COORDINATION AND ORGANISATIONAL MECHANISMS

APPLIED TO THE DEVELOPMENT OF A DYNAMIC,

CONTEXT-AWARE INFORMATION SERVICE

Manel Palau, Luigi Ceccaroni

TMT Factory, C/ Moià 1, planta 1, E08006, Barcelona, Spain

Ignasi Gómez-Sebastià, Javier Vázquez-Salceda, Juan Carlos Nieves

Knowledge Engineering & Machine Learning Group, UPC, C/Jordi Girona 1-3, E08034, Barcelona, Spain

Keywords: Multi-agent systems, Coordination and organisational theory, Context awareness, Personalised

recommendation, Semantic Web services.

Abstract: A multi-agent system design-methodology is used to address the highly dynamic, regulated, complex,

distributed environment of interconnected services. A framework of three interconnected levels is applied to

tackle this issue relying on coordination and organisational techniques, as well as on Web-services and new

methodologies to design, deploy and maintain a distributed system. This paper presents results based on a

real use case: interactive community displays with touristic information and services, dynamically

personalised according to user context and preferences.

1 INTRODUCTION

Urban information services are often provided in

ways which have not changed much in a century.

This scenario should inevitably evolve, bringing up

the opportunity to improve services provided to

people living in or visiting a city, with the novel

possibility of ubiquitously accessing personalised,

multimedia content (Greenfield, 2006).

On the one hand, this scenario presents

numerous, dynamic services that have to be

composed and coordinated in order to provide

higher-value services. For instance, an advanced

entertainment service can be provided by combining

information coming from cinema, restaurant and

museum services, along with transport and mapping

services. These services are not static, as existing

services can leave the system and new ones can

enter it; the service used in a given moment for a

given task might not be available later or it might

happen that a more suitable service becomes

suddenly available.

On the other hand, the scenario has to be able to

filter and adapt the incoming content, making it

compatible with user’s preferences and location, and

with existing regulations. For example, a

recommendation system should not suggest a pub to

an underage user if local laws do not allow it.

An additional challenge for systems in highly

dynamic environments, where unexpected events

can arise at any time (e.g., transport not in time due

to a traffic jam), is to be able to react and adapt to

these events.

We consider that this complex context can

benefit from the combination of multi-agent

techniques, semantic Web services and machine

learning (Comas et al., 1999) to enable dynamic,

context-aware service composition (Vallée et al.,

2005), thus providing users with relevant high-level

services depending on their current context.

Moreover, technologies concerning organisational

and coordination theories applied to (intelligent)

Web services (Luck et al., 2006) are also important

in order to effectively maintain a system operating in

such a constrained (due to user’s preferences and

local laws) and dynamic environment.

Additionally, the scenario presents the need of

integrating new functionalities, new services or new

actors (humans or AI systems) into an existing

running system. This integration is especially

difficult taking into account that the scenario

Palau M., Ceccaroni L., Gómez-Sebastià I., Vázquez-Salceda J. and Carlos Nieves J. (2010).

COORDINATION AND ORGANISATIONAL MECHANISMS APPLIED TO THE DEVELOPMENT OF A DYNAMIC, CONTEXT-AWARE INFORMATION

SERVICE.

In Proceedings of the 2nd International Conference on Agents and Artiﬁcial Intelligence - Agents, pages 88-95

DOI: 10.5220/0002731600880095

 SciTePress

presents a system formed by active, distributed and

interdependent processes.

In the ALIVE European project a new software-

engineering methodology is being explored

(Vázquez-Salceda et al., 2009). This approach has as

objective to bring together leading methods from

coordination technology, organisation theory and

model driven design (MDD) to create a framework

for software engineering addressing a new reality

composed of live, open systems of active services.

Examples of these services are the ones described in

this paper or the ones involved in the crisis

management scenario described in Quillinan et al.

(2009). The ALIVE’s framework is a multi-level

architecture composed of three levels:

 the organisational level, which provides

context for the other levels, supporting an

explicit representation of the organisational

structure (composed by patterns and rules) of

the system, and effectively allowing a

structural adaptation of distributed systems

over time;

 the coordination level, which provides the

means to specify, at a high level, the patterns

of interaction among services, transforming

the organisational representation (including

information flows, constraints, tasks and

agents) coming from Organisational Level

into coordination plans; and

 the service level, which allows the semantic

description of services and the selection of the

most appropriate services for a given task

(based on the semantic information contained

in the service description), effectively

supporting high-level, dynamic service

composition.

In this paper, results of the application of the

ALIVE’s approach (see section 3) to the design of

dynamic, adaptive systems are presented. In

particular, the design for each level is described (see

section 4). First, the complete design of the

organisational level of an interactive community

display (ICD) scenario (see section 2) is presented.

This design shows that the correct identification of

roles in the organisational level allows a dynamic

adaptation in the coordination level. Then, the

complete design of the coordination level of the ICD

scenario is outlined. This design shows how, given a

set of landmarks (which are states of special

interest), a system can be dynamically adaptable.

Finally, the specifications of the Web services

involved are presented. In Sections 5 and 6, a

discussion of related and future work is outlined.

2 THE SCENARIO

We will use a personalised recommendation tool for

entertainment and cultural activities as the basis for

exemplifying the scenario. The personalisation is

offered via ICDs, which are multimedia information

points offering interactive services in public areas

(Ceccaroni et al., 2009; Gómez-Sebastià et al.,

2009). The aim is to bring city services closer to

people living in or visiting a city by interconnecting

people, service providers and locations. In the

scenario, it is taken into account that services and

information provided, and how user information is

stored, processed and distributed, are all subject of

various municipal, national and European

regulations.

The starting point of the scenario is a user

interacting with the system’s interface (the ICD) in

search for entertainment or cultural activities around

the city. The user identifies herself. Then the system

accesses the user profile (if available) from a remote

repository, and adapts the interface format and the

interaction mode according to user preferences. If

the user requests a service, the system composes an

initial recommendation considering user preferences,

requirements and, above all, time and location. The

environmental context includes components such as

weather and traffic reports. Ratings and reviews

about venues (restaurants, cinemas, shops and

museums) are also taken into account, as well as

legislation. Finally, activities related to the service

requested are presented, located on a map together

with basic information, such as a brief description,

address and pictures.

When the user requests information about a

venue, for instance a cinema, the system shows its

detailed description (e.g., movies, sessions and

prices). Moreover, it informs on transportation (e.g.,

bus and metro) to reach the venue and, if time is

appropriate, it suggests a restaurant along the way,

thus composing information from different services

(cinemas, restaurants, maps and buses).

3 THE ALIVE FRAMEWORK

The ALIVE framework is being developed in

collaboration by several universities and enterprises

within the frame of European project ALIVE. It

combines MDD and agent-based system engineering

with coordination and organisational mechanisms,

providing support for “live” (that is, highly

dynamic)

and open systems of services. ALIVE’s

COORDINATION AND ORGANISATIONAL MECHANISMS APPLIED TO THE DEVELOPMENT OF A DYNAMIC,

CONTEXT-AWARE INFORMATION SERVICE

Figure 1: Main components of the ALIVE architecture.

multi-level approach (see Figure 1 and following

sections) helps to design, deploy and maintain

distributed systems by combining, reorganising and

adapting services. As shown in Section 4, this

framework is suitable for scenarios with new

services entering the system and existing services

leaving it at run-time.

3.1 Organisational Level

The organisational level provides an explicit

representation of the organisational structure of the

system. The organisational model is the main

component of the organisational level, representing

the organisation as a social system created by

autonomous actors (i.e. they have their own

interests) to achieve common goals. Stakeholders

and their relations are represented, together with

formal goals, requirements and restrictions. The

model is formalised according to the Opera

methodology (Dignum, 2004). The model includes

objectives, or goals, of the organisation (e.g. receive

personalised content, choose a suitable content

provider or provide content); the roles (e.g. user,

service broker or content provider) that are groups of

activity types played by actors (i.e. the agents, or

human users); and the landmarks (e.g. content

provider chosen and content provided).

Objectives are assigned to roles, among which

three kinds of relations exist: the hierarchical

relation, where a parent role can delegate an

objective to a child role; the market relation where a

child role can request the assignment of an objective

to the parent role; the network relation, where both

parent and child roles can request an objective to the

other one. Each relation is assigned according to

what interaction type the designer expects to happen.

For instance, in the presented scenario, there is a

market relation when the service broker looks for the

most suitable content provider. The set of all roles

and the relations among them is the social structure.

Landmarks are important states in the

achievement of a goal, and landmark patterns

impose an ordering over landmarks to be reached. A

set of landmarks and their relations is known as

scene (see

Figure 5). Relations between scenes (i.e.

scene transitions) can be modelled by organising

them in an interaction structure (see

Figure 4).

The organisational level supports the definition

of norms, rights and obligations of the actors

forming a normative structure. Norms are suitable

for highly regulated scenarios like the one presented

(e.g. deadlines on user-system interactions and strict

regulations applying to users’ personal data).

The social structure, the interaction structure

and the normative structure are the three

components of the organisational model.

The OperettA tool (Okouya and Dignum, 2008)

supports system designers in specifying and visually

analysing an organisational model, whereas, the

model checker verifies its consistency. The domain

ontology represents the shared understanding of the

domain, providing a common vocabulary about all

concepts and their properties, definitions, relations

and constraints, and can be defined using existing

ontology-editors (e.g., Noy et al., 2003; Ceccaroni

and Kendall, 2003).

The organisational model is used by the multi-

agent system (MAS) generator in the coordination

level to create the agents that populate the system.

For each role defined in the organisational model

one or more agents are generated. The plan

synthesiser uses information from the organisational

model and the domain ontology in order to generate

the plans the agents will enact. These plans are

stored in the plan repository where they are

retrieved when required.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

3.2 Coordination Level

The coordination level provides the patterns of

interaction among actors, transforming the

organisational model into coordination plans, or

workflows. Workflows are defined using

(generalised) partial global planning (GPGP), a

framework for coordinating multiple AI systems that

are cooperating in a distributed network (Lesser et

al., 2004). Workflows bring the system from a

landmark state to the next one (see

Figure 2). Tasks

contain both pre- and post-conditions that describe

the state of the system before and after the task is

performed.

Figure 2: Workflow example, connecting two landmarks.

A set of intelligent agents (MAS) deployed on

the AgentScape platform (Overeinder and Brazier,

2004) enacts the workflows in a coordinated and

distributed fashion. Agents analyse and monitor

workflow execution, reacting to unexpected events,

either by enacting other workflows or by

communicating the incident to other levels.

Each agent includes the following components:

the brain module, implemented in 2APL (Dastani et

al., 2007), which provides reasoning and decision-

making capabilities; the normative plan analyser,

which scans the workflows in order to determine if

enacting them will violate any of the norms defined

in the organisational model; the ACL module, which

provides agents with the capability of

communicating with other agents in the system by

sending messages; the GPGP scheduler, which

provides an interface for the agents to coordinate

and distribute tasks; and the enactment component,

which facilitates the invocation of services.

3.3 Service Level

Appropriate services are selected for each abstract

task in the workflows, using the information

included in the service description and in the task

description. These descriptions are defined in terms

of OWL-S service profiles (Martin et al., 2004),

facilitating the process of composing services (Sirin

et al., 2004) and finding alternative services. The

reassignment of services to tasks, when a given

service is not available, is carried out on the fly.

The match maker component receives an

abstract task description from an agent and looks for

services that can fulfil this task. It queries the service

directory and selects the most appropriate one (if

several ones are available), based on the task’s

semantic description and on quality of service

parameters (such as average response time). The

service chosen is returned to the agent, and the task

is executed and monitored.

At the service level, service composition within

the scope of a task is carried out, too. For instance, if

a given task requires providing information of a

venue on a map, and there are two available

services, one to obtain venue information and

another to show information on a map, then the task

can be bound to the composition of these services.

4 MODELLING THE SCENARIO

This section describes how the ALIVE framework

has been applied to the scenario presented in Section

2, and how the system manages user’s petitions by

means of the defined roles, objectives, scenes,

norms, landmarks and workflows.

An organisational social structure is defined

using OperettA (see

Figure 3). It includes several

(external/internal) roles (e.g. content adaptor and

content provider) in the domain, and their

dependencies in terms of delegated sub-goals (e.g.

provide content). Roles are represented as nodes and

sub-objective dependencies as directional arrows.

The possible objectives or goals considered for the

user role (sign up, sign in, receive personalised

content, change profile, interests or requirements

and change the interface format or the interaction

mode) have been related with the petitions to be

managed by the system. Each user’s goal is

subdivided into sub-goals and delegated to other

roles. These roles can delegate sub-goals to other

roles, too.

Figure 3 shows how user’s objectives presented

above are delegated to the interface that collects the

petition (sub-objective collect_petitions) and the

interaction task manager that manages them (e.g.

sub-objective manage_sign_up_petition). Then, in

the case of sign up, the petition is managed by

subdividing and delegating it to the user modeller

(create_new_profile) and the authenticator

(assign_open_id).

The sign in petition is delegated to roles

authenticator (validate_user), user modeller

(assign_session_preferences). Roles interface

(adapt_interface_format) and multimodal interaction

COORDINATION AND ORGANISATIONAL MECHANISMS APPLIED TO THE DEVELOPMENT OF A DYNAMIC,

CONTEXT-AWARE INFORMATION SERVICE

Figure 3: Operetta model of the social structure, showing roles (nodes) and their sub-objective dependencies (arrows).

manager (adapt_interaction mode) can also get

delegated objectives if the output is to be adapted to

user’s preferences and device features.

The objectives of the content adaptor role are

collect, personalise and filter (if necessary) the

content according to laws, user preferences/

requirements and context (information such as the

weather conditions). To this end, the role relies on

information provided by other roles: the interface

(providing information about the user context, such

as localisation), the user modeller (providing the

user’s interests and requirements gathered from the

user model) and external information-providers

(providing content, and information related to the

environmental context and the legislation). The

connection with these external information-

providers is performed by the service broker.

The interaction structure for the chosen scenario

(see

Figure 4) is composed by four scenes

representing the petitions requested by the users.

The transitions among these scenes, starting on the

left circle and finalising on the right one, show

several sequences of interactions. The user model

creation scene contemplates the sign up petition,

while, within the user model adaptation scene, the

updating of user’s profile, interests and requirements

is managed.

The interaction interface adaption scene

contains the following petitions: sign in (just for

registered users), new interface format and new

interaction mode. From then on, the information

about the user can be either uploaded or not,

depending on her authentication status (either

authenticated or not authenticated). Finally, the

content adaption scene contemplates the content

petition, which happens in a state where the

interaction and interface preferences have been

loaded, so interface format and interaction mode are

already adapted.

Figure 4: Interaction structure.

Each scene can be defined by a landmark pattern

imposing an ordering over the important states

(landmarks) that should be reached in the

achievement of the goals in the scene. For instance,

within the content adaption scene, the system passes

from one landmark (state) to the next one as

depicted in

Figure 5.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

Figure 5: Landmark pattern of the content adaption scene.

The plan repository within the coordination level

contains the workflows the agents must follow in

order to accomplish the landmarks defined in the

organisational model. Focusing now in the content

adaption scene (see

Figure 8), there is a distributed

plan where several agents coordinate performing

tasks in parallel, in order to maximise the

performance. For instance, the interface provides

user’s context and the user modeller provides user’s

interests and requirements (if available).

Figure 6: Norm example.

There are several norms applied to the

interaction among roles. For instance, as seen in

Figure 6, the content provider has the obligation of

performing his task before a deadline (10 slots time).

This norm has the objective of preventing the user

from having to wait too long before its petition is

processed.

Then, the content adaptor personalises the

petition and collects content taking into account the

gathered user interests and requirements. Meanwhile

legal body and environmental context manager will

provide external information such as legal info or

weather and traffic reports. Once all information has

been gathered, including content information

coming from content providers as well as users’ and

external information, the content adaptor role filters

and adapts the content according to laws, user’s

model and current state (weather forecast, traffic,

actual time…). The content is composed on a list

and on a map to finally provide it to the user through

the interface role.

Figure 7: Mapping task to service.

Some tasks are implemented as web services

which are semantically annotated; describing them

in terms of OWL-S service profiles (i.e. inputs,

outputs, preconditions and effects). In order to

perform the task “choose a suitable content

provider” the matchmaking component maps

“abstract content providers” to concrete ones (e.g.

“Google cinemas”). Doing it this way allows system

to readapt dynamically to failures of a “concrete

content provider” (e.g. remapping to “Yahoo

cinemas” if “Google cinemas” is not available).

Figure 7 shows an example of this mapping.

The modelling presented in this section covers

the levels presented in Section 3: organisational

level, coordination level, and service level.

5 RELATED WORK

Several tourism-related projects, such as E-travel

(Gordon and Paprzycki, 2005) and Deep Map

(Malaka and Zipf, 2000), take advantage of agent

technology integrated with the semantic Web. The

ALIVE project is also close to the Interactive

Collaborative Information Systems (ICIS) project

(Ghijsen et al., 2007), in which a MAS takes into

consideration unexpected events that happen in the

real world in order to obtain a steady and reliable

system in dynamic and changing environments, and

a higher level view is used to take advantage of the

service orchestration and the re-planning.

6 FUTURE WORK

Future work will be on integrating further services,

such as booking, payment or planning routes.

COORDINATION AND ORGANISATIONAL MECHANISMS APPLIED TO THE DEVELOPMENT OF A DYNAMIC,

CONTEXT-AWARE INFORMATION SERVICE

Figure 8: Workflows to achieve the state “content provided”.

Considering the unexpected events of the real world,

for instance, it might happen that user does not

arrive on time to a booked cinema session because

of a traffic jam. In this case alternatives must be

provided to the user, for instance, booking a ticket in

a session that starts a couple of hours later, and

suggesting some nearby shops to kill some time

while the new session starts. Furthermore, work on

the integration of on-time reorganisation

mechanisms and Model Driven Design will be

performed to be able to promote reliability and

stability for services, enabling to keep slowly

changing elements separate from dynamic aspects of

the environment.

7 CONCLUSIONS

Users’ presence and context can be exploited to

provide personalised, dynamic and composed

services fulfilling their expectations, needs and

functional diversity. As seen on Section 3, this paper

presents the design of a multi-agent system that

adapts its behaviour according to the environment

and the user, and takes the initiative to make

suggestions and proactive choices.

Dynamic service-composition is an issue that has

been tackled via pre-defined workflow models

where nodes are not bound to concrete services, but

to abstract tasks at runtime. This work presents a

similar approach (through the mapping performed

by the matchmaker component) with the difference

that workflows used are not predefined, but

dynamically generated from the information

provided by an organisational level, and thus,

workflows evolve as the organisational information

evolves. Due to the connection among levels, a

change in the organisational level can trigger

changes both in the coordination level (via plan and

agent generators) and in the service level (new plans

will result in the execution of new tasks and,

possibly, the invocation of new services).

As outlined in Section 4, intelligent agents at the

coordination level present an option for providing

both exception handling and organisational-

normative awareness capabilities to the system.

Exception handling is common in other SOA

architectures. However, most approaches tend to

focus on low-level (i.e. service) exception handling.

The ALIVE approach enables managing of

exceptions at multiple levels either substituting

services (service level) looking for alternative

workflows to connect two landmarks (coordination

level) or even looking to achieve alternative

landmarks among the same scene (organisational

level). Agents at coordination level enable this

medium and high-level exception handling which

are not commonly seen in other SOA approaches.

Regarding organisational-normative awareness, to

the best of authors’ knowledge, no attempts have

been made to include normative information in

workflows. However, normative agents are common

in the literature (Castelfranchi et al., 1999). Making

normative agents reason about the workflows (and

the tasks included in them) before performing them,

and discarding the ones that do not comply with

organisational norms, adds organisational awareness

to the execution of the workflows.

ACKNOWLEDGEMENTS

This work has been partially supported by the FP7-

215890 ALIVE project. Javier Vázquez-Salceda's

work has been also partially funded by the Ramón y

Cajal program of the Spanish Ministry of Education

and Science. The authors would like to acknowledge

the contributions of their colleagues from the

ALIVE consortium (http://www.ist-alive.eu). The

views expressed in this paper are not necessarily

those of the ALIVE consortium.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

REFERENCES

Castelfranchi, C., Frank, D., Jonker, C. and Treur, J.

(1999). Deliberative Normative Agents: Principles

and Architecture. Proceedings of ATAL’99, 6

International Workshop on Agent Theories,

Architectures and Languages. LNCS, vol. 1757, pp.

364-378. Springer.

Ceccaroni, L., Codina, V., Palau, M. and Pous, M. (2009).

PaTac: Urban, ubiquitous, personalized services for

citizens and tourists. Proceedings of ICDS’09

conference, 3

International Conference on Digital

Society, pp. 7-12. IEEE Computer Society.

Ceccaroni, L. and Kendall, E. (2003). A graphical

environment for ontology development. Proceedings of

the second international joint conference on

autonomous agents and multiagent systems (AAMAS

2003), pp. 958-959. ACM New York, NY, USA.

Comas, Q., R-Roda, I., Ceccaroni, L. and Sànchez-Marrè,

M. (1999). Semiautomatic learning with quantitative

and qualitative features. Proceedings of the VIII

Conference of the Spanish Association for Artificial

Intelligence (CAEPIA-99), pp. 17-25. Asociación

Española para la Inteligencia Artificial, Spain.

Dastani, M., Hobo, D. and Meyer, J.C. (2007). Practical

extensions in agent programming languages.

Proceedings of AAMAS'07, 6

International joint

conference on Autonomous Agents and Multiagent

Systems. Article No. 138. ACM.

Dignum, V. (2004). A Model for Organizational

Interaction: Based on Agents, Founded in Logic. SIKS

Dissertation Series 2004-1. Utrecht University. PhD

Thesis.

Ghijsen, M., Jansweijer, W. and Wielinga, B.B. (2007).

Towards a Framework for Agent Coordination and

Reorganization, AgentCoRe. Proceedings the

COIN’07, International Workshop on Coordination,

Organizations, Institutions, and Norms in Agent

Systems. LNCS, vol. 4870, pp. 1–14. Springer.

Gómez-Sebastià, I., Palau, M., Nieves, J.C., Vázquez-

Salceda J. and Ceccaroni, L. (2009). Dynamic

orchestration of distributed services on interactive

community displays: The ALIVE approach.

Proceedings of PAAMS’09, 7

International

Conference on Practical Applications of Agents and

Multi-Agent Systems. ASC, vol. 55, pp. 450-459.

Springer.

Gordon M. and Paprzycki M. (2005). Designing Agent

Based Travel Support System. Proceedings of

ISPDC’05, 4

International Symposium on Parallel

and Distributed Computing, pp. 207-216. IEEE

Computer Society.

Greenfield, A. (2006). Everyware: The Dawning Age of

Ubiquitous Computing (1

ed.). New riders publishing

in association with AIGA.

Lesser, V., Decker, K., Wagner, T., Carver, N. and

Garvey, A. (2004). Evolution of the GPGP/TÆMS

Domain-Independent Coordination Framework.

Autonomous Agents and Multi-Agent Systems, vol. 9,

num. 1, pp. 87-143. Kluwer Academic Publishers.

Luck, M., McBurney, P., Shehory, O. and Willmott, S.

(2005). Agent Technology: Computing as Interaction.

A Roadmap for Agent Based Computing. AgentLink.

Malaka, R. and Zipf, A. (2000). Deep Map challenging IT

research in the framework of a tourist information

system. Proceedings of ENTER 2000, 7

International

Congress on Tourism and Communication, pp.15-27.

Springer.

Martin, D., Burstein, M., Hobbs, J., Lassila, O.,

McDermott, D., McIlraith, S., Narayanan, S.,

Paolucci, M., Parsia, B., Payne, T. R., Sirin, E.,

Srinivasan, N. and Sycara, K. (2004). OWL-S:

Semantic Markup for Web Services. Technical Report,

Member Submission, W3C.

Noy, N.F., Crubezy, M., Fergerson R.W., Knublauch, H.,

Tu, S.W., Vendetti, J. and Musen, M.A. (2003).

Protege-2000: an open-source ontology-development

and knowledge-acquisition environment. AMIA Annu

Symp Proc. 2003; 2003: 953.

Okouya, D.M. and Dignum, V. (2008). OperettA: A

prototype tool for the design, analysis and

development of multi-agent organizations.

Proceedings of AAMAS’08, 7

International

Conference on Autonomous Agents and Multiagent

Systems, pp. 1677-1678. ACM.

Overeinder, B. J. and Brazier, F. (2004). Scalable

middleware environment for agent-based internet

applications. Proceedings of PARA’04, Workshop on

State-of-the-Art in Scientific Computing. LNCS, vol.

3732, pp. 675–679, Springer.

Quillinan, T. B., Brazier, F., Aldewereld, H., Dignum, F.,

Dignum, V., Penserini, L. and Wijngaards, N. (2009).

Developing Agent-based Organizational Models for

Crisis Management. Proceedings of AAMAS-IT’09,

International conference on Autonomous Agents

and Multiaget Systems.

Sirin, E., Parsia, B. and Hendler, J. (2004). Filtering and

selecting semantic web services with interactive

composition techniques. IEEE Intelligent Systems,

vol. 19, num. 4, pp.42-49. IEEE Computer Society.

Vallée, M., Ramparany, F. and Vercouter, L. (2005). A

Multi-Agent System for Dynamic Service Composition

in Ambient Intelligence Environments. Proceedings of

PERVASIVE’05, 3

International Conference on

Pervasive Computing. Doctoral Colloquium.

Vázquez-Salceda, J., Ceccaroni, L., Dignum, F.,

Vasconcelos, W., Padget, J., Clarke, S., Sergeant, P.

and Nieuwenhuis, K. (2009). Combining

Organisational and Coordination Theory with Model

Driven Approaches to develop Dynamic, Flexible,

Distributed Business Systems. Proceedings of

Digibiz’09, 1

International Conference on Digital

Businesses. Springer.

COORDINATION AND ORGANISATIONAL MECHANISMS APPLIED TO THE DEVELOPMENT OF A DYNAMIC,

CONTEXT-AWARE INFORMATION SERVICE