ѧ
JOURNAL OF SOFTWARE
1999 10 7ڡNo.7 Vol.10 1999



ֶ֧AgentͨŵչBDI߼
½ССƺӥ
ժҪBDIͼ߼ǵǰagentоȵ֮һ,о漰agentͨŵ.ּһܱʾagentΪ۵ͨΪչBDI߼.߼һࡢʲз֦ʱṹһ׶ģ̬߼,˰׼BDI߼ģ̬ʱ̬֮,һЩҪ߼Ӻ͹,ڿ̻agentΪʵͼеĹ滮,ԼԶagentͨŸȷý.߼﷨,Ϊ߼ͽ.
ؼʡBDI agent߼,agentͨ,Ϊ.
ͼšTP18
An Extended BDI Logic for Modeling Multi-agent Communication
CHEN Jian-zhong1,2LIU Da-you1,2TANG Hai-ying1,2HU Ming3
1(Department of Computer ScienceJilin UniversityChangchun130023)
2(Open Symbol Computation and Knowledge Engineering Laboratory of State Education Department
Jilin UniversityChangchun130023)
3(Department of Computer ScienceJilin Institute of TechnologyChangchun130012)
Abstract Although BDI approach is currently a subject of much ongoing research, less relative works about modeling communication have been involved. The aim of this paper is to extend classical BDI logic for representing and reasoning about speech act interactions between multi agents. A many-sorted, quantified, first-order, multi-modal logic with branching time structure is presented, including standard BDI modal and temporal operators like belief, goal, intention, etc. In addition, some operators and constructs are introduced for characterizing agents' know-how or having plans about how to achieve their intentions and for interpreting the pragmatic meanings of inter-agent communication. While discussing the syntax and semantics of the logic, some kinds of illocutionary speech acts are logically classified and interpreted.
Key wordsBDI agent logic, multi-agent communication, speech act.
ֲʽ˹DAI(distributed artificial intelligence)ҪоɶagentɵķֲʽϵͳеĸΪ.ǵĹ۵1,DAIɷΪ3:ֲʽDPS(distributed problem solving)agent IA(intelligent agents)ͶagentϵͳMAS (multi-agent system),IAMASǵǰDAIоȵ.һagentӦ:һ,ʼеһЩĳЩ,ܾμʱύ;һ,һagent,ͨ뻷agent,ĳַʽ໥.
о,һһagentԱӵ(beliefs)(desires)滮(plans)ͼ(intentions)Ⱦ״̬νʶϵͳ(intentional system).ⷽо,ĵɰĴǵRaoGeorgeffBDI߼BDI agentṹ.90,Bratman2,CohenLevesque3ݵ˵ǰڹϣһϵصоɹ4,5,ΪagentодԵĻ.
һ͵BDI agentͨ4Ҫݽṹ:,ڱʾagentģ,еϢܲȫȷ;Ŀ(goals),ӦڷagentҪִе;ͼ,ʾagentѾŵɵ,agentһͼĳŵ,ֱŸͼ,ĳЩԭͼٿɣ3;ܹ滮,һԤ滮򷽷,ʾagentΪɸͼѭĹ֪ʶ.֮,ǷֱʾagentĳԡѡԡԺ͹״̬.һΪagentĿģ͹Щݽṹ֮ݺͿ,γӵǰѡͼѡȡʵ滮ִӦͼ.
ео46,ǲڶͼ3״̬ʽ߼߼ϵ,Թ滮agentͨŵʾȴ漰.ͼӽһ,һЧʾagent滮ͨŵչBDI߼.
1agentͨźΪ(speech act theory)
DAIеagentͨӵйȫϢ֪ʶ,ͨŵĻþΪagentṩ罻ҪϢֶ.agentϵͳеЭͬЭЭ̡ȫһµȶͨЧagentͨʵ.
Werner7ݰѶagentͨŷΪ5:ͨ;źͨ;ۻڼϢ,Actorϵͳ;ܹ滮,粿ȫֹ滮(PGP);ݻΪͨ.лΪĶagentͨǵǰоȵ,һ̶ģͨΪ,мʵۺͼ,,agentͨ,֪ʶѯ(KQML)֪ʶʽ(KIF)ȶֱӻӵػΪ.
 ΪԴѧ,ѳΪ˹ܺѧ۵ĺ.ҪоԵ,ǰǵճ￴˵ĴĳͼĶ,ЩνﶯҲЩһ״̬ı仯.һΪٰ3ද:ָ,˵һ﷨ϿɽܵĻ;,ûĳ֮,ܰ˵ߵͼ;Գ,µĽһЩһЩЧ.ͨΪָµΪ.һΪ:֮Ϊ.֮ΪĲͬ,˵״̬ĸı.Singh8ݰ֮Ϊ6:ٶ;ָ;۳ŵ;;ݽֹ;.
ĸʽ߼ͨһģ̬,agentͼ͹滮,Ϊʽ.õԭ:ٰΪ˵ߣϢߣһʱĶ;ߣϢߣ˵˫״̬յϢ;߼,˵߽ǶȸÿΪķΪִж˫״̬ĿӰȵý.
2
ȸһЩķʽ˵:Dag,DacDsaֱʾϵͳagentļϡԭӶļϺΪļ.ѭBDI߼4,չBDI߼ھз֦ʱṹĿģ;Tһʱ㣨ʱ̣,RʾһTϳĶԪ֦ʱϵ,RTT,RT߼Ϲһ÷֦ʱ;DacеһDsaеһΪͳΪTеʱִ,ǶһźActڰѷ֦ʱеÿһ;躯Agtִÿһagent.,INʾ0Ȼ;(s)ʾһsݼ.
1. TǷǿʱ㼯,RTTTϵһȫġԵķ֦ʱϵ,(T,R)Ϊϵһ,WʾTļ,wW,TwRwֱʾwеʱ㼯Ϻʱϵ;tTw,(w,t)Ϊwеһ,SwʾwоƵļ,ʾоƵļ.
2. wW,wϵһ·һʱ(t1,t2,...,tk,...),kIN(tk,tk+1)Rw,Pwʾw·ļ;pPwkIN,pkʾ·pϵkʱ,pkʾpʼpk·.
3. һ滮һԪ顴plan-goal,plan-premise,plan-body,plan-result,й滮Ŀplan-goalһν,ʾ滮͸ù滮Ĵ;滮plan-bodyʾù滮ʱҪִеĳУ,ɼDacDsaԪأԪУ滮Ŀ깹,Dpbʾй滮ļ;滮ǰplan-premiseSʾִpbDpbľƼ;滮plan-resultSʾڹ滮ǰһִpbܲ¾Ƽ;:һ滮ִйе(w1,t1)plan-premise,(w2,t2)plan-result,˵w1=w2,Ᵽ֤˹滮ִйǷ,.һagentĹ滮һ޻ͼ,ͼеĶǹ滮,ÿһźGoal-to-plan,Ӹ߲滮滮пɰ滮Ŀ꣩ܵӹ滮߲滮ĿӦ.һ붥㱻ɹִ,ҽÿӹ滯ִ;һ򶥵㱻ɹִ,ҽһӹ滯ִ.
Dplʾһagent滮й滮ļ.4Pg,Pp,Pb,Prһ滮plDplΪ,ֱ𷵻Ӧplan-goal,plan-premise,plan-body,plan-result.һ滮BNF﷨ṹ,Сa+aһλγ,;,͡*ֱ˳򡢲к͵͸϶ϳɷ.

plan=plan-goalplan-premiseplan-bodyplan-result
plan-goal=Lеν
plan-premise=P(S)һԪ
plan-result=P(S)һԪ
plan-body=ANDplan-expORplan-exp+
plan-exp=DacһԪأDsaһԪأplan-goal
plan-exp;plan-expplan-expplan-expplan-exp*

Ϊʾ˵滮ִе,ν:perf(pb,p,u,v)exec(pe,p,u,v)ֱʾһ滮pbDpbpbеһ滮ʽpe·pϵĵu͵vʱ֮䱻ִ(u,vIN uv)ʵ.
perf(pb,p,u,v)iffpb,pbеpe,exec(pe,p,u,v);
(pbǻ)pbһpe,exec(pe,p,u,v);
exec(ac,p,u,v)iffv=u+1Act(pu,pv)=ac,acDacacDsa;
exec(pg,p,u,v)iffperf(Pb(Goal-to-plan(pg)),p,u,v),pgһ滮Ŀ;
exec(pe1;pe2,p,u,v)iffn{u,...,v},exec(pe1,p,u,n)exec(pe2,p,n,v);
exec(pe1pe2,p,u,v)iffexec(pe1,p,u,v)exec(pe2,p,u,v);
exec(pe*,p,u,v)iffu=vexec(pe;(pe*),p,u,v).
躯AgtԹ滮滮ʽΪ,ִиöеagent:
,peж,pbpe.
Ϊ˷ͨŵ,MsgʾͨϢļ,FC={assertive,directive,commissive,permissive,prohibitive,declarative}ʾΪĸ֮ļ;һϢmMsg򵥱ʾΪһԪ顴fc,pp,fcFC,ppһ;DsaеԪsa(i,j,m)ʾ,mMsg,exec(sa(i,j,m),p,u,v)Ϊĳ,agent iʱuʱvִһagent j˵ϢmΪ;һνcomm(i,j,m)ĺ:comm(i,j,m)Ϊ,iffĳiȷʵj˵m. 
3ֶ֧agentͨŵչBDI߼
߼ϵͳһࡢʵġڿ塢з֦ʱṹһ׶ģ̬߼.һЩ߼ӵĺɷʽؽ:trueӱʾ߼;(Bel i ),(Goal i )(Int i )ֱʾagent iһţ,һĿպһͼҪյʵ,Цһ߼ʽ,ע,Ŀ,3ģ̬ӵͨӦĿɴϵ,׼ģ̬ϵͳKD45,Ŀͼ㹫KD;ӡ=ʾһ׵ȼ۹ϵ;(ig)ʾagent iagentgһԱ;(HasPlan i pl)ʾڵǰ״̬,iһ滮pl,pliĹ滮;ڷ֦ʱṹ,йءֳʱʵĹʽΪ״̬ʽ,Ƶķչйءֳ·ʵĹʽΪ·ʽ;·ʽ(Perf i pb)Ϊiڵǰ·ִһ滮pb,(Exec i ac)Ϊiڵǰ·ִһac;·ʽ(PragInt comm(i,j,m))ڸiڵǰ·һΪý;Aȫ·,AΪ·ʽڴӵǰ״̬ʼδ磨·п;ʱ̬ӷUNڴ״̬ʽ·ʽ·ʽ,ʽUΪտֱ׿,NΪսһ״̬;,(߼)(߼)(ȫ)Ծ侭.
3.1 
4. ַ:(1) ӷtrue,Bel,Goal,Int,=,,A,HasPlan,Perf,Exec,PragInt,UN;(2) ߼ӷ()͡(),ȫʷ;(3) Predνʷ;(4) Func=FuncagFuncacFuncsaFuncplFuncpbFuncgrFuncSFuncU,кźͷֱֵΪagentԭӶΪ滮滮塢agent顢ƺĺ;(5) Var, Var=VaragVaracVarsaVarplVarpbVargrVarSVarU;(6) š)(*͡,;(7) ϶ϳɷ;͡*.arity:PredFuncINÿνʺͺŵԪ,0ԪνʼΪ,0Ԫǳ.
5. Lеһ,һ;Termʾļ;һ֮һ:ag,ac,sa,pl,pb,gr,S,U;һ,Term=FuncҡVarҺTerm=TermagTermacTermsaTermplTermpbTermgrTermSTermU.
Lĺʽʽ(wff)﷨(νʺͺŶѾӦĲ).


3.2
6. D߼ϵͳ󹹳ɵļ,D=DagDacDsaDplDpbDgrDSDU,,ǿռDgrǽDag֮ϵagentļ,DSпܾƵļ,DU󼯺,༯ϵĺͬǰ;DnʾDnԪļ(nIN).
Ϊ﷨ͬϵһ,ͺ.
7. νʽͺP:PredWTp,,QPred,nIN,wW,tTw,arity(Q)=n,P(Q,w,t)Dn,PӦֱνʵԪ;ͺF:Fun(D),,fFun,nIN,arity(f)=n,F(f)Dn,FӦֱͺԪ಻;ָV:VarD,,vVar,V(v)D,VӦֱָɱ಻.
һĺӡF,VһӵָΪFV.
8. ӡTerm,ӡFun,򡬦ӡF,VF(),V(),޶Ե,ǱF,Vͨʡ.
9. ģMһԪṹT,R,W,D,Act,Agt,Msg,PL,MI,BR,GR,IR,F,P,T,R,W,D,Act,Agt,Msg,F,PĶͬǰ;PL:DagWTp(Dpl)ÿagentÿµĹ滮;ML:DagMsgWTp(WT)ÿagent䵱ǰ״̬ͽյÿϢ,͵ĽǾ״̬תΪ״̬;BR:Dagp(WTW)ÿagentɴϵ,ԡԺŷ;GR:Dagp(WTW)IR:Dagp(WTW)ֱÿagentĿɴϵͼɴϵ,.
(*:ӳ亯,ӳ临д),䶨ɽͽṹ͹ʽ,ͨϵ.״̬ʽĽͽṹ硴M,V,w,t,Mһģ,VǱָ,wW,tTw;·ʽĽͽṹ硴M,V,w,p,pwϵһ·.


3.3Ӻӷ
ӺӷõӺӷǿ߼ʽ׶.,źͿ(߼),(߼),(߼ȼ)();Aɶ·E,Ϊ״̬ʽĳ·Ͽ;Ŀʱ̬ӡֱ͡ʾҪ͡Զ:=deftrueU,,·ʽ,()δĳ·Ͽ,ڸ·ϵĳ(ȫ)ʱ̿.Achvʾʵ:agent iִжacͬʱﵽĿ,M,V,w,pn,kIN,nk*(Exec i ac))(
3.4Ϊý
ϸPragInt,Ϊý.˵:ٸͨ˫֪״̬,ÿϢĿ,Ϊܱɹִе;νcomm(i,j,m)ʹýڹʽ,ֱڶ;Ϊ֮ͬ,ýҲͬ.·pϴʱnʱk,һΪsa(i,j,fc,pp),˵ߣagentiߣagentj˵Ϣfc,pp,fcʾΪ֮,ppΪ,i͸Ϊ.
.M,V,w,p(PragInt comm(i,j,assertive,pp)) iff n,kIN*nkexec(sa(i,j,assertive,pp),p,n,k)ҡM,V,w,pkpp.һΪ,硰ſ.,,Ϊppڻ˵ʼΪ.
.M,V,w,p(PragInt comm(i,j,directive,pp)) iffn,kIN*nkexec(sa(i,j,directive,pp),p,n,k)(M,V,w,pk(Bel i ((Int j pp)(HasPlan j Goal-to-plan(pp)))Upp).һָΪ,硰ţ,,i:(1) ·pʱkʼ,jҪͼ֪й滮ﵽppʹpp;(2) ppδĳʱ̽Ϊ.
.M,V,w,p(PragInt comm(i,j,commissive,pp)) iffn,kIN*nkexec(sa(i,j,commissive,pp),p,n,k)ҡM,V,w,pk(((Int i pp)(HasPlan i Goal-to-plan(pp)))Upp).һŵΪ,硰Ҫ.,,(1) ·pʱkʼ,iҪ֪ﵽpp;(2) ppδĳʱ̽Ϊ.
.M,V,w,p(PragInt comm(i,j,permissive,pp)) iffn,kIN*nkexec(sa(i,j,permissive,pp),p,n,k)(ac.(Bel i ((Achv j ac true)(Achv j ac (HasPlan j Goal-to-plan(App)))))).һΪ,硰Կ.,һ·Ͽ,Ϊڸ·Ͻĳʱj,jһЩǰĶac,˵Ϊִнһ״̬,ڸ״̬,jܱ⶯acִķ.
.M,V,w,p(PragInt comm(i,j,prohibitive,pp))iffn,kIN*nkexec(sa(i,j,prohibitive,pp),p,n,k)(ac.(Bel i ((Achv j ac true)A(Achv j ac (HasPlan j Goal-to-plan(App)))))).һֹΪ,硰㲻Կ.,,ij֪ĳЩacԴﵽ(App),jӦбﵽppĹ滮,ppڸ·ϲ.
.M,V,w,p(PragInt comm(i,j,declarative,pp)) iffn,kIN*nkexec(sa(i,j,declarative,pp),p,n,k)ҡM,V,w,pkpp(uIN,nukM,V,w,pu pp)ҡM,V,w,p(((Int i pp)(HasPlan i Goal-to-plan(pp)))Upp).һΪ,硰Ϸʼ.,,(1) ppδĳʱk״Ϊ;(2) ·pʱnʼ,˵,iҪ֪ʹpp.
j䵱ǰ״̬,úMLյÿϢ,ӶԼ״̬,Ȼ,ִлܾϢаĻĶ.ƪ,jԽյΪ۴. 
4 
Ŀǰ,йagentͶagentϵͳķͼԽԽܵѧҵע,ԽϵͳۻҪҲս.ĵĹԴΪ,οصо46,8,9,Ҫ:ٸһչBDI agentʽ߼,ͼֶ֧agentͨΪ;߼а˶agent滮͹滮ṹ;ͨһϵ߼Ӻ͹,ϢͷǶ,˻Ϊ۵agentͨŵý.عȽ,ǶԾBDI߼ף8ݹչ.KQMLͨʵֽǶȶΪ˸ϸµķӦ,δ߼ý.ף9agentЭ;ϵͳΪ,˻Ϊ۵agentͨŹ.ĵĹ,θ,һ.
ĵоܳ,ԻΪagentͨŵвȫͳ.һо:߼ϵͳиĽչ,Ϊǰл,ͳ￼ͨ˫Ϣýͺʶ״̬ĸı;Ը߼Ϊۻ,оagentһЩ,agentṹ,agentЭЭͬЭ,agentƵ.
оõȻѧ𡢹863߿ƼĿ𡢹ҽʿͼִѧż֪ʶ̹ҽʵʽ.߳½,1968,ʿ,,ҪоΪAgentϵͳ,ֲʽ˹,ݽṹ㷨.,1942,,ʿʦ,ҪоΪ˹,Agentϵͳ,רϵͳ,֪ʶ.ƺӥ,1971,ְʿ,ʦ,ҪоΪAgentϵͳ,ֲʽ˹.,1963,ְʿ,,ҪоΪAgentϵͳ,ֲʽ˹.
ͨѶϵ:½,130023,ִѧѧϵ
ߵλ½ССƺӥִѧѧϵ130023ִѧż֪ʶ̹ҽʵҡ130023(ֹѧԺϵ130012)
ο
1Chen Jian-zhong, Liu Da-you, Tang Hai-ying. A template architecture for modeling intelligent agents.In:Sirirucha-tapong P, Lin Zong-kai eds. Proceedings of the International Workshop on CSCW in Design. Beijing: International Academic Publishers, 1997. 196201
2Bratman M E. Intentions, Plans, and Practical Reason. Massachusetts: Harvard University Press, 1987
3Cohen P R, Levesque H J. Intention is choice with commitment. Artificial Intelligence, 1990,42(3):213261
4Rao A S, Georgeff M P. Modeling rational agents within a BDI-architecture. In: Fikes R, Sandewall E eds. Proceedings of the 2nd International Conference on Principles of Knowledge Representation and Reasoning (KR & R-91). San Mateo, CA: Morgan Kaufmann Publishers, 1991. 473484
5Rao A S, Georgeff M P, Sonenberg E A. Social plans: a preliminary report.In: Werner E, Demazeau Y eds. Decentralized AI 3Proceedings of the 3rd European Workshop on Modeling Autonomous Agents in a Multi-agent World (MAAMAW-91). Amsterdam: Elsevier Science Publishers, 1992. 5776
6Wooldridge M, Jennings N R. Formalizing the cooperative problem solving process. In: Klein M ed. Proceedings of the 13th International Workshop on Distributed Artificial Intelligence (IWDAI-94). Seattle, WA: AAAI Press, 1994. 403417
7Werner E. Cooperating agents: a unified theory of communication and social structure. In: Gasser L, Huhns M N eds. Distributed Artificial Intelligence, Vol . Los Altos, CA/Pitman: Morgan Kaufmann Publishers, 1989. 336
8Singh M P. Multiagent Systems: A Theoretical Framework for Intentions,Know-how, and Communication (LNAI Volume 799). Berlin: Springer-Verlag, 1994
9Haddadi A. Reasoning about cooperation in agent systems: a pragmatic theory Ph.D. Thesis. University of Manchester Institute of Science and Technology (UMIST), 1995
1998-05-04յԭ 
1998-07-09յ޸ĸ
