Multi-Agent SystemsLecture 3

University “Politehnica” of Bucarest2005-2006

Adina Magda Floreaadina@cs.pub.ro

http://turing.cs.pub.ro/blia_06

Agent communicationLecture outline

The nature of communication Indirect communication Direct communication

Agent Communication Languages• Content languages• Ontologies• Theory of speech acts• KQML• FIPA and FIPA-ACL

Interaction protocols

1. The nature of communication

1.1 Human communication Communication is the intentional exchange of information brought

about by the production and perception of signs drawn from a shared system of conventional signs (AIMA, Russell&Norvig) language

Action (communicative act); intentional stance

1.2 Component steps of communicationSpeaker Hearer Intention Perception Generation Analysis Synthesis Disambiguation

Incorporation

3

SyntaxSemanticsPragmatics

1.3 Artificial Communication low-level language vs high-level languages direct communication vs. indirect communication

Agent communication/ MAS communication low-level communication: simple signals, traces, low-level languages high-level communication - cognitive agents, mostly seen as intentional

systems Communication in MAS Implies interaction The environment provides a computational infrastructure where

interactions among agents take place. The infrastructure includes protocols for agents to communicate and

protocols for agents to interact

4

Communication protocols = enables agents to exchange and understand messages

Interaction protocols = enable agents to have conversations, i.e., structured exchanges of messages

Aim Communication enables agents to: coordinate their actions and behavior, a property of

a MAS performing some activity in a shared environment

attempt to change state of the other agents attempt to make the other agents perform some

actions

5

2. Indirect communication2.1 Signal propagation - Manta, A. Drogoul 1993 An agent sends a signal, which is broadcast into the environment, and

whose intensity decreases as the distance decreases At a point x, the signal may have one of the following intensities

V(x)=V(x0)/dist(x,x0) or V(x)=V(x0)/dist(x,x0)2

6

x0

SS - stimulus

Agent A(stimulus triggersbehavior P)

Agent B(stimulus triggersbehavior P)

2.2 Trails - L. Steels, 1995 agents drop "radioactive crumbs" making trails an agent following a trail makes the trail faint until it disappears

Reactive agentsReactive agents

2.3 Blackboard systems, Barbara Hayes-Roth, 1985

Blackboard = a common area (shared memory) in which agents can exchange information, data, knowledge

Agents initiates communication by writing info on the blackboard Agents are looking for new info, they may filter it Agents must register with a central site to receive an access

authorization to the blackboard Blackboard = a distributed knowledge computation paradigm Agents = Knowledge sources (KS)

7

Cognitive agentsCognitive agents

KS

KS

KS

KS

KSAR

Control

Blackboard

3. Direct communicationSending messages method invocation – Actors exchange of partial plans – coordination of cooperative

agents

ACL = Agent Communication Languages Need to communicate knowledge knowledge

representation Need to understand the message in a context ontologies Communication is seen as an action - communicative acts

8

Concepts (distinguish ACLs from RPC, RMI or CORBA, ORB): An ACL message describes a desired state in a declarative

language, rather than a procedure or method invocation ACLs handle propositions, rules, and actions instead of

objects with no associated semantics - KR ACLs are mainly based on BDI theories: BDI agents

attempt to communicate their BDI states or attempt to alter interlocutor's BDI state – Cognitive Agents

ACLs are based on Speech Act Theory – Communicative Acts

ACLs refer to shared Ontologies Agent behavior and strategy drive communication

and lead to conversations - Protocols

9

3.1 Agent Communication Languages

10

Origins of ACLsOrigins of ACLsKnowledge Sharing Effort - DARPA, 1990 External Interface Group - interaction between KBS - KQML Interlingua - common language of KB - KIF Shared, Reusable Knowledge Bases - Ontolingua

ACL

Content language

Ontology

Communication primitivesand protocols

Content languages• KIF• Prolog• Clips• SQL• FIPA-SL, FIPA-KIF

Ontologies DAML OWL

DARPA Agent Markup Language (August 2000).The goal of the DAML effort is to develop a languageand tools to facilitate the concept of the Semantic Web.

11

3.2 Content languages for ACL3.2 Content languages for ACL(Knowledge representation)

Description Logics (DL) - a formalism for expressing concepts and their interrelationships.

In DL, concepts are organized into IS-A hierarchies.Concepts are specifications such that given an individual

(object instance) a DL system can recognize the individual and determine which concepts it belongs to.

DL systems also perform subsumption checking.

Knowledge Interchange Format (KIF) - is based on first order predicate logic and has a LISP-like prefix syntax.

KIF is capable of expressing facts and rules.KIF provides constructs for describing procedures, i.e.

programs to (possibly) be executed by an agent.

12

Knowledge Interchange Format (KIF)

Facts(salary 015-46-3946 john 72000)(salary 026-40-9152 michael 36000)(salary 415-32-4707 sam 42000)

Asserted relation (> (* (width chip1) (length chip1)) (* (width chip2) (length chip2)))

Rule(=> (and (real-number ?x) (even-number ?n)) (> (expt ?x ?n) 0))

Procedure(progn (fresh-line t)

(print "Hello!")(fresh-line t))

13

3.3 OntologiesOntology = a specification of objects,

concepts, and relationships in a particular domain

It comprises a vocabulary, a domain theory and a conceptual schemata to describe organization and interpretation

Lexicalized ontologies (WordNet, EuroWordNet, BalkanNet, FrameNet, MikroKosmos).

Ontologies for knowledge representation

Person

PupilStudEmpl

Sun_EIBM_E

Person

Empl

Woman

Stud

Man

Joe Alice

Joe Alice

Ontology Languages Wide variety of languages for “Explicit Specification”

• Graphical notations– Semantic networks

Ontology Languages Wide variety of languages for “Explicit Specification”

• Graphical notations– Topic Maps

Ontology Languages Wide variety of languages for “Explicit Specification”

• Graphical notations– UML

Ontology Languages Wide variety of languages for “Explicit Specification”

• Graphical notations– RDF

Ontology Languages Wide variety of languages for “Explicit

Specification” • Logic based

– Description Logics (e.g., OIL, DAML+OIL, OWL)

– Rules (e.g., RuleML, LP/Prolog)

– First Order Logic (e.g., KIF)

Objects/Instances/Individuals• Elements of the domain of discourse• Equivalent to constants in FOL

Types/Classes/Concepts• Sets of objects sharing certain characteristics• Equivalent to unary predicates in FOL

Relations/Properties/Roles• Sets of pairs (tuples) of objects• Equivalent to binary predicates in FOL

Such languages are/can be:• Well understood• Formally specified• (Relatively) easy to use• Amenable to machine processing

Many languages use “object oriented” model based on:

Web “Schema” Languages Existing Web languages extended to facilitate content

description• XML XML Schema (XMLS)• RDF RDF Schema (RDFS)

XMLS not an ontology language• Changes format of DTDs (document schemas) to be XML• Adds an extensible type hierarchy

– Integers, Strings, etc.– Can define sub-types, e.g., positive integers

RDFS is recognisable as an ontology language• Classes and properties• Sub/super-classes (and properties)• Range and domain (of properties)

RDF and RDFS RDF stands for Resource Description Framework It is a W3C candidate recommendation

(http://www.w3.org/RDF) RDF is graphical formalism ( + XML syntax + semantics)

• for representing metadata• for describing the semantics of information in a machine-

accessible way RDFS extends RDF with “schema vocabulary”, e.g.:

• Class, Property• type, subClassOf, subPropertyOf• range, domain

Problems with RDFS RDFS too weak to describe resources in sufficient detail

• No localised range and domain constraints– Can’t say that the range of hasChild is person when applied to persons and elephant

when applied to elephants

• No existence/cardinality constraints– Can’t say that all instances of person have a mother that is also a person, or that

persons have exactly 2 parents

• No transitive, inverse or symmetrical properties– Can’t say that isPartOf is a transitive property, that hasPart is the inverse of isPartOf

or that touches is symmetrical

• … Difficult to provide reasoning support

• No “native” reasoners for non-standard semantics• May be possible to reason via axiomatisation

Web Ontology Language Requirements

Desirable features identified for Web Ontology Language:

Extends existing Web standards • Such as XML, RDF, RDFS

Easy to understand and use• Should be based on familiar KR idioms

Formally specified Of “adequate” expressive power Possible to provide automated reasoning support

OWL - Web Ontology Language

W3C OWL - designed for use by applications that need to process

the content of information instead of just presenting information to humans.

OWL facilitates greater machine interpretability of Web content than that supported by XML, RDF, and RDF Schema (RDF-S) by providing additional vocabulary along with a formal semantics.

OWL has three increasingly-expressive sublanguages: OWL Lite supports a classification hierarchy and simple constraints. OWL DL supports maximum expressiveness while retaining

computational completeness (all conclusions are guaranteed to be computed) and decidability (all computations will finish in finite time).

OWL Full supports maximum expressiveness and the syntactic freedom of RDF with no computational guarantees.

OWL - Web Ontology Language

Ontology header<owl:Ontology rdf:about="">

<rdfs:comment>An example OWL ontology</rdfs:comment>

<owl:priorVersion rdf:resource="http://www.w3.org/TR/2003/CR-owl-guide-20030818/wine"/>

<owl:imports rdf:resource="http://www.w3.org/TR/2003/PR-owl-guide-20031215/food"/>

<rdfs:label>Wine Ontology</rdfs:label>

…

Simple Named Classes

Class, rdfs:subClassOf<owl:Class rdf:ID="Winery"/>

<owl:Class rdf:ID="Region"/>

<owl:Class rdf:ID="ConsumableThing"/>

3.4 Theory of Speech ActsJ. Austin - How to do things with words, 1962, J. Searle - Speech acts,

1969

A speech act has 3 aspects: locution = physical utterance by the speaker illocution = the intended meaning of the utterance by the speaker (performative) prelocution = the action that results from the locution

Alice told Tom: "Would you please close the door"locution illocution content

prelocution: door closed (hopefully!)

Illocutionary aspect - several categories Assertives, which inform: the door is shut Directives, which request: shut the door, can pelicans fly? Commissives, which promise something: I will shut the door Permissive, which gives permission for an act: you may shut the door Prohibitives, which ban some act: do not shut the door Declaratives, which causes events: I name you king of Ruritania Expressives, which express emotions and evaluations: I am happy 26

3.5 KQML - Knowledge Query and Manipulation Language

A high-level, message-oriented communication language and protocol for information exchange, independent of content syntax (KIF, SQL, Prolog,…) and application ontology

KQML separates: semantics of the communication protocol (domain independent) semantics of the message (domain dependent)3 (conceptual) layers

27

Content

Communication

Message

Describes low levelcommunication parameters:- identity of sender and receiver- an unique id associated with the communication

Core of KQML- identity of the network protocol with which to deliver the message- speech act or performative Optional- content language- ontology

Syntax S-expressions used in LISPKQML performatives are classified:

Queries - These performatives are used to send questions for evaluation somewhere. Generative - Used for controlling and initiating the exchange of messages. Response - Used by a agent in order reply to queries. Informational - Informational performatives are used to transfer information. Capability definition - Allows an agent to learn about the capabilities of other agents and to announce its

own to the agent community. Networking - Networking performatives make it possible to pass directives to underlying communication

layers.

Example(ask-one :sender joe

:receiver ibm-stock:reply-with ibm-stock:language PROLOG:ontology NYSE-TICKS:content (price ibm ?price) )

28

(tell :sender willie:receiver joe:reply-with block1:language KIF:ontology BlockWorld:content (AND (Block A)(Block

B)(On A B)) )

29

1. Query performatives:ask-one, ask-all, ask-if, stream-all,...

(stream-all :sender willie:receiver ibm-stock

:content (price ?VL ?price ) )

ask-one(P)

tell(P)

tell(P1)

stream-all(P)

tell(P2)eos

tell(P1,P2,...)

ask-all(P)

B

B

B

A

A

A

30

4. Generic informational performatives:tell, untell, insert, delete, ...

In fact, KQML contains only 2 types of illocutionary acts: assertives and directives

+ facilitator and network-related performatives (no necessarily speech acts)

tell(P)

untell(P)

delete(P)

insert(P)

2. Generative performatives:standby, ready, next, rest, discard, generate,...

6. Network performatives:register, unregister, forward, route, ...

3. Response performatives:reply, sorry ...

5. Capability performatives:advartise, subscribe, recommend...

Facilitator

A

A

B

B

31

Facilitator agent= an agent that performs various useful communication services:

maintaining a registry of service names (Agent Name Server) forwarding messages to named services routing messages based on content matchmaking between information providers and clients providing mediation and translation services

tell(P)

ask(P)

subscribe(ask(P)) tell(P)

tell(P)

advertise(ask(P))

tell(P)

recruit(ask(P))

reply(A)

recommend(ask(P)) advertise(ask(P))

tell(P)

ask(P)

reply(B)

A

A

A

A

B

B

B

B

point-to-point

32

Semantics of KQML (Labrou & Finin) Use preconditions and postconditions that govern the use of a performative + the

final state for the successful performance of the performative Uses propositional attitudes: belief, knowledge, desire, intentions

Preconditions: the necessary states for an agent to send a performative and for the receiver to accept it and successfully process it; if the precondition does not hold, the most likely response is error or sorry

Postconditions - describe the state of the sender after successful utterance of a performative and of the receiver after the receipt and processing of a message

Completion condition - the final state after a conversation has taken place and that the intention associated with the performative that started the conversation has been fulfilled

Propositional attitudes

Bel(A,P) Know(A,S) Want(A,S) Int(A,S)

Instances of action

Proc(A,M) SendMsg(A,B,M)

33

tell(A,B,X)A states to B that A believes the content X to be true, Bel(A,X)

Pre(A): Bel(A,X) Know(A, Want(B, Know(B, Bel(A,X))))

Pre(B): Int(B, Know(B, Bel(A,X)))

or Bel(A,X)

Post(A): Know(A, Know(B, Bel(A,X))) no unsolicited information

Post(B): Know(B, Bel(A,X))

Completion: Know(B, Bel(A,X))

advertise(A,B,M)A states to B that A can and will process the message M from B, if it receives one

Int(A, Proc(A,M)) commisive act

Pre(A): Int(Proc(A,M))

Pre(B): NONE

Post(A): Know(A, Know(B, Int(A, Proc(A,M)))

Post(B): Know(B, Int(A, Proc(A,M)))

Completion: Know(B, Int(A, Proc(A,M)))

34

3.6 FIPA and FIPA - ACL Foundation for Intelligent Physical Agents, 1996

Goal of FIPA = make available specifications that maximize interoperability across agent-based systems

FIPA Committees: ACL, agent specification, agent-software interaction

As KQML, FIPA ACL is based on speech act theory; it sees messages as communication acts (CA); syntax similar to KQML

Differs in: the names of CAs, set of CAs, and semantics

35

FIPA standard define normative specifications for: agent management (or agent platform services)

white pages via an Agent Name Server (ANS) yellow pages via a Directory Facilitator (DF) registration in a given DF defines a domain (agent community) an Agent Platform defines a logical "place" containing an ANS,

DF, management tools, and a collection of agents interplatform communication takes place via an Agent

Communication Channel (ACC), which defaults to CORBA IIOP agent communication language - FIPA ACL

based on speech acts has a formal semantics also included are several predefined protocols (e.g., contract-net

negotiation and auction protocols), and the concept of application-specific protocols

36

agent-software integration defines Agent Request Broker and Wrapper roles allows an agent system to integrate non-agent software details of how wrapper communicates with wrapped

software are left to the implementor (a Wrapper is an agent, and communicates with other agents via ACL)

several reference applications for: personal travel assistance personal assistant network provisioning and management audio/video entertainment and broadcasting

37

FIPA has extended these specifications, including work on: agent management support for mobility (identifying the relationship

between this work and MASIF is explicitly targeted) an ontology service, supporting

- translation of terms between different ontologies

- downloading meanings of terms, axioms, and relationships between terms

- querying for relationships between ontologies uploading and updating of ontologies additional applications, e.g., product design and manufacturing

agents

FIPA does not currently constrain the low-level implementation of agents to any great extent, nor, except for defining agent platform services, does it constrain the infrastructure a great deal.

38

FIPA - ACL

FIPA communicative acts

Informatives - query_if, subscribe, inform, inform_if, confirm, disconfirm, not_understood

Task distribution - request, request_whenever, cancel, agree, refuse, failure

Negotiation - cfp, propose, accept_proposal, reject_proposal

39

FIPA-SL(inform :sender Agent1

:receiver Agent2:content (price good2 150):in-reply-to round-1: reply-with bid03: language S1:ontology hp-auction:reply-by 10:protocol offer:conversation-id conv-1 )

40

FIPA-SL(request :sender (agent-identifier :name i) :receiver (set (agent-identifer :name j)

:content ((action (agent-identifier :name j)(deliver box7 (loc 10 15))))

:protocol fipa-request :language fipa-sl :reply-with order56 )

(agree :sender (agent-identifier :name j) :receiver (set (agent-identifer :name i)

:content ((action (agent-identifier :name j)(deliver box7 (loc 10 15))) (priority order56 low))

:protocol fipa-request :language fipa-sl :in-reply-to order56 )

41

FIPA - SemanticsSL (Semantic Language) - a quantified, multi-modal logic, with

modal operators

Allows to represent:beliefsuncertain beliefsdesiresintentions

B - belief D - desire U - uncertain belief

PG - intention

Bif - express whether an agent has a definite opinion one way or another about the truth or falsity of

Uif - the agent is uncertain about

42

FIPA - SemanticsThe semantics of a CA is specified as a set of SL's formulae that

describe: Feasibility preconditions - the necessary conditions for the sender -

the sender is not obliged to perform the CA Rational effect - the effect that an agent can expect to occur as a

result of performing the action; it also typically specifies conditions that should hold true of the recipientThe receiving agent is not required to ensure that the expected effect comes aboutThe sender can not assume that the rational effect will necessary follow

<A, inform(B, )>

Pre: BA BA (BifB UifB )

Post: BB

43

KQML and FIPA ACL

The two ACLs are essentially the same Although FIPA ACL requires agents to have a limited knowledge

of SL, both the ACLs do not have fixed semantics. The FIPA ACL does not provide for facilitator agents. This is a major drawback as using facilitator is one of the best

way to overcome different systems using different content language and providing matchmaking service.

KQML provides for brokering and recommendation service, whereas FIPA ACL don't really take this into account.

4. Interaction protocols

Interaction protocols = enable agents to have conversations, i.e., structured exchanges of messages

Finite automata Conversations in KQML Petri nets FIPA IP standards:

• FIPA-query, FIPA-request, FIPA-contract-net, ...

4.1 Finite state automata

45

A:B<<ask(do P)

B:A<<accept(do P)

B:A<<refuse(do P)

B:A<<result(do P) B:A<<fail(do P)

proposeS(P)

acceptR(P) rejectR(P)

counterR(P)counterS(P)

acceptS(P) rejectS(P)Winograd, Flores, 1986

COOL, Barbuceanu,95

46

4.2 Conversations in KQMLUse Definite Clause Grammars (DCG) formalism for the specification of conversation policies for KQML performatives

DCGs extend Context Free Grammars in the following way: non-terminals may be compound terms the body of the rule may contain procedural attachments, written as "{" and "}"

that express extra conditions that must be satisfied for the rule to be valid

Ex: noun(N) [W], {RootForm(W,N), is_noun(N)}

S s(Conv, P, S, R, inR, Rw, IO, Content), {member(P, [advertise, ask-if]}

s(Conv, ask-if, S, R, inR, Rw, IO, Content) [ask-if, S, R, inR, Rw, IO, Content] |

[ask-if, S, R, inR, Rw, IO, Content], {OI is inv(IO)},

r(Conv, ask-if, S, R, _, Rw, OI, Content)

r(Conv, ask-if, R, S, _, inR, IO, Content) [tell, S, R, inR, Rw, IO, Content] |

problem(Conv, R, S, inR, _, IO) Labrou, Finin, 1998

DAA wants to do P,A cannot do P

Request do(P)

Refuse do(P)

Accept/request do(P)

Fail to do(P)

Notification of end(P)

DB

AR1

AR2

FA2FA1

BR

FBFailure Satisfaction

Impossibleto do(P)

B does not wantto do(P)

B is willingto do(P)

Completed(P)

4.3 Petri netsPetri net = oriented graph with 2 type of nodes:places and transitions;there are moving tokens through the net - representation of dynamic aspect of processes.Tokens are moved from place to place, following firing rules.A transition T is enabled if all the input places P of T posses a token (several other rules may be defined).A marking is a distribution of tokens over places. Colored Petri-nets

Ferber, 1997

47

Success

48

References M. Huhns, L. Stephens. Multiagent systems and societies of agents. In

Multiagent Systems - A Modern Approach to Distributed Artificial Intelligence, G. Weiss (Ed.), The MIT Press, 2001, p.79-120.

M. Wooldrige. Reasoning about Rational Agents. The MIT Press, 2000, Chapter 7

Y. Labrou, T. Finin. Semantics and conversations for an agent communication language. In Readings in Agents, M. Huhns & M. Singh (Eds.), Morgan Kaufmann, 1998, p.235-242.

J. Ferber - Multi-Agent Systems. Addison-Wesley, 1999, Chapter 6 T. Finnin, R. Fritzson - KQML as an agent communication language. In Proc. of

the Third International Conference on Information and Knowledge Management (CIKM'94), ACM Press, 1994.

M. Singh. Agent communication languages: Rethinking the principles. IEEE Computer, Dec. 1998, p.40-47.

Y. Labrou, T. Finnin, Y. Peng. Agent communication languages: The current Landscape. IEEE Computer, March/April 1999, p. 45-52.

FIPA97. "Agent Communication Language" Specification FIPA, 11/28/97

49

Web References

DARPA KSE http://www-ksl.stanford.edu/knowledge-sharing/

KQML http://www.cs.umbc.edu/kqml/

KIF http://logic.stanford.edu/kif/

Ontolingua http://www-ksl-svc.stanford.edu:5915/&service=frame-editor

FIPA http://www.fipa.org/

DAML http://www.daml.org/

OWL http://www.w3.org/TR/owl-guide/

References for Ontologies (due to prof. Stefan Trausan) Constandache, G.G., Ştefan Trăuşan-Matu, Ontologia şi hermeneutica

calculatoarelor, Ed. Tehnică, 2001 Gruber, T., What is an Ontology, http://www.kr.org/top/definitions.html J. Sowa, Ontologia şi reprezentarea cunoştinţelor, în (Constandache şi Trăuşan-

Matu, 2001) http://www.w3.org/2001/sw/WebOnt/ http://www.cs.man.ac.uk/~horrocks/Slides/index.html