Multi-agent Systems and Ontologies Applied to New Industrial Domains

Case Study: Ornamental Plants

Lorena Otero-Cerdeira, Francisco J. Rodr

ıguez-Mart

ınez,

Tito Valencia-Requejo and Loxo Lueiro-Astray

Laboratorio de Inform

atica Aplicada 2, University of Vigo, Campus As Lagoas, Ourense, Spain

Keywords:

Multi-agent System, Heterogeneity, Information Integration, Ontology.

Abstract:

This paper describes a real solution applied to an enterprise of ornamental plant selling and distribution. The

platform that we propose uses intelligent agent technologies and ontologies to meet the special needs of an

enterprise of this kind. We present the architecture deﬁned with the agents involved in both parties, the plant

wholesaler and the plant producers. A description of the ontologies that these agents use to interact is also

provided. In the ﬁnal section some relevant issues detected and conclusions will be presented.

1 INTRODUCTION

The market of ornamental plants has always been

very important in northwestern Spain and yet it is

one of the sectors where fewer technological change

has been incorporated in recent years. This paper de-

scribes a real solution applied to an enterprise of or-

namental plant selling and distribution. This enter-

prise, a plant wholesaler, needs to interact daily with

the producers of the plants they sell and with the cus-

tomers that buy them. Our platform uses intelligent

agent technologies and ontologies to meet the special

needs of an enterprise of this kind.

In business context it is essential to track all in-

formation and business processes, even more, since

this enterprise is working with a perishable product

as plants are, making an efﬁcient and just-in-time con-

trol is crucial.

The platform that has been designed allows the en-

terprise to carry out tasks such as, Production plan-

ning, Order management logistics and Catalog pro-

cessing.

When taking into account all the requirements and

issues of the enterprise, no software product or plat-

form was found that integrated all of them in a ful-

ﬁlling way. So there was the need to develop a com-

pletely new platform from scratch to satisfy all the

requirements of such a system without the help or ex-

pertise that the existence of other software products

or platforms would provide. Here lies the strength

and the importance, but also the main difﬁculty of this

project.

The development of such a platform was ad-

dressed as a multi-agent system (MAS), which we

chose to divide as well in different smaller MAS for

the sake of the design and categorization, obtaining

this way a layered representation of the MAS.

MAS have usually been used in simulation ap-

plications, real-world interactions or adaptive struc-

tured information systems (Valckenaers et al., 2006).

Within these contexts MAS have been used for ﬁre ac-

cident detection(Gowri et al., 2010), emergency evac-

uation simulation (Sharma, 2009)(Murakami et al.,

2002), meetings planning (Macho et al., 2000), etc.,

but never to develop a platform as the described in

this paper.

It is important to state that not the whole system

was addressed as a MAS, other pieces of software

were developed to cover aspects of the system where

the features of the MAS were not needed, such as hu-

man resources management or accounting. For the

purposes of this paper only the part of the system re-

garding the MAS will be described

The platform is divided into three different MAS.

The Plant Wholesaler MAS is composed by the agents

that take care of the enterprise’s tasks mentioned

above. The Plant Nursery MAS is composed by the

agents that take care of all the processes related to

plant producing, such as catalog processing or pro-

duction following up. Both of these MAS are tightly

related to the legacy systems existing in both types

of organizations. The Communication MAS holds the

agents that undertake the actions related to mutual ex-

clusion preservation and information integration.

357

Otero Cerdeira L., J. Rodríguez Martínez F., Valencia Requejo T. and Lueiro Astray L..

Multi-agent Systems and Ontologies Applied to New Industrial Domains - Case Study: Ornamental Plants.

DOI: 10.5220/0004136303570364

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2012), pages 357-364

ISBN: 978-989-8565-30-3

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

The rest of this paper is structured as follows, in

Section 2 a detailed view of the system’s goals will

be provided. In Section 3 the builded platform will be

fully analyzed. In Section 4 we will describe some is-

sues that were discovered during the development of

the platform and ﬁnally in Section 5 we will summar-

ize the main conclusions of the development as well

as further improvements that will be undertaken.

2 SYSTEM’S GOALS

The project’s main goal was to obtain a platform that

could improve the production management, cost con-

trol in the producers and marketing support in foreign

countries by using MAS and Business Intelligence al-

gorithms.

Nowadays there are some commercial applica-

tions that centralize all the information related to plant

nursing and plant selling processes allowing queries

over this information. These applications are simply

aggregative or storage systems and do not provide

support to the enterprise’s manager in relevant tasks

such as production management, real-time order man-

agement, stock following up and production planning.

Some of these solutions do have modules that take

care of tasks such as ﬁeld control, commercial in-

voices management and, customers and providers tra-

cing.

From plant producers point of view, such applic-

ations do not achieve the requirements established,

since they don’t take into account the special char-

acteristics of plants as products. Plants are a per-

ishable item and so its production planning is not as

straight and pre-established as other product’s plan-

ning would be. Plants grow, die, bloom, etc., these

kind of changes and others plants experience turn

products into new ones. The same plant in the same

ﬂowerpot can be different products as it changes. Be-

sides in these applications costs are stablished follow-

ing ﬁnancial and accounting parameters, which works

ﬁne in other types of productive areas but not on this

one.

From the plant wholesaler point of view, there

are several systems that do marketing management

but that are simply Enterprise Resource Planning sys-

tems (ERP) which have been adapted from other pro-

ductive areas and that do not consider the special cir-

cumstances of such an heterogeneous market as plant

selling is.

The following needs have been identiﬁed as pro-

ject goals as they don’t coexist in a single software ap-

plication in this environment: Dynamic catalog pro-

cessing, Dynamic price estimation, Dynamic route

planning, Order processing, Production cost man-

agement, Production planning, Reservation of future

products, Sales forecasting and Stock following up.

Besides the functional requirements identiﬁed

above, there were also other non-functional require-

ments that were tagged as mandatory for the ﬁnal plat-

form. Among them, the most remarkable are: Ef-

ﬁciency, Expandability, Flexibility, Maintainability,

Performance, Reliability, Robustness, Scalability, Se-

curity and Usability.

After analyzing all the requirements listed above,

both functional and non-functional, the most suitable

approach seemed to develop a multi-agent system.

At this point it is important to state that each plant

nursery has its own information structure and con-

fers different semantic meaning to the concepts of

the plant nursing ﬁeld. Information integration was

one of the major problems that the legacy system had.

People that ﬁrst developed the legacy system were not

experts in the ﬁeld and so they used concepts without

fully understanding its meaning. Without a common

ontology over the concepts, the use and maintenance

of the legacy system has turned incomprehensible. It

is important to remark that the MAS itself will not

cover for the information integration and so, means to

overcome this information heterogeneity needed to be

studied.

In order to achieve all the requirements listed

above a MAS by itself could not replace the previ-

ously existing legacy system, it was also necessary a

deep reengineering over the information system.

After considering several alternatives to rebuild

the information system, the chosen one involves re-

building the database structure for both types of en-

tities. To perform the communication and data ex-

change between them the solution was found in onto-

logies.

By using this approach, agents deﬁned for the

MAS would use an ontology to describe the meaning

of concepts in the communication processes (Wies-

man et al., 2002), avoiding this way the heterogeneity

that exists between the involved knowledge sources.

3 SOLUTION’S DESCRIPTION

The platform that we propose consists in the deﬁni-

tion of a layered MAS whose agents use ontologies to

provide interoperability between them. As mentioned

before, the global MAS is divided into three MAS that

will now be described.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

358

Figure 1: MAS and their relations.

3.1 MAS Description

The plant wholesaler MAS consists of a group of

agents performing tasks for which a traditional solu-

tion would not provide the appropriate functional-

ity. Accordingly, intelligent behavior will accomplish

tasks such as order processing, catalog processing or

route planning.

In ﬁgure 1 the whole structure of the Plant Whole-

saler MAS, the Plant Nurseries MAS and the interac-

tions among all these MAS are depicted.

3.1.1 Order Processing

Order processing involves several groups of agents.

Those from the plant wholesaler compose its Order

Processing MAS. The same happens in the plant nurs-

eries, every agent involved in order processing integ-

rates the Order Processing MAS. The agents deﬁned

for the plant wholesaler are of two different types, Or-

der Agent and Negotiator.

Order Agent. An order agent is responsible for split-

ting an order between its negotiators and compos-

ing the shipping with the results of each negotiator.

When a new order arrives to the plant wholesaler, a

new order agent is created to manage it. This agent

reports its existence to a mediator agent, which acts as

an intermediary between the plant wholesaler’s agents

and the plant nurseries’s agents. For every item in the

order list, the order agent activates a negotiator agent

to manage it. Once every negotiator is done, the order

agent composes the ﬁnal order with the information

that each one of its negotiators provides.

Negotiator. A negotiator agent is responsible for

dealing with a provider to get a product. A negoti-

ator agent is responsible for getting the best offer for

a product. The best offer might not only be determ-

ined by the price of the product. Indeed, it is neces-

sary allowing that other constraints could be applied,

such as provider’s proximity or even constraints in the

product.

For order processing, in the plant nurseries side

two types of agents were deﬁned, Provider and

Broker.

Provider. A provider agent supplies products from the

plant nursery. When a new provider is created it com-

municates its existence to a mediator agent. Since all

mediator agents are federated, the knowledge about

providers and negotiators is shared among all of them.

A provider agent has a group of broker agents that ne-

gotiate the different products. A provider has at most

a broker agent for each product that is sold in the plant

nursery. The provider regularly reports its state to a

mediator, so the mediators can optimize the distribu-

tion of requests from the negotiators among the dif-

ferent brokers. The provider reports on the products

that their brokers are currently selling and on those

that they have in stock for selling.

Broker or Delegate. A broker agent is responsible

for dealing with plant wholesaler’s negotiators. A

broker agent sells to negotiator agents its product, re-

specting the boundaries that the provider agent may

have established. If the broker belongs to an internal

provider, it deals the conditions of the sale with the

negotiator agent speciﬁed by a mediator. In this case

we refer to it as delegate. If the broker agent belongs

to an external provider, the selling process works as

an auction between the negotiator agent and all the

broker agents that sell the product.

The last type of agent involved in order processing

is the Mediator. This type of agent does not belong to

the plant wholesaler’s MAS or to the plant nurseries’s

MAS, it belongs to the Communication MAS.

Mediator. A mediator agent is responsible for linking

negotiators to brokers. In the communication MAS

there are several federated mediators, which means

that the knowledge that one of them has is immedi-

ately shared with the other mediators. A mediator

agent is responsible for discovering the providers and

for linking the negotiators to the brokers.

In ﬁgure 2 the whole structure of the order pro-

cessing procedure is reﬂected. In ﬁgure 3 a detailed

view of all the communications established among the

agents is speciﬁed. In this ﬁgure the communication

between Negotiator - K and Delegate - Q is performed

inside the Communication MAS, as reﬂects the line

labeled as 4. The ontologies necessary for commu-

nication to be successful will be described in section

3.2.

Multi-agentSystemsandOntologiesAppliedtoNewIndustrialDomains-CaseStudy:OrnamentalPlants

359

Figure 2: Order Processing.

Figure 3: Communication during Order Processing.

3.1.2 Catalog Processing

Catalog processing involves agents in the plant

wholesaler and in the plant nurseries, additionally

agents in the Communication MAS are necessary for

mediation purposes.

The types of agents involved in catalog processing

are Catalog Combiner and Seeker for the plant whole-

saler, and Catalog Manager for each plant nursery.

Catalog Combiner. A catalog combiner is respons-

ible for obtaining an updated catalog for the plant

wholesaler. The product catalog of the plant whole-

saler must be daily revised and modiﬁed to provide

the customers with an updated version. The catalog

combiner agent builds the catalog from the existing

one, taking into account both internal and external in-

formation. As internal information, the sales of the

day and the reservations must be considered, and as

external information, the updates on every plant nurs-

eries’s catalog that is a provider for the plant whole-

saler must be introduced.

Seeker. A seeker agent is responsible for gather-

ing the updated information about a provider’s cata-

log. A seeker agent is responsible for getting the up-

dated version of a provider’s catalog. Once a seeker

is created it already knows the provider that it must

deal with. By default the seeker gets the plant nurs-

ery’s whole catalog, but it can also be conﬁgured to

query the provider’s Catalog Manager for an speciﬁc

product o for a product that fulﬁlls a set of require-

ments.

Catalog Manager. A catalog manager agent is re-

sponsible for updating a plant nursery’s catalog. The

catalog manager is responsible for updating a plant

nursery’s catalog by taking the existing one and modi-

fying it with the sales of the day and the modiﬁcations

that the products may have experienced. The catalog

manager reports its existence to a mediator agent, so

when a catalog combiner agent queries the mediator

for the list of plant nurseries, it is included. From that

moment on, the catalog manager is available for tak-

ing the requests from the seekers.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

360

Figure 4: Catalog Processing. Communication and Structure.

As line labeled as 3 in ﬁgure 4 reﬂects, the

communication between a seeker and a catalog

manager takes place inside the communication MAS.

This communication is be successful since both

agents involved are using same ontology. This will

be detailed in section 3.2.

Mediator. A mediator agent is responsible for provid-

ing indexing information. A mediator agent holds the

information over which catalog managers are avail-

able to be queried from the seekers. Since in the com-

munication MAS several mediators exist, they must

be federated so there is no difference in the informa-

tion that they hold.

3.1.3 Route Planning

Another important task of the system is the dynamic

route planning. It was also addressed as a Multi-Agent

Resource Allocation (MARA) problem. In this type

of problems a group of agents share a common re-

source which requires a coordination mechanism that

will manage its usage (Cicortas and Iordan, 2011).

Dynamic route planning can be addressed as an inde-

pendent task that lies outside the scope of this paper,

and that will be developed in further papers.

3.2 Ontologies Deﬁned

The existing information structure that the different

entities had, was very different, causing the interac-

tion between them to be very hard. In addition we

have identiﬁed several cases where the same term

was used with different meanings, not only between

the different entities but also within the same en-

tity. The MAS deﬁned in this project would not suc-

ceed if the problem of the semantic gap were not ad-

dressed. To accomplish this task, two different on-

tologies were deﬁned, one for the plant wholesaler

and the other one for the plant nurseries, this two

ontologies share the semantic meaning over the con-

cepts and so, agents are able to interact. In multi-

agent environments ontologies are expected to com-

plement mutual understanding and interactive beha-

vior between such agents (Laera et al., 2007).

The ontologies used in the MAS were deﬁned with

the Ontology Web Language (OWL2) (Grau et al.,

2008) which is the standard recommended by the

W3C. This language is used to formalize the domain,

deﬁning the concepts as classes, and the properties

that these classes have (Roussey et al., 2011). Onto-

logies were identiﬁed as the most accurate technique

to share common knowledge among a group of soft-

ware agents.

3.2.1 Plant Wholesaler Ontology

In ﬁgure 5 the global structure of the Plant Wholesaler

Ontology is depicted.This ontology reﬂects the spe-

ciﬁc situation of the enterprise that hosted the project

and it may not be applied to every plant wholesaler

due to the peculiarities of the enterprise.

Figure 5: Plant Wholesaler Ontology.

Multi-agentSystemsandOntologiesAppliedtoNewIndustrialDomains-CaseStudy:OrnamentalPlants

361

The plant wholesaler ontology is composed by the

following classes and properties.

Class: wholesaler:Catalog. It represents a collection

of products that are available. The catalog is com-

posed by integrating the list of available products that

each provider of the plant wholesaler has. The prop-

erty wholesaler:generatedIn links a catalog to its cre-

ation date.

Class: wholesaler:Product. A product is the smal-

ler entity that composes a catalog. Products are usu-

ally plants, but they could also be goods related to the

business, such as fertilizers or ﬂowerpots. Products

are identiﬁed by a plant variety and the deﬁning fea-

tures of that plant. The property wholesaler: integ-

ratedBy deﬁnes the products that are part of a catalog.

Class: wholesaler:Item. It identiﬁes the actual plant

variety of a Product. In the botanical hierarchy the

item should be equivalent to the plant species (Class:

wholesaler:Species), however in most cases this does

not happen, and the equivalence is set at the genus

level (Class: wholesaler:Genus) or even at the fam-

ily level (Class: wholesaler:Family). Fixed attributes

would be: Flowerpot, Size and Price, which means

that for every product the values of this attributes must

always be set. The property wholesaler: basedIn in-

dicates that the item is the base of a product.

Class: wholesaler:Features. It represents a col-

lection of attributes that distinguish the different

products. There are some attributes that every product

must enclose which are included in the Class: whole-

saler:Fixed, and there are other features that depend

on the family, genus and species of the product, which

are encompassed in Class: wholesaler:Variable. This

subclasses are deﬁned as disjoint classes. The prop-

erty wholesaler:deﬁnedBy indicates that a product is

modiﬁed by a collection of features.

Class: wholesaler:Order. It represents the collection

of elements that a customer requests from the plant

wholesaler. The property wholesaler:orderedIn links

an order to its creation date.

Class: wholesaler:Customer. It encloses the know-

ledge about the plant wholesaler’s clients. The prop-

erty wholesaler:orderedBy identiﬁes the customer has

requested an order.

Class: wholesaler:OrderElement. It represents each

one of the elements in an order. Each OrderElement

is only linked to a product in the catalog and to a

provider, that would be the nursery that provides the

product. The property wholesaler:composedOf de-

notes that an order is integrated by a collection of or-

der elements.

Class: wholesaler:Provider. It identiﬁes the

nurseries that interact with the plant wholesaler.

The providers must be split into two differ-

ent and disjoint subclasses, internal providers

(Class: wholesaler:Internal) and external providers

(Class: wholesaler:External). The property whole-

saler:providedBy indicates the provider that supplies

a certain product.

3.2.2 Plant Nursery Ontology

In ﬁgure 6 the global structure of a nursery’s ontology

is depicted. This is the ontology that every agent in a

nursery’s MAS will use. This ontology is composed

by the following classes and properties.

Figure 6: Plant Nursery Ontology.

Class: nursery:Customer. It represents the clients

that the plant nursery may have. There are some

customers that are considered as internal, and there-

fore deﬁned in (Class: nursery:Internal). This class

is disjoint to the subclass (Class: nursery:External)

that represents the external clients of the nursery. To

relate a customer with the orders, there is the property

nursery:makes, deﬁned to be asymmetric and irreﬂex-

ive.

Class: nursery:Order. It represents the collection of

elements that a customer requests from the plant nurs-

ery. The property nursery:involves indicates that an

order groups a collection of plants.The property nurs-

ery:orderedIn relates an order to the date it was made.

Class: nursery:Plant. It represents the smaller en-

tity that the plant nursery works with. The property

nursery:locateIn indicates the location of a plant by

relating them to one of the different available ﬁelds.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

362

Class: nursery:State. It denotes the different situ-

ations a plant may have. The plant nurseries do not

have a catalog itself but a group of plants that satisfy

some quality requirements to be sold. This state has

a high frequency change. The property nursery:hasA

identiﬁes that a plant has a certain state among all the

possible ones.

Class: nursery:Field. It represents a piece of land

where different types of plants may be cultivated.

Each ﬁeld is identiﬁed by a group of properties and

it has a cost assigned. The features of the ﬁeld have

an impact over its cost, which therefore affects plant’s

cost.

Class: nursery:Properties. It represents a group of

attributes that distinguish the different ﬁelds. The

property that relates a ﬁeld to its feature is nurs-

ery:identiﬁedBy.

Class: nursery:Features. It represents a collec-

tion of attributes that distinguish the different plants.

There are some attributes that every plant must en-

close which are included in the Class: nursery:Fixed,

disjoint to the features deﬁned in Class: nurs-

ery:Variable. The property nursery:deﬁnedBy indic-

ates that a plant is modiﬁed by a collection of features,

it is an irreﬂexive and asymmetric property.

Class: nursery:Procedures. It denotes a collection of

actions or operations that can be made over the plants.

The procedures are composed by atomic instructions

deﬁned in Class: nursery:Action. The property nurs-

ery:composedBy represents the link between a pro-

cedure and its composing actions. The property nurs-

ery:suffers indicates that procedure is applied to a

plant or group of plants. Every procedure consumes

resources such as, fertilizers, water, sawdust, that

are enclosed in the class Class: nursery:Material.

These concepts are related by the property nurs-

ery:consumes.

Class: nursery:Action. It deﬁnes the atomic actions

that could be made on a plant or set of plants.

Class: nursery:Cost. It denotes the value that plants,

procedures or ﬁelds have. Its domain is deﬁned as

the intersection of these classes. The property nurs-

ery:worthA relates a class in the domain to its cost.

The cost of a plant is determined taking into account

the cost of the procedures that it may have suffered

and the cost of the ﬁeld where the plant is being

grown. It is important to remark that the cost of a

plant is not its price. The price is determined apply-

ing a percentage of proﬁt over the cost of production.

This margin of beneﬁt is crucial to the agents nego-

tiation processes that take place to accomplish order

processing.

4 IDENTIFIED ISSUES AND

FUTURE WORK

The platform deﬁned to ensure proper operation and

networking between plant nurseries and plant whole-

saler fully meets expectations as far as regards inner

functioning.

Plant nurseries that adopted the platform de-

scribed have experienced an improvement in response

times, cost estimation and production following up.

In turn in the plant wholesaler the effort to keep the

catalog updated and to process the costumer’s orders

has been reduced. Problems showed up when some

plant nurseries chose to use the platform but keeping

their own information representation. Agents were no

longer able to interact correctly since their ontologies

where not compatible.

To overcome the situation the most suitable solu-

tion was to force an ontology matching process be-

fore agents started their communication. Imposing a

single shared ontology would be, not only impractical

because it would force the parties to use a standard

communication vocabulary; but also limiting since it

would not consider the requirements of agents that

could be developed in the future. Ontology match-

ing is a way of guaranteeing the interoperability of

the parties involved in a communication process, i.e,

to ascribe to each important piece of knowledge the

correct interpretation (Euzenat, 2001). In our case,

the purpose is to ﬁnd semantic mappings between the

concepts of the different ontologies that the agents

use.

The problem of ontology matching has been ex-

tensively studied in recent years as stated in the works

of (Noy, 2004), (Euzenat and Shvaiko, 2007), (Tro-

jahn et al., 2008) or (Shvaiko and Euzenat, 2012).

And also its applications in Multi-Agent Systems

(Laera et al., 2007) (Wiesman et al., 2002).For gener-

ating the matches between ontologies several frame-

works and techniques have been proposed, as those

in the works of (Falconer and Noy, 2009) (Klein,

2001) (Shamsfard and Barforoush, 2003) (K

opcke

and Rahm, 2009) (David et al., 2010).

The next step of the development of this platform

will be the integration of ontology matching tech-

niques, to ensure the communication with agents that

use an ontology different to the proposed one. The

process that we will use to accomplish this challenge

is as follows. First, the state of the art of ontology

matching (Kalfoglou and Schorlemmer, 2003) will be

deeply studied to identify the new trends and research

ﬁelds. Then a framework will be developed to test

different ontology matching algorithms, the purpose

is to determine which set or combination works the

Multi-agentSystemsandOntologiesAppliedtoNewIndustrialDomains-CaseStudy:OrnamentalPlants

363

best for this problem. Finally, once the algorithms are

determined, they will be integrated in the system.

5 CONCLUSIONS

In this paper we presented a Multi-Agent System ap-

plied to a new ﬁeld, ornamental plant selling and dis-

tribution, where this technologies have never been

used before. The MAS deﬁned takes into account

the particularities of this environment and it obtains

the most proﬁt of MAS features to deal with it. The

system allows the different parties to operate not only

coordinated but also independently. To allow this co-

ordination two ontologies were deﬁned in order to

transmit the knowledge of this domain to the agents

involved in the communication processes. This onto-

logies by themselves were not enough to cover every

possible communicative scenario, since other sys-

tems, or agents could use ontologies different from

the ones deﬁned. A further step of including ontology

matching has been identiﬁed as necessary and it is the

core of the ongoing research.

ACKNOWLEDGEMENTS

This work has been supported by the project

10MRU007E supported by Xunta de Galicia.

REFERENCES

Cicortas, A. and Iordan, V. (2011). Multiagent systems for

resource allocation.

David, J., Euzenat, J., Scharffe, F., and dos Santos, C. T.

(2010). The Alignment API 4.0. IOS Press, 1:1 – 8.

Euzenat, J. (2001). Towards a principled approach to se-

mantic interoperability.

Euzenat, J. and Shvaiko, P. (2007). Ontology Matching.

Springer.

Falconer, S. M. and Noy, N. F. (2009). Interactive Tech-

niques to Support Ontology Matching. Data-Centric

Systemsand Applications - Springer.

Gowri, R., Kailas, A., Jeyaprakash, R., and Anirudh, C.

(2010). Development of multi-agent system for ﬁre

accident detection using gaia methodology. Interna-

tional Journal of Computer Science and Information

Security (IJCSIS).

Grau, B. C., Horrocks, I., Motik, B., Parsia, B., Patel-

Schneider, P., and Sattler, U. (2008). Owl 2: The next

step for owl.

Kalfoglou, Y. and Schorlemmer, M. (2003). Ontology map-

ping: The state of the art. Science, 2:3.

Klein, M. (2001). Combining and relating ontologies: An

analysis of problems and solutions.

opcke, H. and Rahm, E. (2009). Frameworks for entity

matching: A comparison. Data and Knowledge En-

gineering - Elsevier Science.

Laera, L., Blacoe, I., and Tamma, V. (2007). Argumentation

over ontology correspondences in mas.

Macho, S., Torrens, M., and Faltings, B. (2000). A multi-

agent recommender system for planning meetings. In

In Proc. of the Agents’2000 workshop on Agent-based

recommender systems (WARS’2000).

Murakami, Y., Minami, K., Kawasoe, T., and Ishida, T.

(2002). Multi-agent simulation for crisis management.

Noy, N. F. (2004). Semantic Integration: A Survey Of

Ontology-Based Approaches. SIGMOD, 33(4).

Roussey, C., Pinet, F., Kang, M. A., and Corcho, O. (2011).

An Introduction to Ontologies and Ontology Engin-

eering. Advanced Information9 and Knowledge Pro-

cessing - Springer.

Shamsfard, M. and Barforoush, A. A. (2003). The State

of the Art in Ontology Learning: A Framework for

Comparison.

Sharma, S. (2009). Avatarsim: A multi-agent system for

emergency evacuation simulation. Journal of Compu-

tational Methods in Science and Engineering.

Shvaiko, P. and Euzenat, J. (2012). Ontology matching:

state of the art and future challenges. IEEE Transac-

tions on Knowledge and Software Engineering.

Trojahn, C., Moraes, M., Quaresma, P., and Vieira, R.

(2008). A cooperative approach for composite onto-

logy mapping.

Valckenaers, P., Sauter, J., Sierra, C., and Rodriguez-

Aguilar, J. A. (2006). Applications and environments

for multi-agent systems. Springer Science Business

Media.

Wiesman, F., Roos, N., and Vogt, P. (2002). Automatic

ontology mapping for agent communication. ACM.

KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

364