ies o e en s. Du ing his a e sal, he s a e machine ex-
ecu es a se ies o ac i i ies associa ed wi h a ious elemen s o he s a e
machine.
P o ocolS a eMachine: A p o ocol s a e machine is always de ined in he
con ex o a classi ie . I speci ies which ope a ions o he classi ie can
be called in which s a e and unde which condi ion, hus speci ying he al-
lowed call sequences on he classi ie ’s ope a ions. A p o ocol s a e machine
p esen s he possible and pe mi ed ansi ions on he ins ances o i s con ex
classi ie , oge he wi h he ope a ions which ca y he ansi ions. In his
manne , an ins ance li ecycle can be c ea ed o a classi ie , by speci ying he
o de in which he ope a ions can be ac i a ed and he s a es h ough which
an ins ance p og esses du ing i s exis ence.