USE OF SEMANTIC TECHNOLOGY TO DESCRIBE AND REASON

ABOUT COMMUNICATION PROTOCOLS

Miren I. Bag¨u´es, Idoia Berges

∗

, Jes´us Berm´udez, Alfredo Go˜ni and Arantza Illarramendi

†

University of the Basque Country, Donostia, Spain

Keywords:

Protocols, Agent communication, Communication Acts.

Abstract:

Nowadays there is a tendency to enhance the functionality of Information Systems by appropriate information

agents. Those information agents communicate through communication acts expressed in an Agent Commu-

nication Language. Moreover, the aim is to achieve interoperation of those agents through standard commu-

nication protocols in a distributed environment such as that supported by the Semantic Web.

In this paper we present a proposal to describe those protocols using a Semantic Web language. Two are the

main features of that proposal. On the one hand, the communication acts that appear in the communication

protocols are described by terms belonging to a communication acts ontology called COMMONT. On the

other hand, protocols are represented by state transition systems described using OWL-DL language. This

type of description provides the means to reason about the communication protocols in such a way that several

kinds of structural relationships can be detected, namely if a protocol is a preﬁx, a sufﬁx or an inﬁx of another

protocol and that relationship taken in a sense of equivalence or specialization. Furthermore, equivalence and

specialization relationships can also be detected for complete protocols. Those relationships are captured by

subsumption of classes described with a Semantic Web language.

1 INTRODUCTION

NOWADAYS the need for interoperability among

agent based information systems in order to inter-

change information and services is increasing. But

it seems to be difﬁcult to achieve it due to the hetero-

geneity in Agent Communication Languages (ACL).

Moreover, once the understanding of atomic com-

munication acts at a semantic level is achieved, the

next step is to deal with agent communication pro-

tocols. We consider a context where administrators

of information systems ﬁrst, select protocols from ex-

isting repositories of protocols and then customize

them. Later, they assign those selected protocols to

their software agents. Therefore, when agents from

two different information systems want to interoper-

ate between them, it is relevant to see if the protocol

used by one of them has structural relationships with

∗

The work of Idoia Berges is supported by a grant of the

Basque Government.

†

All authors are members of the Interoperable

DataBases Group, online at http://siul02.si.ehu.es. This

work is also supported by the University of the Basque

Country, Diputaci´on Foral de Gipuzkoa (cosupported by

the European Social Fund) and the Spanish Ministry of

Education and Science TIN2007-68091-C02-01.

the protocol used by the other. We represent proto-

cols by state transition systems and describe them us-

ing the OWL-DL language (Bechhofer et al., 2005).

So, we can consider structural relationships in a sense

of equivalence or specialization applicable to com-

plete protocols or to parts of them (preﬁx, sufﬁx and

inﬁx). In a sense, our approach creates a classiﬁcation

of protocols.

In relation to other related works, we can observe

two general aspects: 1) many of them also use tran-

sition systems for representing communication proto-

cols and, 2) they do not consider the structural rela-

tionships among protocols managed by our proposal.

In the following we show the main differences among

those works and our proposal.

The closer related work is (Mallya and Singh,

2007), where protocols are represented as transition

systems and subsumption and equivalence of proto-

cols are deﬁned with respect to three state similarity

funtions. We share some goals with that work, but

the protocol description formalism used by them is

not considered in the paper and there is no references

to how protocol relationships are computed. In con-

trast, we describe protocols with a description logic

language and protocol relationships are computed by

I. Bagüés M., Bergesâ

U I., Bermúdez J., Goñi A. and Illarramendi A. (2008).

USE OF SEMANTIC TECHNOLOGY TO DESCRIBE AND REASON ABOUT COMMUNICATION PROTOCOLS.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - SAIC, pages 67-72

DOI: 10.5220/0001677600670072

 SciTePress

a description logic reasoner.

The works of (Yolum and Singh, 2002) and

(Fornara and Colombetti, 2003) are quite similar one

to each other. Both capture the semantics of commu-

nication acts through agents’ commitments and rep-

resent communication protocols using a set of rules

that operate on these commitments. Moreover those

rule sets can be compiled as ﬁnite state machines.

Nevertheless, they do not consider the study of rela-

tionships between protocols. In addition, in (Desai

et al., 2005), protocols are also represented with a set

of rules with terms obtained from an ontology, but

their main goal is protocol development and, in order

to reason about protocol composition, they formal-

ize protocols into the π-calculus. Then, equivalence

through bisimulation is the only process relationship

considered. Notice that in our approach, reasoning

is made with protocols’ own representation, without

demanding another different formalism.

In the context of Web Services composition, (Be-

rardi et al., 2005) leads to a description logic based

speciﬁcation of ﬁnite state machines that is very dif-

ferent in purpose to our approach. In their scenario,

states and transitions descriptions are not prepared to

be confronted in a comparison. In constrast, in our

case, state and transition descriptions are carefully

modelled as class descriptions such that subsumption

relation captures structural relationhips.

In (Kagal and Finin, 2007) protocols are deﬁned

as a set of permissions and obligations of agents par-

ticipating in the communication. They use an OWL

ontology for deﬁning the terms of the speciﬁcation

language, but their basic reasoning is made with an

ad hoc reasoning engine. We share their main goal

of deﬁning protocols in a general framework that al-

lows reutilization. Nevertheless, they do not consider

relationships between protocols.

Finally, (d’Inverno et al., 1998) and (Mazouzi

et al., 2002) use ﬁnite state machines and Petri

nets, respectively, but without taking into account

the meaning of the communication acts interchanged,

neither considering relationships between protocols.

In the rest of this paper we present ﬁrst the main

features of a communication ontology called COM-

MONT. Then, in section 3 we show how we can de-

scribe protocols using OWL. Next, in section 4 we

explain the way we deal with cycles in protocols. The

structural relationships among protocols arepresented

in section 5. We ﬁnish with some conclusions in sec-

tion 6.

2 COMMUNICATION

ONTOLOGY

Among the different models proposed for represent-

ing protocols one which stands out is that of State

Transition Systems (STS). In our proposal we use

STS where each transition is labeled with a communi-

cation act class described in a communication ontol-

ogy called COMMONT. Therefore, we present in the

following the main features of that ontology.

The goal of COMMONT ontology is to favour the

interoperation among agents belonging to different

Information Systems. The leading categories of that

ontology are: ﬁrst, communication acts that are used

for interaction by actors and that have different pur-

poses and deal with different kinds of contents; and

second, contents that are the sentences included in the

communication acts.

The main design criteria adopted for the commu-

nication acts category of the COMMONT ontology is

to follow the speech acts theory (Searle, 1969), a lin-

guistic theory that is recognized as one of the main

sources of inspiration for designing the most familiar

standard agent communication languages. Following

that theory, every communication act is the sender’s

expression of an attitude toward some possibly com-

plex proposition. A sender performs a communica-

tion act which is expressed by a coded message and

is directed to a receiver. Therefore, a communication

act has two main components. First, the attitude of

the sender which is called the illocutionary force (F),

that expresses social interactions such as informing,

requesting or promising, among others. And second,

the propositional content (p) which is the subject of

what the attitude is about. In COMMONT this F(p)

framework is followed, and different kinds of illocu-

tionary forces and contents leading to differentclasses

of communication acts are supported.

COMMONT is divided into three interrelated lay-

ers: upper, standards and applications, that group

communication acts at different levels of abstraction.

Classes of the COMMONT ontology are described us-

ing the Web Ontology Language OWL (more specif-

ically the OWL DL sublanguage).

In the upper layer, according to Searle’s speech

acts theory, ﬁve upper classes of communication

acts correspondingto Assertives, Directives, Commis-

sives, Expressives and Declaratives are speciﬁed. But

also the top class

CommunicationAct

is deﬁned that

represents the universal class of communication acts.

Every particular communication act is an individual

of that class. In COMMONT, components of a class

This type

style refer to terms speciﬁed in the ontol-

ogy.

ICEIS 2008 - International Conference on Enterprise Information Systems

are represented by properties. The most immediate

properties of

CommunicationAct

are the content and

the actors who send and receive the communication

act. There are some other properties related to the

context of a communication act such as the conver-

sation in which it is inserted or a link to the domain

ontology that includes the terms used in the content.

A standards layer extends the upper layer of

the ontology with speciﬁc terms that represent

classes of communication acts of general purpose

agent communication languages, like those from

KQML or FIPA-ACL. With respect to FIPA-

ACL, we can observe that it proposes four prim-

itive communicative acts (FIPA, 2005): Conﬁrm,

Disconﬁrm, Inform and Request. The terms

FIPA-Confirm

FIPA-Disconfirm

FIPA-Inform

and

FIPA-Request

are used to respectively represent

them as classes in COMMONT. Furthermore, the

rest of the FIPA communicative acts are derived from

those mentioned four primitives. Analogously, com-

munication acts from KQML can be analyzed and

the corresponding terms in COMMONT can be speci-

ﬁed. It is of vital relevance for the interoperability aim

to specify ontological relationships among classes of

different standards.

Finally, it is often the case that every single in-

formation system uses a limited collection of com-

munication acts that constitute its particular agent

communication language. The applications layer re-

ﬂects the terms describing communication acts used

in such particular information systems. The applica-

tions layer of the COMMONT ontology provides a

framework for the description of the nuances of such

communication acts. Some of those communication

acts can be deﬁned as particularizations of existing

classes in the standards layer and maybe some others

as particularizations of upper layer classes.

3 PROTOCOL DESCRIPTIONS

In order to represent protocols using OWL-DL, we

have deﬁned three different classes:

Protocol,

State,

and

Transition

, which respectively repre-

sent protocols, states and transitions in protocols.

We consider propositions that can change their

truth values due to events. A state is described by the

propositions that holds in it plus the transitions that

can take place from it. Some states are declared ﬁnal

states. A transition is associated to the communica-

tion act that is sent and to the state reached with that

transition. Only one communication act can be asso-

ciated to a transition and only one state is reached by a

transition. Then, with these descriptions, a protocol is

Figure 1: The KAMP protocol.

characterized by its initial state. An actual conversa-

tion following a protocol is an individual of the class

Protocol.

Following are some OWL axioms presented in a

logic notation

instead of the more verbose OWL/XML

syntax.

Protocol

≡ ∃

hasInitialState.State

⊓

∀

hasInitialState.State

State

≡ ∀

hasTransition.Transition

∃

hasPredicate.Proposition

⊓

∀

hasPredicate.Proposition

Transition

≡

=1 hasCommAct.CommunicationAct

⊓

=1.hasNextState.State

FinalState

⊑

State

CommunicationAct

⊑ ∀

hasSender.Actor

⊓

∀

hasReceiver.Actor

⊓

∀

hasContent.Content

⊓

∀

hasSubject.Subject

Following, we introduce one example using the

previous kind of description to represent a con-

crete protocol, in particular the KAMP protocol.

This KAMP protocol is our specialization of the

well-known protocol Request conversation policy of

KAoS (Bradshaw et al., 1997)(KAoSAppointment-

ModifyProtocol(KAMP)) and its graphical represen-

tation can be seen in Fig. 1.

Two agents take part in our protocol: Patient, la-

belled with p, and Doctor, labelled with d. First, agent

p requests agent d to modify the date of an appoint-

ment, using the communication act Appointment-

Modify, or shortly, AM (described in the COMMONT

ontology). Then, several situations can arise: On the

one hand, some communication acts can lead us to

a ﬁnal state. This happens when d agrees to modify

the date (AppointmentInform,AI), when d declines to

modify the date (AppointmentModifyDecline, AMD)

or when p withdraws the request (AppointmentMod-

ifyWithdraw,AMW). On the other hand, both agents

can keep countering one to another (Appointment-

ModifyCounter, AMC) until they reach an agreement

in the date of the appointment.

This notation is common in the Description Logics

(DL) ﬁeld. See (Baader et al., 2003) for a full explanation.

USE OF SEMANTIC TECHNOLOGY TO DESCRIBE AND REASON ABOUT COMMUNICATION PROTOCOLS

Following we present our initial description of

KAMP. Notice that the complete description of a pro-

tocol comprises a set of deﬁnitions, although the pro-

tocol is represented by the class name KAMP. More-

over, the description of a protocol is directed by the

graph that represents that protocol. Class names for

states are preﬁxed by

and class names for transi-

tions are preﬁxed by

and formed with the letters of

source and target states.

KAMP

≡

Protocol

⊓ ∃

hasInitialState.S-I

S-I

≡

State

⊓ ∃

hasTransition.T-IA

⊓ ∃

hasPredicate.PI

S-A

≡

State

⊓ ∃

hasTransition.(T-AC

⊔

T-AD

⊔

T-AB

⊔

T-AG)

⊓∃

hasPredicate.PA

S-B

≡

State

⊓ ∃

hasTransition.(T-BA

⊔

T-BD

⊔

T-BE)

⊓∃

hasPredicate.PB

S-C

≡

FinalState

⊓ ∃

hasPredicate.PC

S-D

≡

FinalState

⊓ ∃

hasPredicate.PD

S-E

≡

State

⊓ ∃

hasTransition.T-EG

⊓ ∃

hasPredicate.PE

S-F

≡

FinalState

⊓ ∃

hasPredicate.PF

S-G

≡

FinalState

⊓ ∃

hasPredicate.PG

T-IA

≡

Transition

⊓ ∃

hasCommAct.AppointmentModify

⊓∃

hasNextState.S-A

T-AC

≡

Transition

⊓ ∃

hasCommAct.AppointmentModifyDecline

⊓∃

hasNextState.S-C

T-AD

≡

Transition

⊓ ∃

hasCommAct.AppointmentModifyWithdraw

⊓∃

hasNextState.S-D

T-AB

≡

Transition

⊓ ∃

hasCommAct.AppointmentModifyCounter

⊓∃

hasNextState.S-B

T-AG

≡

Transition

⊓ ∃

hasCommAct.AppointmentInform

⊓∃

hasNextState.S-G

T-BA

≡

Transition

⊓ ∃

hasCommAct.AppointmentModifyCounter

⊓∃

hasNextState.S-A

T-BD

≡

Transition

⊓ ∃

hasCommAct.AppointmentModifyWithdraw

⊓∃

hasNextState.S-D

T-BE

≡

Transition

⊓ ∃

hasCommAct.AppointmentModifyAccept

⊓∃

hasNextState.S-E

T-EG

≡

Transition

⊓ ∃

hasCommAct.AppointmentInform

⊓

∃

hasNextState.S-G

Notice that communication acts that appear in the

transitions of the protocol are deﬁned in the COM-

MONT ontology in the following way:

AppointmentModify

≡

Request

⊓

=1 hasContent.AppointmentOverwrite

AppointmentOverwrite

≡

Overwrite

⊓∃

hasSubject.Appointment

AppointmentInform

≡

Responsive

⊓

=1 hasContent.AppointmentInformation

⊓∃

inReplyTo.AppointmentModify

AppointmentInformation

≡

Proposition

⊓∃

hasSubject.Appointment

AppointmentModifyCounter

≡

Counter

⊓

=1 hasContent.AppointmentOverwrite

⊓

=1 inCounterTo.AppointmentOverwrite

AppointmentModifyAccept

≡

⊓

=1 hasContent.AppointmentOverwrite

AppointmentModifyDecline

≡

Decline

⊓

=1 hasContent.AppointmentOverwrite

AppointmentModifyWithdraw

≡

Withdraw

⊓

=1 hasContent.AppointmentOverwrite

4 CYCLES: CAN WE MANAGE

THEM?

As we said in section 3, our OWL-DL description

of a protocol is directed by the graph that represents

that protocol. Notice also that a protocol with cy-

cles can be represented by different graphs that de-

ploy the same runs of communication acts. Unfor-

tunately, when state classes of protocols are involved

in cycles our descriptions of them does not result ad-

equate to capture equivalence relationships between

states of different graphs allowing for the same set

of runs beginning on that state (notice that the transi-

tive clousure operator of regular languages is not sup-

ported by OWL-DL).

Fortunately, we have found a way to bypass this

inconvenienceassuming that cycles in protocols mean

repetition of a deliberation process in such a way that

a complete cycle execution returns to a state class

equivalent to the initial class of the turn. Therefore,

considering the task of comparing protocol structure

in order to decide compliance of protocol runs, it is

not important the number of times a cycle is walked

around but only if it is possible to go for one turn.

Consequently, we systematically modify the descrip-

tions of protocols with cycles in order to achieve cor-

rect answers with respect to the structural relation-

ships deﬁned in section 5. It is worth mentioning that

this modiﬁcation is systematic and it is only carried

out for the reasoning process.

Moreover, we also assume that, for every proto-

col run, each communication act involved in a cycle

maintains invariant the class of its content. The repe-

tition of deliberation assumed for every cycle in a run

is represented by the fact that the content of an occur-

ring communication act in a transition involved in a

cycle, removes from the state the content of the previ-

ously occurrence of a communication act in the same

transition. A formalization of those two assumptions

is represented by the following rule.

RULE 1: e(x) in a cycle ∧ HoldsAt(P(i), t) ∧

owner(P(i),O) ∧ Happens(e(x), t) ∧ Initiates(e(x),

P(j), t) ∧ owner(P(j),O) ∧ i, j belong to the same class

→ Terminates(e(x), P(i), t)

By the owner of a proposition we refer to the

sender of the message whose content is that propo-

sition itself. Rule 1 declares that when the owner of

a proposition that is now holding causes an event that

initiates holding of a new proposition that belongs to

the same class than the previous one, and the event is

in a cycle, the previous proposition ceases to hold.

ICEIS 2008 - International Conference on Enterprise Information Systems

Figure 2: Equivalence of protocols.

5 PROTOCOL STRUCTURAL

RELATIONSHIPS

When two agents of different systems want to inter-

communicate using a particular protocol each, both

protocols must match properly. Thanks to our proto-

col description, a reasoning process can decide if two

protocols are equivalent, or if one is a specialization

of the other, otherwise they are incompatible proto-

cols. Moreover, we have developed a reasoning ser-

vice that decides if a protocol is a preﬁx, a sufﬁx or

an inﬁx of another protocol, and those relationships

taken also in the sense of equivalence or specializa-

tion. Notice also that due to the fact that protocols

are described as OWL-DL classes, structural relation-

ships among protocols can also be deﬁned in terms of

DL semantics. We write ⊢

⊑

(respectively ⊢

≡

)

to mean that class

is subsumed by class

(respec-

tively

is equivalent to

Deﬁnition 1. Two protocols A and B are equivalent

(A≡

tot

B) if ⊢A≡B.

By specialization of protocols we mean special-

ization of the classes of communication acts appear-

ing in the protocol.

Deﬁnition 2. Protocol A is a specialization of proto-

col B (A⊑

tot

B) if ⊢A⊑B.

For example, the KAMP protocol in Fig. 1 is a

specialization of the KAoS Request protocol shown

in Fig. 2a. Every communication act that appears in

the transitions of the KAMP Protocol is a subclass of

the communicationact in the corresponding transition

of the KAoS Request protocol.

In order to look for a preﬁx relationship, we

can modify a protocol description by converting

some states into ﬁnal states. Let S be a state sub-

class, we denote ﬁnal(S) to the expression resulting

from adding by conjunction the class

FinalState

to the deﬁning expression of S. For example, if

S-E

≡

State

⊓ ∃

hasTransition.T-EG

then ﬁ-

nal(

S-E

) is the expression

FinalState

⊓

State

⊓ ∃

hasTransition.T-EG

. Moreover, we can sup-

press some transitions by removing their axiom def-

initions and removing the corresponding expressions

∃

hasTransition.T

from the state deﬁnitions which

include it. We denote Prune[P/(S

...S

),(T

...T

)]

to the new protocol described by the set of de-

scriptions that results from replacing deﬁnition ex-

pressions of S

...S

, respectively, by expresssions

ﬁnal(S

)... ﬁnal(S

), and removing occurrences of

transitions T

...T

in the set of descriptions of pro-

tocol P.

Deﬁnition 3. Protocol A is a preﬁx of protocol B

(A≡

pre

B) if there exist states S

...S

and transitions

...T

in the set of equations deﬁning B such that

⊢A≡ Prune[B/(S

...S

),(T

...T

)].

Analogously, protocol A is a specialization

of a preﬁx of protocol B (A⊑

pre

B) if ⊢A⊑

Prune[B/(S

...S

),(T

...T

)].

For example, the subKAoS Request protocol

shown in Fig. 2(b) is a preﬁx of the KAoS Request

protocol in Fig. 2(a) because it is equivalent to that

protocol after removing two transitions: one from

state

to state

and another from state

to state

With respect to sufﬁx relationship, we modify a

protocol description converting some states into ini-

tial states. Let S

...S

states in the deﬁnition of

protocol P. We denote ChangeInit[P/S

...S

] to

the new protocol described by the set of descrip-

tions that results from modifying the P protocol def-

inition in the following way: adding by conjunction

∃hasInitialState.S

(i = 1. . . n).

Deﬁnition 4. Protocol A is a sufﬁx of pro-

tocol B (A≡

suf

B) if there exist states S

...S

in the set of equations deﬁning B such that

⊢A≡ChangeInit[B/S

...S

Analogously, protocol A is a specializa-

tion of a sufﬁx of protocol B (A⊑

suf

B) if

⊢A⊑ChangeInit[B/S

...S

We denote Prot[S] to a new protocol deﬁned as

Prot[S] ≡

Protocol

⊓ ∃

hasInitialState.S

Deﬁnition 5. Protocol A is an inﬁx of protocol B

(A≡

inf

B) if there exist a state S in the set of equations

deﬁning B such that A≡

pre

Prot[S] and Prot[S]≡

suf

Analogously, protocol A is a specialization of an

inﬁx of protocol B (A⊑

inf

B) if A⊑

pre

Prot[S] and

Prot[S]⊑

suf

USE OF SEMANTIC TECHNOLOGY TO DESCRIBE AND REASON ABOUT COMMUNICATION PROTOCOLS

6 CONCLUSIONS

In this paper we have presented a proposal to describe

communication protocols, represented by STS, using

the OWL-DL language. The main contribution of

this proposal is the possibility that it brings to reason

about communicationprotocols insuch a way that dif-

ferent kinds of structural relationships -equivalence or

specialization- between complete protocols or parts

of them: preﬁx, inﬁx and sufﬁx can be detected. The

formal deﬁnition of those structural relationships has

been included in the paper within a justiﬁcation of

how cycles can be managed.

REFERENCES

Baader, F., Calvanese, D., McGuinness, D., Nardi, D., and

Patel-Schneider, P., editors (2003). The Description

Logic Handbook. Theory, Implementation and Appli-

cations. Cambridge University Press.

Bechhofer, S., van Harmelen, F., Hendler, J., Horrocks,

I., McGuiness, D., Patel-Schneider, P., and Stein, L.

(2005). OWL Web Ontology Language Reference.

World Wide Web Consortium.

Berardi, D., Calvanese, D., Giacomo, G. D., Lenzerini, M.,

and Mecella, M. (2005). Automatic service composi-

tion based on behavioral descriptions. Int. J. Cooper-

ative Inf. Syst., 14(4):333–376.

Bradshaw, J. M., Dutﬁeld, S., Benoit, P., and Woolley,

J. D. (1997). Kaos: toward an industrial-strength open

agent architecture. pages 375–418.

Desai, N., Mallya, A. U., Chopra, A. K., and Singh, M. P.

(2005). Interaction protocols as design abstractions

for business processes. IEEE Trans. Softw. Eng.,

31(12):1015–1027.

d’Inverno, M., Kinny, D., and Luck, M. (1998). Interaction

protocols in agentis. In In Proceedings of the Third

International Conference on Multi-Agent Systems (IC-

MAS98), pages 261–268.

FIPA (2005). FIPA communicative act library speciﬁcation.

http://www.ﬁpa.org/specs/ﬁpa00037/SC00037J.html.

Fornara, N. and Colombetti, M. (2003). Deﬁning interac-

tion protocols using a commitment-based agent com-

munication language. In AAMAS ’03: Proceedings

of the second international joint conference on Au-

tonomous agents and multiagent systems, pages 520–

527, New York, NY, USA. ACM Press.

Kagal, L. and Finin, T. (2007). Modeling conversation poli-

cies using permissions and obligations. Autonomous

Agents and Multi-Agent Systems, 14(2):187–206.

Mallya, A. U. and Singh, M. P. (2007). An algebra for com-

mitment protocols. Autonomous Agents and Multi-

Agent Systems, 14(2):143–163.

Mazouzi, H., Seghrouchni, A. E. F., and Haddad, S. (2002).

Open protocol design for complex interactions in

multi-agent systems. In AAMAS ’02: Proceedings of

the ﬁrst international joint conference on Autonomous

agents and multiagent systems, pages 517–526, New

York, NY, USA. ACM Press.

Searle, J. R. (1969). Speech Acts: An Essay in the Philos-

ophy of Language. Cambridge University Press, New

York.

Yolum, P. and Singh, M. P. (2002). Flexible protocol speci-

ﬁcation and execution: Applying event calculus plan-

ning using commitments. In Proceedings of the 1st In-

ternational Joint Conference on Autonomous Agents

and MultiAgent Systems (AAMAS), pages 527–534.

ACM Press.

ICEIS 2008 - International Conference on Enterprise Information Systems