A SERVICE-ORIENTED FRAMEWORK FOR MAS MODELING

Wautelet Yves, Achbany Youssef and Kolp Manuel

Info. Syst. Research Unit (ISYS), University of Louvain, Belgium

Keywords:

Requirements Engineering, service oriented modeling, i*, multi-Agent system.

Abstract:

This paper introduces an analysis framework made of different models proposing complementary views. First

view - the strategic services diagram (SSD) - is the most aggregate knowledge level; it uses services for rep-

resenting the system to be in a global manner. Second view offers a more detailed perspective of the agents

involved into the services using traditional i* Strategic Dependency (SDD) and Strategic Rationale Diagrams

(SRD). Finally, the third view - the Dynamic Services Hypergraph (DSH) - offers a dynamic approach of ser-

vices realization paths at various Qualities of Service (QoS). Using those models at analysis level is interesting

in a model driven software development approach. Indeed, the project stakeholders can share a common vision

through a lookup at the services the system has to offer, at the depender and dependee agents for the services

constituents and at the different realization paths and their QoS. The framework also offers the fundamental

elements to develop a complete model driven software project management framework. It can be considered

as a startup for a global broader software engineering method.

1 INTRODUCTION

Multi-Agent Systems (MAS) have become an impor-

tant and promising research area in computer science

in general and in software engineering in particular.

Applying agent technology to huge enterprise infor-

mation systems engineering is nevertheless seldom

done. One may wonder where this report comes from.

MAS development is maybe not mature enough to

be applied in large industrial software development.

Indeed, most of the research is focused on develop-

ing particular aspects of modeling, specifying, de-

signing or implementing collaborating agents. Global

frameworks for MAS model driven development and

software project management are, until now, seldom

approached in literature. Researchers need conse-

quently to understand what dimensions are required

to bring a technology to be used on a large scale and

to ﬁll the gap between the research ﬁeld and its prac-

tical application.

This paper is part of this effort: on the basis of

the weaknesses identiﬁed in the i*/Tropos MAS de-

velopment methodology, a generic framework for or-

ganizational modeling and requirements engineering

using higher abstraction level elements: the services.

The importance of distinguishing the system behavior

at such level will be characterized and a set of mod-

els allowing services representation through multiple

complementary views will be introduced. This frame-

work is composed of a static high level services view,

of a static and a dynamic decomposed services views.

The paper is structured as follows, Section 2

points the limitations of the traditional Strategic De-

pendencyand Strategic Rationale Diagrams for model

driven software development. Section 3 points to the

use of a framework based on services that can be ex-

tended for software project management. Section 4

formalizes the new diagrams included into the frame-

work through the use of a meta-model. Section 5 con-

cludes the paper.

2 PROBLEM STATEMENT

Some modern software development methodologies

are said to be Model Driven. In such methods, the en-

tire development process is deduced from the features

modelled in high level diagrams. Object-oriented de-

velopment methodologies such as the Uniﬁed Process

(IBM, 2003; Jacobson et al., 1999; Jacobson and By-

lund, 2000; Kruchten, 2003) are for example said to

be use case driven. This means that the entire devel-

120

Yves W., Youssef A. and Manuel K. (2008).

A SERVICE-ORIENTED FRAMEWORK FOR MAS MODELING.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - ISAS, pages 120-128

DOI: 10.5220/0001708101200128

 SciTePress

opment process is driven by the use cases identiﬁed

in the analysis phases. Moreover effort estimation

techniques as Use-Case Points (Schneider and Win-

ters, 1998) directly use high level diagram elements -

the use cases - as fundamentals for effort estimation.

Such model elements called scope elements are con-

sequently usefull not only to share a common high

level vision with stakeholders, but also to estimate

the project effort on a non redundant basis, to eval-

uate related risks, to ﬁx an acceptable level of qual-

ity, etc. Classical i*/Tropos developments using tra-

ditional Strategic Dependency (SD) and Strategic Ra-

tionale (SR) diagrams (Yu, 1995) for organisational

modelling and requierements engineering do unfortu-

nately not possess ideal elements to drive the devel-

opment process, this will be shown in this section.

2.1 Related Work

As pointed out earlier, considerable work has been

done in agent-orientedsoftware developmentmethod-

ologies. The traditional i* modeling framework has

however been poorly criticized and weaknesses as the

lack of higher abstraction level elements has been

only recently pointed out in literature (Pastor et al.,

2007). Similarly, considerations on software project

management using Tropos or other MAS methodolo-

gies are not easy to ﬁnd in MAS literature. In the

same way, even if some attempts such as AUML ex-

ist, no dynamic dimension implying i* concepts has

been proposed and formalized. We willreference here

three approaches, each one addressing one of the pre-

viously mentioned aspects. We will also brieﬂy com-

ment our view in regard to the aspects described in

those papers.

• The process followed in this paper has been

strongly inspired by the concepts introduced in

(Estrada et al., 2006; Pastor et al., 2007). Those

papers deﬁne a series of features for evaluating

the i* framework on the basis of case studies. The

empirical study addresses the weaknesses of the

approach and points to the use of higher level ab-

stractions called business services into i* devel-

opments. Our conclusions are rather the same and

we take their approach as a starting point for mul-

tiple view modeling framework.

• (Dubois et al., 2007) proposes to model threats

and vulnerabilities directly into SD and SR mod-

els and relates them directly with actors and goals.

We also propose a framework where threats (and

even opportunities) are incorporated into high

level models but we relate then to services not in

SD and SR diagrams. Furthermore operational so-

lutions are incorporated to keep/improve quality

of service.

• (Mouratidis and Giorgini, 2007) adresses i* se-

curity limitations and claims for the adoption of

secure Tropos, an extention to the Tropos pro-

cess incorporating security issues. Secure Tro-

pos offers extentions for modeling requirements

with security constraints. Moreover, in that pro-

cess, security is an issue only considered at late re-

quirement stage, the methodology partially iden-

tiﬁes conﬂicts between security and other require-

ments. Our approach does not model security as

constraint but identiﬁes the potential threats to de-

termine the most risky issues at the earliest stages

of the project.

2.2 The Lack of Scope Elements

One of the main problems in using the Tropos soft-

ware development methodology is to ﬁnd an adequate

scope element to drive the software process.

Actors from the SD and SR models are involved in

dependencies of different intentional elements. The

goal is the only candidate as scope element since it

represents the most abstract functional issue. A com-

plete description of these elements can be found in

(Yu, 1995). However, goals are not considered as the

ideal scope element candidate since:

• The atomicity of the goal is not clearly deﬁned

so that a goal decomposition can involve atomic

goals part of higher level goals consequently (i)

a great amount of goals – too many for keeping

an easily understandable vision of the software

project – can be present for a huge software pro-

jet; (ii) the goals are often at different levels of

atomicity so that they can be redundant which is a

major drawback to keep the simplest vision possi-

ble;

• The distinction between a task and a goal is not

always trivial; different persons can model simi-

lar behavior by interchanging those different con-

cepts.

Consequently, we need an element located at suf-

ﬁciently high level of abstraction to represent a ser-

vice the agents should provide in a non redundant

way. This element should provide a clear vision of

the project, allow to identify environmental threats for

risk management, time management, etc. This ele-

ment is called a service and is depicted in the next

section.

A SERVICE-ORIENTED FRAMEWORK FOR MAS MODELING

121

2.3 Contributions

Compared to the approach followed into (Estrada

et al., 2006; Pastor et al., 2007), we provide:

• A complete multiple view analysis framework

for service-oriented modeling using i*.The frame-

work also offers the fundamental elements to de-

velop a complete model driven software project

management framework. It can be considered as

a startup for a global broader software engineer-

ing method;

• A dynamic model for representing services real-

ization at deﬁned quality of service level;

• Enhancements into the Strategic Services Model

to model system threats;

• A case study.

3 A UNIFIED-FRAMEWORK FOR

AGENT-ORIENTED

REQUIREMENTS

ENGINEERING

This section introduces the services approach and de-

picts the framework in an informal manner.

3.1 The Service Approach

In (Estrada et al., 2006; Pastor et al., 2007), the au-

thors describe business services as “an abstract set of

functionalities that are provided by a speciﬁc actor”,

they also specify that “an actor can be an organiza-

tional entity ... that uses or offers services”.

Services will be formalized in Section 4.2. In our

approach, they present the following properties:

• they represent the highest abstraction level entities

of the project, a service can encapsulate a large

amount of atomic agregate goals or tasks;

• they are independent but complementary to each

other in the sense that they do not overlap, but rep-

resent the whole of the system functionalities;

• they involve at highest level two super actors that

can be reﬁned in multiple sub actors at lower lev-

els;

• their realization induces of different quality of ser-

vice (QoS) areas.

Moreover, we also propose to add concepts to

model environmental elements addressing a threat –

Figure 1: A Multiple View of the Framework.

potentially lowering QoS – or an opportunity poten-

tially raising QoS. Once identiﬁed, we can also rep-

resent in a generic and in a more detailed manner,

the measures that should be taken to ﬁght against the

threats (to maintain QoS) or to put the opportunities

into practice (to raise QoS).

3.2 The Framework

To ﬁx the weaknesses pointed out previously, a

service-oriented framework for organizational mod-

eling will be introduced. This framework is made of

two complementary views represented by four differ-

ent diagrams. Figure 1 depicts the framework dia-

grams hierarchy.

The ﬁrst view called the static view allows to rep-

resent the services at two complementary abstraction

levels:

• the highest level is materialized through the

Strategic Services Diagram (SSD) described and

formalized in section 4.1. This level provides

a synthetic description of the services provided

at deﬁned QoS by the organization and that will

be operationalized into the application. Environ-

mental factors inﬂuencing positively or negatively

the provided services as well as means to opera-

tionalize them are represented here. This diagram

can be used as a reference vision document for

stakeholders as well as a basis for iteration plan-

ning, risk management, quality management, ef-

fort planning, etc.(see Section 3.3);

• the lowest level is materialized through the tradi-

tional Strategic Dependency and Strategic Ratio-

nale Diagrams as described in (Yu, 1995). This

level provides a more detailed description of the

ICEIS 2008 - International Conference on Enterprise Information Systems

122

tasks, goals and resources involved in the achieve-

ment of all the services.

The second view called the dynamic view allows

to represent the service achievement through the use

of the Dynamic Service Hypergraph (DSH) described

and formalized in Section 4.2. The goals and tasks of

the multiple agents involved are sequentialized in the

graph to represent possible paths for service realiza-

tion at given QoS.

Taken as a whole, these views allow to represent

the multiple aspects of the services the system should

offer.

3.3 The FaMOs Framework - A Project

Management Perspective

As described previously, one of the main goals of the

framework is to allow model driven software devel-

opment. The main advantage of this type of approach

is to furnish structured developments methodologies

to software professionals. The framework was con-

ceived to provide this kind of achievement but also

furnish a broader range of possibilities.

Those services include:

• Risk Management. By allowing modeling ser-

vices threats and the operational solutions de-

signed to maintain QoS when those occur, the

framework includes some essential features for

risk management. The framework includes dif-

ferent levels of detail (abstraction) for modeling

those features.

• Quality Management. Quality indicators are

present through the use of a QoS-level so that

quality benchmarks for each service can be ﬁxed

early into the project. Furthermore, through

the “services opportunities”, improvement fac-

tors that, once operationalized, allow to higher

QoS can be modelled at different levels. Finally,

the framework constitutes an extension of tradi-

tional i*/Tropos modelling diagrams so that soft-

goal modeling into the Strategic Dependency and

Strategic Rationale Diagrams remains possible.

• Time Management. Using the services as high

level abstractions and the DSH for the service

complexity evaluation, an approximate service

effort can be computed (as for example in the

Use Case Points methodology (Anda et al., 2002;

Schneider and Winters, 2001)). On the basis of

this evaluation, a project planning using waterfall

or better an iterative development life cycle can

be created. By including the number of human re-

sources, their role, their wages, etc. a project cost

evaluation can also be computed. The precision of

those evaluations will successively increase from

a project to another on the basis of the empirical

statistics collected during the previous ones.

Due to a lack of space, this paper will only fo-

cus on the models formalization. (Wautelet et al.,

2007) introduces a case study to illustrate the use

of the framework. it only uses elements of risk and

quality management but do not involve time and soft-

ware process management. A complete presentation

of the project management capabilities of the FaMOs-

framework is currently under development.

4 SERVICE DRIVEN AGENT

MODELING: A

FORMALIZATION

To drive the business and user services acquisi-

tion, we propose a meta-model which provides

modeling elements relevant for specifying both

strategic and operational aspects of the organiza-

tional context in which the future information sys-

tem will be deployed. This meta-model is made

of meta-concepts (e.g., “Actor”, “Dependencies”,

“Services”, etc.), meta-relationships relating meta-

concepts (e.g., “Performs”, “Operationalize”, “Act”,

etc.), meta-attributes of meta-concepts or meta-

relationships (e.g., “Probability of happening”, “Im-

provement rate”, etc.), and meta-constraints on meta-

concepts and meta-relationships (e.g., “An actor oc-

cupies a position if and only if that actor possesses all

the capabilities required to occupy it”).

4.1 Strategic Services Model

The Strategic Services Model (or SSM) supports the

acquisition, representation and reasoning about the

services that should be provided by the agents of the

application.

Figure 1 depicts the SSM. Actors are intentional

entities used to model people, physical devices or

software systems that are processor for some actions.

The inheritances from the Actor into Position, Agent

and Role as well as the linking with the Dependency

class have been inspired by the Tropos metamodel

presented in (Susi et al., 2005). A Position can cover

1 to n Roles, an Agent can play 0 to n Roles and can

occupy 0 to n Positions. Actors achieve some Ser-

vices which are functionalities they offer to others in

order to fulﬁl a portion of their goals and tasks (for

further formalization see section 4.2). The Service is

always achieved through an Actor Dependency. The

A SERVICE-ORIENTED FRAMEWORK FOR MAS MODELING

123

Figure 2: Strategic Services Diagram: A Meta-Model.

later relates two Actors: a depender and a dependee

as well as a dependum, the Service.

Since the aim of the diagram is to offer a high level

view of an under development enterprise information

system, there are basically two types of services that

need to be modelled: Business Services – business

processes achievedinto the company business domain

– as well as User Services – services provided to the

end user.

An Environmental Factor can be an external

Threat for the adequate fulﬁlment of the service (both

in terms of achievement and quality of service degra-

dation) or an opportunity that can potentially raise

the QoS. A service and its QoS are operationalized

into a digraph as depicted in section 4.2. Moreover a

Probability of happening is associated to each Threat

as well as an Improvement rate to each Opportunity.

Those features will be useful for the Threats and Op-

portunities prioritization into project planning.

An Operational Solution operationalizes an Envi-

ronmental Factor. In other words Threats and Op-

portunities are external factors that can lower or raise

QoS. To ﬁght against a Threat or to beneﬁt from an

opportunity measures need to be taken. Those can

be represented through the Operational Solution. The

later is specialized into the Risk Measure and the

Improvement Solution respectively aimed to model

the operational solution putted into practice to ﬁght

against a threat and to take beneﬁt of a potential op-

portunity. Note that once again an element (the oper-

ational solution) distinguished at SSD level induces a

higher level of abstraction that will be reﬁned further

into the frameworkthrough SD/SR Models and DSH.

The elements of the SSM are instantiated to pro-

vide a Strategic Service Diagram (or SSD). An SSD

presents on the highest aggregationlevel of the frame-

work the services provided by each agent as well as

the potential threats and opportunities they are facing.

It also supports the identiﬁcation of operational so-

Figure 3: Meta-Model of the Dynamic View.

lutions that can be taken to ﬁght against the Threats

potentially lowering QoS and to take beneﬁts of Op-

portunities potentially raising QoS. It shows the po-

tential problems that will be encountered when at-

tempting to achieve Services and, with the guidance

of the methodology, helps operationalizing them. By

essence, the SDD helps to understand the purpose of

the software project in terms of Services, the prob-

lems (i.e. Threats) they can face, the potential im-

provements (i.e. Opportunities) and the social struc-

ture (i.e. the Dependencies) which govern Actor in-

teractions.

4.2 Dynamic Service Model

In this section, we will bring the Dynamic Service

Model (DSM) to further formalization Figure 3 de-

picts the DSM. As the only properties of the agent

relevant for this dynamic view are the capabilities in

terms of tasks and goals it can execute along with its

advertised quality values, an agent will be deﬁned as

follows

Deﬁnition 1. A tuple

h{(cp

, q

), . . . , (cp

i+m

, q

i+m

)}, Ag

i is called

an agent a, where cp

is a capability. The agent

advertises its capability to execute a task or goal to

the QoS level and cost q

. The advertised level is

a vector of QoS- and cost-property and value pairs

following a QoS ontology. Ag

is assumed to contain

all additional properties of the agent irrelevant for

the present discussion, yet necessary when building a

MAS.

Format and content of Ag

will depend on the

agent programming language being used. A Capa-

bility is the generalization of Goal and Task. As it is

unimportant to know that an agent can perform more

Note that since, in this view, the service realization is

documented we refer to agent rather than to actors as in the

SSD.

ICEIS 2008 - International Conference on Enterprise Information Systems

124

than one capability, a specialist agent is deﬁned over

a single capability.

Deﬁnition 2. ha, cp

, q

i associating a capabil-

ity cp

to a quality level q

advertised by the

agent a is a specialist agent a

. The agent must

be capable of perfoming the capability: ∀a

ha, cp

, q

i, (cp

, q

) ∈ a.

Any agent a that can accomplish m > 1 capabil-

ities can also be seen as a set of specialist agents:

, . . . , a

i+m

}. It is necessary to have a more precise

idea of how capabilities and services are conceptual-

ized.

Deﬁnition 3. A capability cp

is hcp

pre

, τ

, cp

post

where cp

pre

describes the capability precondition, τ

is a speciﬁcation (in some language) of how the agent

is to execute the capability, and cp

post

describes the

conditions true after the capability is executed. Ca-

pabilities belong to the set CP.

Deﬁnition 4. hs

, s

, servTransit

, servState

i is a

service s

, where s

provides the details of the func-

tional speciﬁcation of the service, (s

, s

) deﬁnes

a directed acyclic graph. Nodes represent states,

and edges transitions between states. The two func-

tions label nodes and edges with capability infor-

mation: servTransit

: s

7−→ CP is a partial func-

tion returning the capability for a given edge in

the graph, while servState

: s

7−→ {cp

pre

}

∈CP

∪

{cp

post

}

∈CP

maps each edge to a condition from the

set of all capability preconditions (i.e., {cp

pre

}

∈CP

)

and postconditions (i.e., {cp

post

}

∈CP

). The capa-

bility speciﬁed on an edge must have the precondition

and postcondition corresponding to conditions given,

respectively, on its origin and its destination node.

A service can therefore be understood as a pro-

cess, composed of a set of Tasks and Goals (i.e.,

these are specializations of capability) ordered over

the graph representing the service. The functional

speciﬁcation of the service, i.e., s

is not of interest

here, but involves in practice, e.g., a speciﬁcation of

interfaces, and other implementation considerations.

Requesting a service requires the speciﬁcation of ex-

pected QoS, in addition to a deadline for providing

the service, minimal level of reputation for agents

that are to participate in service execution, the max-

imal monetary cost, and explicit user preferences on

agents to select (e.g., users may prefer globally the

services of some providers over others, regardless of

actual performance—this may occur with preferen-

tial threatment resulting from environment constraints

such as, e.g., legal constracts on cooperation between

organizations and/or individuals).

Figure 4: The service depicted as a labeled hypergraph.

Deﬁnition 5. A service request ˆs

, s

QoS

, s

cost

, s

pref

i, where:

• s

is the service to provide.

• s

QoS

speciﬁes expected qualities and their re-

quired level. Its deﬁnition follows a QoS ontology,

such as, e.g., the FIPA QoS ontology speciﬁcation

(for Intelligent Physical Agents, 2002). What-

ever the speciﬁc QoS ontology, expected quali-

ties are likely to be speciﬁed as (at least) s

QoS

h(p

, d

, v

, u

), . . . , (p

, d

, v

, u

)i, where:

– p

is the name of the QoS parameter (e.g., con-

nection delay, standards compliance, and so

on).

– d

gives the type of the parameter (e.g., nomi-

nal, ordinal, interval, ratio).

– v

is the set of desired values of the parameter,

or the constraint <, ≤, =, ≥, > on the value of

the parameter.

– u

is the unit of the property value.

• s

is a deadline, speciﬁed as a natural.

• s

speciﬁes minimal levels of reputation over task

quality parameters that any agent participating in

the provision of the given service must satisfy. It

is not necessary to specify reputation for all qual-

ities over all tasks, selective reputation expecta-

tions are admitted.

• s

cost

is the maximal monetary cost the user re-

questing the service is ready to pay to obtain the

service.

• s

pref

is a set of expressions that constrain the pool

of potential agents to which the service mediator

can allocate tasks.

By conceptualizing the service as suggested in

Def.4, the service is thus mapped onto a directed hy-

pergraph G where each node is a step in service pro-

vision and an edge in G corresponds to the execution

of a capability cp

by a specialist agent a

k,u

, where u

ranges over specialists that can execute cp

according

to the criteria set in the service request. Each individ-

ual edge is labelled by a function of the criteria set,

A SERVICE-ORIENTED FRAMEWORK FOR MAS MODELING

125

Client

ServiceProvider

Travel Planner

Denial ofService

Lower Response Time

Use

a Task Allocation

Algorithm

Agent

Service

Threat

Opportunity

Operational solution

Security

Figure 5: Strategic Services Diagram - Travel Planner.

c(cp

, a

k,u

), representing the QoS advertised by the

specialist a

k,u

for performing the capability cp

. Each

path from the starting node (e.g., node s1 in Figure 4)

to the destination node (node s13 in Figure 4) thus

corresponds to a sequence of capabilities ensuring

the completion of the service within the prescribed

QoS. The model thus assumes that there are alterna-

tive ways for completing the service. The topology

of the graph—i.e., the node structure and the capabil-

ities associated to edges between the nodes—is pro-

vided by the designer through the service deﬁnition,

so that the graph is a graphical model of the different

ways the service can be performed as a sequence of

capabilities.

5 CASE STUDY

In this section, the framework is applied to a simpli-

ﬁed case study: the travel planning service. Travel

planning regroups all the activities provided by a ser-

vice provider destined to organize a business or hol-

iday trip. Those activities include the booking of fa-

cilities such as an hotel, a ﬂight, a car, a taxi, etc.

The different dimensions of a service description and

fulﬁllment will be envisaged, their complementarities

will be illustrated.

5.1 The Strategic Services Diagram

Figure 5 depicts the Travel Planner’s Strategic Ser-

vices Diagram.

In this service, the Customer addresses to a Travel

Planner for its travel organization. The Travel Planner

is in charge of the Commodities Booking to the Re-

source Provider (which can be an hotel resort, a ﬂight

company, a taxi company, etc. i.e. all the companies

furnishing commodities to the Customer) for the des-

tination of the Customer on the basis of its Budget,

Proposed Date and Agreement.

Figure 6: Strategic Dependency Diagrams: Travel Planner.

Figure 7: Strategic Rationale Diagrams: Travel Planner.

5.2 Strategic Dependency Diagram

Figure 6 illustrates a strategic dependency diagram of

the Travel Planner system’s requirements. The system

is intended to support a customer and a travel planner

in travel related activities. Three actors are introduced

into this diagram: Travel Planner is responsible for

the organization of the travel and the necessary reser-

vations. Customer gives these preferences concerning

its travel which he wants to plan. Resource provider

deals with providing the various resources helping to

planning.

A strategic rationale diagram of the Travel Planner

system’s requirements is depicted in Figure 7. This

Figure details the various actors introduced into the

strategic dependency diagram.

5.3 Dynamic Services Hypergraph

Figure 8 represents fulﬁlment paths for the service

Travel Planner. Each node is a step in service provi-

sion and each edge corresponds to the execution of a

task t

by a specialist agent a

k,u

, where u ranges over

specialists that can execute t

according to the crite-

ria set in the service request. A function of the crite-

ria set, c(t

, a

k,u

), labels each edge and represents the

QoS advertised by the specialist a

k,u

for performing

the task t

. Note that different paths offering different

ICEIS 2008 - International Conference on Enterprise Information Systems

126

Figure 8: Strategic Dynamic Diagram - Travel Planner.

QoS are available. Indeed, path < t

offers alternative ways of fulﬁlling the service. The

following list gives a descritpion of each task used in

the Hypergraph.

• t

:The customer proposes a range of dates ﬁtting

for his travel.

• t

:The customer expresses the maximal budget he

is able to allow for his travel.

• t

:The Resource Provider gets a list of proposition

for commodities on the basis of the customer con-

straints.

• t

:The customer selects its preferredcommodities.

• t

:On the basis of the constraints deﬁned, the Ser-

vice provider organises a trip plan for the cus-

tomer. Finally, the Service Provider books the

commodities (i.e., ﬂight, hotel, ...).

• t

:The customer accepts a list (date, budget, ﬂight

and hotel) proposed.

• t

:The Resource Provider provides the resource to

the customer.

5.4 Lessons Learned

This simpliﬁed case study allowed illustrating:

• The use of different levels of knowledge:

– A ﬁrst level materialized through the SSD

where stakeholders can have an aggregate view

of the services the system offers as well as the

threats and opportunities;

– A second level materialized through the SDM,

SDM and DSD to decompose the service in a

static and dynamic manner. In this view ana-

lysts can see respectively the actors involved as

well as the multiple realization paths.

• The alternative paths in the hypergraph with dif-

ferent QoS.

To demonstrate the real contribution of the frame-

work it should however be applied to a broader case

study implying multiple services. Due to the lack of

space this was not possible here.

6 CONCLUSIONS

As identiﬁed several times in literature, the Tropos

methodology in general and i* framework in partic-

ular lacks proper elements for driving the soft pro-

cess. Indeed, even if the qualities and advantages of

the strategic dependency and strategic rationale mod-

els has been recognized, redundancy, various levels

of aggregation and alternative modeling ways make

them hard to use in the context of model driven soft-

ware development. That is why a broader framework

for analysis stages of i*/Tropos is needed.

The framework furnishes multiple views for

model driven software development. This framework

includes a high level service view allowing to iden-

tify in a non redundant way the services that must be

fulﬁlled by the system as well as elements raising or

lowering QoS. Such framework is much broader than

only allowing organisational modeling and require-

ments engineering. Indeed the framework can be ex-

tended for risk, quality or time management. The aim

is consequently to create a complete software project

management dimension for Tropos based on the mod-

els of the framework; this work is in progress; a CASE

Tool is also under development.

Future research directions will include systematic

forward engineering rules based on the models in

the framework. This work has partially been done

in (Do et al., 2003) but needs to be extended to in-

clude the beneﬁts of dynamic models as the Strategic

Services Hypergraph. Such systematic approach will

provide guidelines to software designers to facilitate

the framework adoption.

REFERENCES

Anda, B., Angevik, E., and Ribu, K. (2002). Improving es-

timation practices by applying use case models. PRO-

FES, pages 383–397.

Do, T., Kolp, M., and Pirotte, A. (2003). Social patterns for

designing multi-agent systems. In Proceedings of the

15thInternational Conference on Software Engineer-

ing and Knowledge Engineering (SEKE 2003), San

Francisco, USA, July.

Dubois, E., Rifaut, A., and Mayer, N. (2007). Improv-

ing risk-based security analysis with i*. in Giorgini

P., Maiden N., Mylopoulos J., Eric Yu editors, Social

Modeling for Requirements Engineering, in Coopera-

tive Information Systems series, MIT Press.

Estrada, H., Rebollar, A., Pastor, O., and Mylopoulos, J.

(2006). An empirical evaluation of the i* framework

in a model-based software generation environment.

Proceedings of CAiSE, pages 513–527.

for Intelligent Physical Agents, F. F. (2002). Fipa quality of

A SERVICE-ORIENTED FRAMEWORK FOR MAS MODELING

127

service ontology speciﬁcation. Doc.No. SC00094A.

FIPA (http://www.ﬁpa.org).

IBM (2003). The rational uniﬁed process. Rational Soft-

ware Corporation, Version 2003.06.00.65.

Jacobson, I., Booch, G., and Rumbaugh, J. (1999). The uni-

ﬁed software development process. Addision-Wesley.

Jacobson, I. and Bylund, S. (2000). The road to the uniﬁed

software development process. Cambridge University

Press.

Kruchten, P. (2003). The rational uniﬁed process : An in-

troduction. Longman (Wokingham), Addison-Wesley,

December.

Mouratidis, H. and Giorgini, P. (2007). Extending i* and

tropos to model security. in Giorgini P., Maiden N.,

Mylopoulos J., Eric Yu editors, Social Modeling for

Requirements Engineering, in Cooperative Informa-

tion Systems series, MIT Press.

Pastor, O., Estrada, H., and Mart´ınez, A. (2007). The

strengths and weaknesses of the i* framework: an ex-

perimental evaluation. in Giorgini P., Maiden N., My-

lopoulos J., Eric Yu editors, Social Modeling for Re-

quirements Engineering, in Cooperative Information

Systems series, MIT Press.

Schneider, G. and Winters, J. (1998). Applying use cases -

a practical guide. Addison Wesley Longman, Inc.

Schneider, G. and Winters, J. (2001). Applied use cases.

Second Edition, A Practical Guide, Addison-Wesley,

Reading, MA.

Susi, A., Perini, A., Giorgini, P., and Mylopoulos, J. (2005).

The tropos metamodel and its use. Informatica,

29(4):401–408.

Wautelet, Y., Achbany, Y., and Kolp, M. (2007). A

uniﬁed framework for multi-agent services-

driven organisational modeling. http://

www.isys.ucl.ac.be/staff/yves/articles/FAMOS07.pdf

Yu, E. (1995). Modeling strategic relationships for process

reengineering. PhD thesis, University of Toronto, De-

partment of Computer Science, Canada.

ICEIS 2008 - International Conference on Enterprise Information Systems

128