A Conceptual Framework for Automated Negotiation Systems

A Concep ual F amewo k o Au oma ed Nego ia ion Sys ems Manuel Resinas, Pablo Fe nandez, and Ra ael Co chuelo ETS Ingenie ia In o ma ica Uni e sidad de Se illa, Spain h p://www. dg-se ille.in o Abs ac . In he las yea s, much wo k ha e been done in he de elop- men o echniques o au oma ed nego ia ion, and, pa icula ly, in he au oma ed nego ia ion o SLAs. Howe e , he e is no wo k ha desc ibes how o de elop ad anced so wa e sys ems ha a e able o nego ia e au- oma ically in an open en i onmen such as he In e ne . In his wo k, we de elop a concep ual amewo k o au oma ed nego ia ions o SLAs ha se es as a s a ing poin by iden i ying he elemen s ha mus be sup- po ed in hose so wa e sys ems. In addi ion, based on ha concep ual amewo k, we epo on a se o p ope ies o au oma ed nego ia ion sys ems ha may be used o compa e diffe en p oposals. 1 In oduc ion In his wo k, we ocus on SLA nego ia ions1. The goal o his kind o nego ia ions is o each an ag eemen be ween a se ice p o ide and a se ice consume abou he e ms and gua an ees o he se ice consump ion. We a e in e es ed in de eloping so wa e sys ems ha a e able o nego ia e in open en i onmen s, such as he In e ne , wi h empo al es ic ions. Tha is, he ag eemen has o be eached in a limi ed amoun o ime. This scena io defines he cha ac e is ics o he nego ia ion ha shall be ca ied ou : i is a non-coope a i e nego ia ion be ween wo o mo e pa ies wi h pa ial in o ma ion o he wo ld and ha d compu a ional cons ain s. In he las yea s, much wo k ha e been done on au oma ed nego ia ions. These wo ks a e ocused on he de elopmen o new decision-making algo i hms o he cons uc ion o new p o ocols ha p esen s ce ain desi able cha ac e is ics o au oma ed nego ia ions. Howe e , much less a en ion has been paid o he so - wa e a e ac s ha a e necessa y o ca y ou his au oma ed nego ia ion. In his wo k, we analyse he p oblem o au oma ed nego ia ion o se ice le el ag ee- men s om a so wa e enginee ing pe spec i e. Specifically, we wan o iden i y he elemen s ha a e equi ed o build a sys em ha de elops an au oma ed nego ia ion o SLAs. To each his goal, we p esen a no el concep ual amewo k o SLA au oma ed nego ia ions. Unlike o he wo ks, we ake a so wa e enginee ing pe spec i e and 1 SLA nego ia ions a e equi alen o he so-called se ice-o ien ed nego ia ions [1]. cen e on he elemen s ha a e equi ed o build an au oma ed nego ia ion sys- em. This concep ual amewo k se les he bases o a la e analysis o he diffe en nego ia ion p oposals and defines a common ocabula y o au oma ed nego ia ion sys ems. In addi ion, we ob ain a se o p ope ies o au oma ed nego ia ion sys ems as a consequence o he concep ual amewo k. These p op- e ies may be used o e alua e and compa e diffe en p oposals on au oma ed nego ia ion o SLAs. This pape is s uc u ed as ollows: fi s , we desc ibe he concep ual amewo k in Sec ion 2, hen, in Sec ion 3, we use he concep ual amewo k o ob ain a se o p ope ies o au oma ed nego ia ion sys ems. Finally, we p esen ou conclusions and u u e wo k in Sec ion 4. 2 A Concep ual F amewo k o Au oma ed Nego ia ions T adi ionally, an au oma ed nego ia ion sys em has been cha ac e ised by h ee elemen s [2]: p o ocol, nego ia ion objec and decision-making model. In ou concep ual amewo k we p opose o ex end his cha ac e isa ion wi h wo addi- ional elemen s: in o ma ion and p e e ences. Addi ionally, we de ail mo e p e- cisely he decision-making and p o ocol elemen s, and we also p o ide a conc e e desc ip ion o he nego ia ion objec . 2.1 Nego ia ion Objec In a SLA nego ia ion, he objec ha is being nego ia ed is an ag eemen be- ween pa ies. An ag eemen defines a dynamically-es ablished and dynamically- managed ela ionship be ween pa ies [3]. The goal o he ag eemen is o es ablish he gua an ees ha mus be obse ed du ing he execu ion o a se ice. An ag eemen is composed a leas by he ollowing: –A specifica ion o he pa ies in ol ed in i . In p inciple, he numbe o pa ies in ol ed in an ag eemen is no cons ained. Howe e , he mos common case is wo-pa y ag eemen s. –A collec ion o e ms ha desc ibes bo h unc ional desc ip ions and non- unc ional gua an ees o he se ice. Addi ionally a e m can also exp ess o he aspec s o an ag eemen such as e mina ion clauses. A e m is com- posed o h ee pa s: •The coun e pa y whom he e m is applied o. Each e m is o be applied o one o he pa ies in ol ed in he ag eemen and he pa y is obliga ed o ulfil wha i is specified in i . Ob iously, he coun e pa y mus be one o hose ha ha e been designa ed in he ag eemen as one o he pa ies ha a e in ol ed in i . •Ase o cons ain s o speci y unc ional o non- unc ional desc ip ions o gua an ees o he se ice. I is expec ed ha he con en o hese cons ain s will be e y b oad and domain-specific. •Ase o compensa ions ha will be applied in case he pa y does no ulfil he cons ain s specified in he e m. This elemen is op ional and i is no suppo ed by he majo i y o he nego ia ion s a egies. Nego ia ion objec Wo ld Model P e e ences P o ocol Decision-making Rules Pe o ma i es In o ma ion exchanged Binding decision Decommi decision Response gene a ion Subjec i i y Mechanisms Subjec Assessmen mechanisms S a emen s Se ice s a emen s Pa y s a emen s Nego ia ion s a emen s Pa ies Te ms Fig. 1. Concep ual amewo k 2.2 P e e ences The ag eemen p e e ences exp ess he da a ha is used o assu e ha he use needs a e co ec ly deal among he nego ia ion p ocess. These p e e ences com- p ise: a se o s a emen s exp essing he ea u es and equi emen s o he use , and an assessmen mechanism o e alua e and compa e ag eemen p oposals. The mos common way o e alua ing p oposals is h ough he defini ion o u il- i y unc ions [4]. S a emen s can be classified depending on he domain in which hey a e applied: – Se ice s a emen s. These s a emen s a e applied o he se ice offe ed (o demanded). They can e e o ei he unc ional o non- unc ional cha - ac e is ics o he se ice such as he se ice in e ace o he se ice p ice. – Pa y s a emen s. In his case, an exp ession abou he pa y is s a ed. These s a emen s can exp ess ei he ea u es o equi emen o e a gi en pa y. Examples o his can be: Pa y Z is loca ed in I an o Pa y X mus ha e a high epu a ion on se ice Y. – Nego ia ion s a emen s. They speci y ea u es abou he nego ia ion p o- cess i sel , such as he nego ia ion deadline. Each s a emen is linked o a se o languages ha gi e seman ics o he ocabula y used wi hin he s a emen . On ologies can be seen as an example o languages desc ibing he ela ionship amongs concep s o a gi en seman ic domain. Usually, he s a emen s a e exp essed using wo diffe en o malisms: ules and cons ain s. Howe e , o he o malisms could be used. 2.3 Wo ld Model While nego ia ing, he mo e in o ma ion we ha e abou o he pa ies, he be e ou pe o mance is [5]. Fu he mo e, i has been shown ha aking he condi- ions o he ma ke in o accoun does imp o e he ou come o nego ia ions wi h se e al simul aneous opponen s [6]. The e o e, i is essen ial o an au oma ed nego ia ion sys em o build a model o he coun e pa ies and he ma ke . We call his model, he wo ld model o an au oma ed nego ia ion sys em and i may co e h ee diffe en domains: –Coun e pa ies: including he cha ac e is ics o he se ice demanded o o - e ed, he nego ia ion p ocess ollowed by he coun e pa y and he coun e - pa y i sel (e.g. epu a ion o geog aphical loca ion). –Ma ke : o ins ance, he ma ke ese a ion p ice, o he p obabili y o ap- pea ing ou side op ions du ing he nego ia ion [7]. –Se ice-domain: such as knowledge abou he ocabula y used in he e ms o he ag eemen . The in o ma ion used o build he wo ld model may be ei he objec i e o subjec i e: objec i e in o ma ion ypically includes he public ea u es abou he se ice demanded/supplied, bu i may also include in o ma ion abou he coun e pa ies hemsel es (e.g. hei geog aphical loca ion) and he ma ke . Sub- jec i e in o ma ion comp ises elemen s such as he epu a ion o a coun e pa y o he ma ke p ice o a ce ain se ice. Finally, he e a e h ee diffe en mechanisms o ob ain he in o ma ion ha is used o build he wo ld model: (i) Di ec ly polling he po en ial coun e pa y. In his case, he sys em mus implemen a compa ible specifica ion o a o ma o exp ess unc ional and non- unc ional ea u es o se ices and a p ocedu e o que y and o inspec se ices. (ii) Que ying a hi d pa y en i y o ob ain in o ma ion ela ed o a specific coun e pa y. Fo ins ance, o ob ain in o ma- ion abou i s epu a ion o i s geog aphical loca ion. (iii) Analysing p e ious in e ac ions wi h a po en ial coun e pa y. The esul s o he analysis may be s o ed in o de o be used la e while making decision abou p oposals ela ed o he po en ial coun e pa y. We en ision ha he fi s p ocedu e shall be commonly used in ga he ing objec i e in o ma ion abou he coun e pa ies, while he second and hi d p o- cedu es shall be mo e common in ob aining subjec i e in o ma ion abou he ma ke and he po en ial coun e pa y i sel . 2.4 P o ocol The nego ia ion p o ocol es ablishes he ules ha go e n he nego ia ion and he way he communica ion be ween he diffe en pa ies in ol ed in he nego ia ion is ca ied ou as well as he in o ma ion exchanged be ween he pa ies. We dis inguish h ee diffe en , al hough s ongly ela ed, aspec s in a nego ia ion p o ocol: ules, pe o ma i es and in o ma ion exchanged. Pe o ma i es. A pe o ma i e is he exp ession o he in en ion o he sende o a message. The se o pe o ma i es used in a nego ia ion p o ocol may di - e significan ly. Howe e , he e a e a minimum subse o pe o ma i es ha a e common o he majo i y o nego ia ion p o ocols. Namely, accep (accep a p oposal), ejec p oposal (wi hd aws he p oposal), ejec nego ia ion (cancel he whole nego ia ion p ocess), and commi 2(commi o a gi en p oposal). De- pending on he specific p o ocol ha is being used, o he pe o ma i es may be necessa y. Fo ins ance, in auc ions, i is common o use he in o m pe o ma i e o no i y e en s occu ed du ing he nego ia ion such as ha a new bid has been done. In p o ocols ha diffe en ia es be ween binding and non-binding p opos- als, he p opose (make a non-binding p oposal) pe o ma i e may be in oduced. O he p o ocols use a o e sys em o decide which is he p e e ed offe [8]. In such p o ocols, a o e pe o ma i e may be used. Finally, in nego ia ion p o- ocols ha use a gumen a ion [9], o he pe o ma i es o in oduce a gumen s suppo ing ou p oposal a e used such as a gue and challenge [9]. Rules. In a nego ia ion p o ocol, he e a e usually some es ic ions ega ding a a ie y o aspec s o he nego ia ion such as how he p oposals mus be buil , when a pa icipan can pos a p oposal, which pe o ma i e can be used in each momen , when a pa icipan may join o he nego ia ion, o when he nego ia ion can finish. In [10], a axonomy o ules o nego ia ion p o ocols is p esen ed. Howe e , al hough i is a ho ough axonomy and he majo i y o he aspec s o au oma ed nego ia ion a e co e ed, we belie e ha i should be ex ended wi h ules o decommi ing om p e iously c ea ed ag eemen s. We a gue ha he e mus be ules ha explici ly speci y whe he a decommi men may ake place, and, i so, when i may occu and which a e he penal ies o be paid as a compensa ion. In o ma ion Exchanged. The hi d aspec in a nego ia ion p o ocol is he ype o in o ma ion exchanged amongs he pa icipan s in he nego ia ion. A a ie y o app oaches has been p esen ed in he li e a u e. Those app oaches may be classified in o h ee b oad g oups: –The in o ma ion exchanged explici ly s a es he pa s o he ag eemen ha a e disliked by he pa y as well as he p oposed changes. –The in o ma ion exchanged consis s only o p oposals. In o he wo ds, he nego ia ion p o ocol is p oposal-based. The ad an age o his app oach is ha i un eils less in o ma ion o he o he pa ies. The disad an age is ha he lack o explici in o ma ion implies a blind sea ch o a mu ually accep able ag eemen ha may lead o longe nego ia ions and e en o no o find any ag eemen a all. –The in o ma ion exchanged includes p oposals, as in p oposal-based p o o- cols, and s a emen s ha a e used o pe suade o con ince he opponen o 2In he li e a u e, his pe o ma i e is usually called p opose meaning making a bind- ing p oposal. Howe e , we p e e o lea e he e m p opose o non-binding p oposals and o use commi o binding p oposals. accep ou p oposal [11]. This app oach is called a gumen a ion and i is a p omising field ha may e en ually o e come he d awbacks o he p oposal- based nego ia ion [9]. Howe e , he nego ia o s ha suppo a gumen a ion end o be e y complex and no a gumen a ion app oach has been applied o a eal scena io ye . 2.5 Decision-Making The decision-making model de e mines he way a pa y beha es while in ol ed in a nego ia ion p ocess. Th ee elemen s o m pa o he decision-making model o an au oma ed nego ia ion sys em: he decision o wha is conside ed an ac- cep able ag eemen and whe he o commi o i , he cons uc ion o esponses o his in o ma ion, and he decision o decommi om a p e iously es ablished ag eemen i possible. Binding Decision. This decision includes de e mining when a binding p oposal mus be submi ed and whe he a binding p oposal ha has been ecei ed should be accep ed. The binding decision depends on se e al ac o s ha may a y depending on whe he i is a se ice consume who is making he decision o i is a se ice p o ide . Ne e heless, we can di ide hese ac o s in o h ee b oad g oups: –P e e ences o he use . These p e e ences may be ela ed o he con en s o he ag eemen , he pa y we a e nego ia ing he ag eemen and he nego i- a ion p ocess. –The in o ma ion he sys em ha e abou he s a us o he ma ke and o he possible concu en nego ia ions. Fo ins ance, we may be mo e eluc an o accep a p oposal i we know i is e y likely ha in a sho amoun o ime we will ecei e a p oposal be e han he cu en one [7]. –Ex e nal ac o s ha may p e en a pa y o commi o an ag eemen . Fo ins ance, he p o ide ’s capabili y o accep new ag eemen s o he exis ence o dependencies amongs he ag eemen s a se ice consume wan s o each. Toge he wi h he decision o making o accep ing a binding p oposal, i has o be decided when his decision ake place. Usually, he decision is made as he p oposals a e ecei ed like in [12]. Howe e , o he app oaches may be ollowed such as making he decision in some ce ain poin s in ime p e iously defined. Response Gene a ion. O he impo an ask in an au oma ed nego ia o is o decide which esponse mus be sen o he o he pa icipan s in he nego ia ion. On he one hand, his esponse is subo dina ed o he binding decision. On he o he hand, he esponse gene a ed mus obey he ules imposed by he nego ia- ion p o ocol. The p ocess ollowed o gene a e he esponses a ies significan ly. Howe e , in gene al, i depends on he pe o ma i es o he nego ia ion p o ocol and he exp essi i y o he in o ma ion exchanged du ing he nego ia ion: –In auc ions, he unique possible esponse is a bid oge he wi h he bidding p ice. The e o e, in his case he p oblem is cen ed on deciding in which auc ion mus be placed he bid [13]. –In bila e al p oposal-based p o ocols, a coun e p oposal mus be gene a ed. A wide a ie y o echniques ha e been de eloped o gene a e hem. The mos significan a e: hose ha use ime-dependan unc ions, esou ce-dependan unc ions, e ce e a o ob ain he coun e p oposal by modi ying he alues o he e ms o he offe [4]; hose ha y o make he coun e p oposal mo e appealing o he opponen by sending he coun e offe wi h he highes simila i y o he ecei ed offe [14]; hose ha use uzzy cons ain echniques [15], and hose ha in e p e he nego ia ion as i i we e a game and use echniques simila o hose used in chess games [16]. Gene ic algo i hms ha e also been used o calcula e offline which is he bes s a egy o use depending on he condi ions o he nego ia ion in a ce ain ins an [17]. –In nego ia ion p o ocols ha suppo s a gumen a ion, he esponse gene a- ion includes wo p oblems [9]. Fi s , he diffe en a gumen s mus be gene - a ed and hen, he bes a gumen om he poin o iew o he speake mus be selec ed. Howe e , bo h he gene a ion o a gumen s and he selec ion o he bes one may occu a he same ime. Ne e heless, whe he o no his is possible depends on he specific a gumen a ion amewo k. Decommi Decision. I has been p o ed [18,12] ha decommi men is a alu- able esou ce when dealing wi h mul iple simul aneous nego ia ions and, hence, i is ano he decision elemen ha mus be included in an au oma ed nego i- a ion sys em. The decommi decision is highly ela ed o he binding decision and depends on he same ac o s. The e o e, in many cases, bo h he binding and decommi decisions a e made by he same elemen [12]. 3 P ope ies o Au oma ed Nego ia ion Sys ems The ollowing p ope ies o au oma ed nego ia ion sys ems (ANS) ha e been ob ained based on he concep ual amewo k (Sec ion 2). We mus ema k ha hese p ope ies a e jus cen ed on high-le el de ails o au oma ed nego ia ion sys ems and do no co e lowe -le el elemen s such as conc e e echnologies, p o ocols o algo i hms. 1. In o ma ion que y: An in o ma ion que y is an inqui y made by one pa y o ano he o ob ain mo e de ailed in o ma ion abou i o abou he se ice i p o ides o demands. 2. Wo ld model: An ANS builds a wo ld model i i analyses p e ious in e ac- ions wi h he elemen s ex e nal o he a chi ec u e and uses he esul s o make be e decisions [2] du ing he nego ia ion. 3. Thi d pa y in o ma ion: An ANS uses hi d pa y in o ma ion i i explic- i ly que ies a hi d pa y en i y o ob ain in o ma ion ela ed o ano he pa y. 4. In o ma ion managed abou he pa ies: The e a e h ee ypes o in o ma- ion ha can be managed: abou he pa ies, he se ice and he nego ia ion p ocess i sel . 5. Mul iple nego ia ion p o ocols suppo ed: An ANS suppo s mul iple ne- go ia ion p o ocols i i is able o use a numbe o nego ia ion p o ocols such as diffe en bila e al nego ia ion p o ocols o se e al auc ion p o o- cols [2]. 6. Decommi men om p e iously es ablished ag eemen s: An ANS suppo s he decommi men [18] om p e iously es ablished ag eemen s i i can e oke p e ious ag eemen s be o e he execu ion o he se ice. 7. Ex e nal ac o s in binding decisions: An ANS may make use o a capaci y es ima o o de e mine whe he he p o ide can p o ision a ce ain ag ee- men be o e commi ing o an ag eemen [19]. 8. Coope a i e o non-coope a i e ag eemen c ea ion: An ANS suppo s non- coope a i e ag eemen c ea ion when i ac s as a sel -in e es ed pa y each- ing ag eemen s wi h o he sel -in e es ed pa ies. Al e na i ely, an ANS suppo s coope a i e ag eemen c ea ion when i can each ag eemen s wi h o he pa ies ying o maximise he social wel a e. 9. Assessmen mechanisms used: The assessmen mechanisms o an ANS is he kind o in o ma ion used o e alua e he goodness o a p oposal o ag eemen in ela ion o some c i e ia p o ided by he use . 10. Fo ms o exp essing p e e ences: The p e e ences abou he se ice and he pa ies can be exp essed in diffe en ways such as ules o cons ain s (see Sec ion 2). 4 Conclusions In his wo k, we analyse he p oblem o au oma ed nego ia ion o se ice le el ag eemen s om a so wa e enginee ing pe spec i e. To de elop ad anced au- oma ed nego ia ion sys ems, i is necessa y o iden i y he elemen s ha a e equi ed in hem. We p esen a concep ual amewo k o au oma ed nego ia- ion o SLAs ha iden ifies he elemen s ha mus be suppo ed by an ad anced au oma ed nego ia ion sys em. We ex end he classical elemen s o au oma ed nego ia ion wi h wo new ones: wo ld model and p e e ences.Fu he mo e,we de ail mo e p ecisely he pa s ha compose he o he h ee: nego ia ion objec , p o ocol and decision-making. In addi ion, aking he concep ual amewo k as a basis, we also p esen a se o p ope ies o au oma ed nego ia ion sys ems ha can be used o compa e hem. Mo eo e , we belie e ha his wo k is a necessa y s ep o de elop a so wa e amewo k o au oma ed nego ia ion sys em ha gi es suppo o he mos significan p oposals o nego ia ion p o ocols and s a egies ha ha e been done in he las yea s. Fu he wo k includes efining he ca ego isa ion o au oma ed nego ia ion p oposals in o de o iden i y which al e na i e is be e in each case. In his way, an au oma ed selec ion o he algo i hms used o ca y ou he nego ia ion can be made depending on he con ex o he nego ia ion. Re e ence 1. Sie a, C., Fa a in, P., Jennings, N.R.: A Se ice-O ien ed Nego ia ion Model Be- ween Au onomous Agen s. In: P oc. o he 8 h Eu opean Wo kshop On Modelling Au on. Agen s in a Mul i-Agen Wo ld, Sp inge -Ve lag (1997) 17–35 2. Jennings, N.R., Fa a in, P., Lomuscio, A.R., Pa sons, S., Woold idge, M., Sie a, C.: Au oma ed Nego ia ion: P ospec s, Me hods and Challenges. G oup Decision and Nego ia ion 10 (2001) 199–215 3. And ieux, A., Czajkowski, K., Dan, A., Keahey, K., Ludwig, H., P uyne, J., Ro ano, J., Tuecke, S., Xu, M.: WS-Ag eemen Specifica ion (2004) 4. Fa a in, P., Sie a, C., Jennings, N.R.: Nego ia ion Decision Func ions Fo Au- onomous Agen s. In . Jou nal o Robo ics and Au onomous Sys ems 24 (1998) 159–182 5. Zeng, D., Syca a, K.: Bayesian Lea ning in Nego ia ion. In . J. Hum.-Compu . S ud. 48 (1998) 125–141 6. Sim, K.M., Wong, E.: Towa d ma ke -d i en agen s o elec onic auc ion. Sys ems, Man and Cybe ne ics, Pa A, IEEE T ansac ions on 31 (2001) 474–484 7. Li, C., Giampapa, J., Syca a, K.: Bila e al nego ia ion decisions wi h unce ain dynamic ou side op ions. In: Elec onic Con ac ing, 2004. P oceedings. Fi s IEEE In e na ional Wo kshop on. (2004) 54–61 8. Klein, M., Fa a in, P., Sayama, H., Ba -Yam, Y.: P o ocols Fo Nego ia ing Com- plex Con ac s. IEEE In elligen Sys ems 18 (2003) 32–38 9. Rahwan, I., Ramchu n, S.D., Jennings, N.R., McBu ney, P., Pa sons, S., Sonenbe g, L.: A gumen a ion-Based Nego ia ion. The Knowledge Enginee ing Re iew 18 (2003) 343–375 10. Ba olini, C., P eis , C., Jennings, N.R.: A chi ec ing Fo Reuse: A So wa e F amewo k Fo Au oma ed Nego ia ion. In: Agen -O ien ed So wa e Enginee ing III: AOSE 2002. Volume 2585 o Lec u e No es in Compu e Science., Sp inge - Ve lag (2003) 88–100 11. Ramchu n, S.D., Jennings, N.R., Sie a, C.: Pe suasi e Nego ia ion Fo Au- onomous Agen s: A Rhe o ical App oach. In: P oc. IJCAI Wo kshop On Compu- a ional Models o Na u al A gumen . (2003) 9–17 12. Nguyen, T., Jennings, N.: Managing commi men s in mul iple concu en nego i- a ions. Elec onic Comme ce Resea ch and Applica ions 4(2005) 362–376 13. An hony, P., Jennings, N.R.: De eloping a Bidding Agen Fo Mul iple He e oge- neous Auc ions. ACM T ans. In e ne Technology 3(2003) 185–217 14. Fa a in, P., Sie a, C., Jennings, N.R.: Using Simila i y C i e ia o Make T ade- Offs in Au oma ed Nego ia ions. A ificial In elligence 142 (2002) 205–237 15. Kowalczyk, R.: Fuzzy e-nego ia ion agen s. So Compu ing 6(2002) 337–347 16. Ka p, A.H., Wu, R., Chen, K.Y., Zhang, A.: A Game T ee S a egy Fo Au oma ed Nego ia ion. In: ACM Con e ence On Elec onic Comme ce. (2004) 228–229 17. Fa ima, S.S., Woold idge, M., Jennings, N.R.: A Compa a i e S udy o Game Theo e ic and E olu iona y Models o Ba gaining Fo So wa e Agen s. A ificial In elligence Re iew 23 (2005) 187–205 18. Sandholm, T., Lesse , V.: Le eled commi men con ac s and s a egic b each. Games and Economic Beha io 35 (2001) 212–270 19. Ludwig, H., Gimpel, H., Dan, A., Kea ney, R.: Templa e-Based Au oma ed Se ice P o isioning - Suppo ing he Ag eemen -D i en Se ice Li e-Cycle. In: ICSOC. (2005) 283–295