doi:10.1016/j.eswa.2007.02.018


Available online at www.sciencedirect.com
www.elsevier.com/locate/eswa

Expert Systems with Applications 34 (2008) 2019–2028

Expert Systems
with Applications
Apply ontology and agent technology to construct virtual observatory

Ruey-Shun Chen a, Duen-Kai Chen b,*

a
Department of Information Management, China University of Technology, Taipei 116, Taiwan, ROC

b
Institute of Information Management, National Chiao Tung University Hsinchu 300, Taiwan, ROC
Abstract

The need to deal with abundance and heterogeneous information is apparent in the astronomy community. The virtual observatory
(VO) concept is the astronomical community’s response to alleviate this problem and Web services serve as one of the most important
VO enabling technologies. However, one of the limitations of Web services is the lack of semantic description of its content, thus pro-
hibits its ability to understand the queries and its inference capabilities. This study proposes to develop a conceptual framework based on
multi-agent systems and ontology technology, in order to create a VO with semantically enriched Web services. Intelligent agents rep-
resent: (1) users to submit requests (2) perform semantic matching in between users’ requests and Web services registered within agent
platform, and (3) activate a serial of Web services. The capabilities offered by multi-agent systems to query and invoke semantically
enriched Web Services is also exploited in this study. To validate the proposed framework, an illustration example is implemented in
JADE agent platform to demonstrate how the proposed framework operates and how it benefits the research regarding to auroral
images.
� 2007 Elsevier Ltd. All rights reserved.

Keywords: Multi-agent systems; Ontology; Web services; Virtual observatory
1. Introduction

The need to retrieve and process exponentially increas-
ing of information is becoming apparent and is having a
huge impact on professions that rely on distributed
archived information. This is true especially for the auroral
physicists who depend largely on the shared image infor-
mation content, for example, images from different loca-
tions, with different sizes, intensity, and morphology of
the auroral oval. Therefore, how to retrieve needed infor-
mation, access to language independent software for image
processing and furthermore, share the research results
becomes an essential topic. The virtual observatory (VO)
concept is the astronomical community’s response to the
information abundance. VO is an emerging, open, web-
0957-4174/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2007.02.018

* Corresponding author. Tel.: +886 3 5712121x57427; fax: +886 3
5723792.

E-mail addresses: rschen@iim.nctu.edu.tw (R.-S. Chen), cdk@iim.
nctu.edu.tw (D.-K. Chen).
based, distributed research environment for astronomy
with massive and complex data sets.

Web services serve as one of VO enabling technologies;
however, there are some limitations of the traditional Web
services. One of the limitations is that Universal Descrip-
tion, Discovery and Integration (UDDI) and Web Services
Description Language (WSDL) do not provide semantic
description of its content. UDDI is characterized for its
lack of semantic description mechanisms, such as semantic
interoperability, explicit semantic models to understand the
queries and inference capabilities (Baousis et al., 2006). The
same limitation also applies to WSDL, for WSDL does not
contain information about the capabilities of the described
Web services.

This paper proposes to develop a conceptual framework
based on multi-agent systems and ontology technology to
assist auroral physicists work with semantically enriched
web services and realize the concept of VO. The above
mentioned framework can be used for retrieving necessary
information, dynamically allow users to define descriptors
and operations most appropriate for their purposes, for

mailto:rschen@iim.nctu.edu.tw
mailto:cdk@iim.nctu.edu.tw
mailto:cdk@iim.nctu.edu.tw


Fig. 1. Conceptual architecture of virtual observatory (Schlenoff et al.,
2005).

2020 R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028
example, comparison of physical features between auroral
images, regardless of their source or data representation.
In other words, the proposed framework complies with
the VO concept. The goal of the proposed framework is
to assemble data archives and services, as well as data
exploration and analysis tools.

Recently, ontology and agent technology have been
widely applied to various domains. For example, Jing, Bi,
and Wu (2005) pointed out that Geospatial information
services may provide the data and function services to han-
dle with basic geo-tasks, in the meanwhile, user and soft-
ware agent should be able to discover, invoke, compose
and monitor geospatial resources offering particular
services and having particular properties. Web services
are turning into the dominant means for connecting remo-
tely executing programs via Internet and commonly used
machine readable representations such as XML. The
Web Ontology Language for Services (OWL-S) is imported
to express the process ontology and process control ontol-
ogy. In virtue of the Web services semantic markup,
develop the agents to support the automatic reasoning
and fulfill the composition of Web Services and inter-oper-
ation is possible.

The contributions of this paper are twofolds. First is to
provide an environment which is easy to obtain and share
information with the assistance of ontology and multi-
agent systems via dynamically discover and invoke seman-
tically enriched Web services. Furthermore, it is believed
that the proposed framework could easily be extended for
similar domains. Second, the proposed framework utilizes
multi-agent systems and ontology for share, retrieve and
process information. Adopting multi-agent systems to
assist information processing can achieve advantages such
as robustness and scalability. The proposed framework
enhances flexibility and extensibility owing to its modular
design. Modular design means that the module within the
architecture can be substituted by better and newer mod-
ules when necessary, thus ensuring the flexibility and exten-
sibility. To facilitate information sharing among different
research projects, the proposed framework develop ontol-
ogy which includes standardized, community-accepted
descriptions of the information. Such ontology served as
the basis for information content definitions that can be
used repeatedly.

The rest of the paper is organized as follows. Section 2
introduces the background on Web services, VO, and
space-based auroral image research; ontology technology;
multi-agent systems-especially focuses on the agent com-
munication and coordination. Section 3 presents how the
proposed framework assists auroral physicists in retrieving
and processing images from heterogeneous data sources by
incorporated ontology and multi-agent systems technol-
ogy. Section 3 also gives the overview of the proposed
framework; the types of agents consisted in the framework,
the functionality of each agent. Section 4 highlights the
software solution and implementation to the proposed
framework and provides practical examples to illustrate
how images from different sources (e.g. ultraviolet imager
(UVI), far-ultraviolet imager (FUV)) can be retrieved and
processed. Section 5 gives the conclusion and identifies
future works.

2. Literature review

2.1. Space-based auroral image, virtual observatory and Web

services

A conceptual architecture of a VO is shown in Fig. 1.
From a space physicists’ viewpoint, one should be able to
discover the available data for their study, which generally
reside in distributed archives, federate them, and pipe the
output into a set of data mining/knowledge discovery in
databases analysis and discovery tools. These services
may be implemented as Web services, and may involve
use of AI and machine learning tools, coupled with sophis-
ticated visualization environments (Djorgovski, 2005).

W3C (2004) defines web service as ‘‘. . . a software sys-
tem designed to support interoperable machine-to-machine
interaction over a network. It has an interface described in
a machine-processable format (specifically WSDL). Other
systems interact with the Web service in a manner pre-
scribed by its description using simple object access proto-
col (SOAP) messages, typically conveyed using HTTP with
an XML serialization in conjunction with other Web-
related standards’’.

A Web service is an accessible application that other
applications and humans can discover and invoke. Web
services are nowadays emerging as a major technology
for deploying automated interactions between distributed
and heterogeneous applications. Various standards back
this deployment, including

• WSDL – support the definition of Web services.
• UDDI – support advertisement to the community of

potential users.
• SOAP – binding for invocation purposes.



R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028 2021
One of the examples for facilitating Web services as an
enabling technology for VO described in Ohishi et al.
(2004). The authors in Ohishi et al. (2004) depicted how
UDDI, which serves as yellow page services, is adopted
to establish Japanese virtual observatory (JVO).

2.2. Ontology

Ontology is an agreement about a shared conceptualiza-
tion, which includes frameworks for modeling domain
knowledge and agreements about the representation of
particular domain theories (Huhns & Singh, 1997). Ontol-
ogy is an esoteric concept borrowed from philosophy by
the Artificial Intelligence/Information Technology commu-
nities. In this study, we adopt the definition of ontology as
‘‘a way of describing all things in common across a collec-
tion of similar objects’’. Ontology served as standardized,
community-accepted descriptions of the information con-
tent. There existed previous attempts to apply ontology
in the field of image processing, in Soo, Lee, Yeh, and Chen
(2002), a framework of utilizing sharable domain ontology
and thesaurus to help the retrieval of historical images of
the First Emperor of China’s terracotta warriors and
horses was presented. Ontology has also been used in Earth
Sciences and Geographic Information Systems (GIS) appli-
cations to organize and allow inter-comparison of remote
sensing data sets from disparate detectors (Fonseca, Egen-
hofer, Agouris, & Camara, 2002).

The reason why ontology is essential to the intelligent
agents is that it can provide a shared virtual world in which
each agent can ground its beliefs and actions. That is why
ontology is becoming increasingly recognized as a crucial
element of scalable multi-agent system technology.
Recently, there are plenty of literatures (Bailin & Trusz-
kowski, 2001; Bravo, Perez, Sosa, Montes, & Reyes,
2005; Obitko & Marik, 2002) which highlight the crucial
role ontology played in the development of multi-agent
systems.

2.3. Agent technology

According to Jennings and Wooldridge’s (1998) defini-
tion, an agent is a computer system situated in some envi-
ronment, and that is capable of autonomous action in this
environment in order to meet its design objectives. Agent
systems can be further classified either single-agent or
multi-agent systems. Agent systems, particularly multi-
agent systems, are subfields of distributed artificial intelli-
gence (DAI) research, and have existed under AI for two
decades (Chen, Chen, & Lin, 2005). Generally, DAI is bro-
ken into distributed problem solving (DPS) and multi-
agent systems. Research on designing and developing
multi-agent systems focuses on the interaction between
agents. Topics frequently discussed in this area of research
include agent action, the relationship between agents,
multi-agent system architecture and the environment in
which the multi-agent systems exist; interactions among
agents within the multi-agent system, and agent adapta-
tion ability. Coordination, negotiation, cooperation are
three common interactions among agents in multi-agent
systems.

Regarding the application of software agents, Jennings
and Wooldridge (1998) noted that ‘‘agent are being used
in an increasingly wide variety of applications, ranging
from comparatively small systems such as email filters to
large, open, complex, mission critical systems such as air
traffic control’’. For example, Lee, Yun, and Jo (2003) pro-
posed an auction agent system using a collaborative mobile
agent and a brokering mechanism called Mobile collabora-
tive auction agent system (MoCAAS), which mediates
between the buyer and the seller and executes bidding asyn-
chronously and autonomously.

In the past decade, academic communities have wit-
nessed a proliferation of software agents with widely vary-
ing specialties. Agent technology is a key area in artificial
intelligence research. Today, a software agent generally
means a software program that accomplishes a task on
behalf of its user. Multi-agent systems along with ontology
have been applied for supporting distributed decision mak-
ing in several fields, such as manufacturing, business and
engineering. For example, a cooperative multi-agent plat-
form was introduced in Soo, Lin, Lin, and Cheng (2005)
to support the invention process based on the patent doc-
ument analysis. There are also studies abundant in inte-
grate ontology with agent technology. Wu, He, and Jin
(2005) proposed to build an open, large-scale and interop-
erable distributed intelligent medical diagnosis and therapy
system in the context of a wide-area network such as the
Internet, the paper studies integrating mobile agent and
ontology with Web Services, using ontology as a standard
web service, which avoids misunderstandings in agent com-
munication. Furthermore, it assures the correctness of
matching agent services according to semantics accurately
in service register server like the UDDI registry. Schlenoff,
Washington, and Barbera (2005) developed intelligent
ground vehicle (IGV) ontology. The goal of their effort
was to develop a common, implementation-independent,
extendable knowledge source for researchers and develop-
ers in the intelligent vehicle community. Malucelli, Palzer,
and Oliveira (2005) in their paper, combines the use of
ontology and agent technologies to help in solving the
semantic heterogeneity problem in e-commerce negotia-
tions. The proposed approach aims at creating a methodol-
ogy that assesses lexical and semantic similarity among
concepts represented in different ontologies without the
need to build an a priori shared ontology. Day et al.
(2005) described an intelligent tutoring agent (ITA) that
uses the ontology, INFOMAP, and question answering
techniques through the Instant Messaging platform for
the ‘‘operating system’’ course.

In virtue of the semantically enriched Web services,
develop the intelligent agents to support the automatic rea-
soning and decision making, as well as fulfill the composi-
tion of Web services and inter-operation is possible.



Fig. 2. Proposed system architecture.

2022 R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028
3. Apply ontology and agent-based architecture to build

virtual observatory

In this section, we first describe how the proposed
framework operates. Afterwards, the functionality and
internal structure of the agents and components are
depicted.

As illustrated in Fig. 2, the proposed framework mainly
consists of the web server and the agent platform for all the
agents to reside in. In a typical operational scenario, the
user may send the request specifying some user defined cri-
teria through the browser, as same as using any of cur-
rently existing web application. The advantages of using
browser to connect to the proposed framework are easy
to use and to ensure that users can be accommodated with-
out having agent platform or any additional programs (e.g.
Java Runtime Environment) installed on their client side.
The web application also provides services such as user
account creation, user login, logout and creation of user
profile.

Web server receives the request and then sends it to the
Gateway Agent. The role of the Gateway Agent here is to
connect web application and agent platform in a transpar-
ent manner; it serves as an intermediary service. Gateway
Agent allows the web application to invoke agent; translat-
ing the HTTP compliant messages to the Foundation for
Intelligent Physical Agents (FIPA) compliant Agent Com-
munication Language (ACL) – and vice versa – in order to
let Web services and agent technology work together. Per-
sonal Agent, as describe in Fig. 3, should be capable of
interpreting user’s intention. Once the Personal Agent
decomposes the user’s request into subtasks, it then send
the request to the Broker Agent and searching for Service
Provider Agents capable of accomplish the aforementioned
subtasks. Broker Agent allows Service Provider Agents to
post its services. After finding appropriate agents to per-
form the subtask, Personal Agent then sends request to
the target agent and starts to negotiate with the target
agent directly. When all of the subtasks had been fulfilled,
then Personal Agent will send result to Gateway Agent and
Gateway Agent then return the result to user via web page.

After introducing the overview of the proposed frame-
work; the types of agents consist in the framework and
its functionalities are describes as follows.

3.1. Personal agent (PA)

The PA serves as a representative for users in agent plat-
form. The main purpose of a PA is finding and executing
services and delivering results to the user. A PA is com-
posed of the following components:

• Agent communication module, agents should be able to
communicate with each other via ACL. Agent commu-
nication module facilitates interaction with Gateway
Agent, Broker Agent, and all potential Service Provider
Agents.

• Domain knowledge base specifies the autonomous
behavior of the PA. It contains the information to guide
the PA to accomplish the task delegated by the user.

• Ontology database is used to store ontology for PA to
facilitate query and invoke semantically enriched Web
services.

• User intention module can be further broken down into
User log database, Learning module and User prefer-
ences database. The aim of this module is to provide



Fig. 3. Personal Agent internal architecture.

R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028 2023
customize services for each user and attempt to be able
to provide the user with more satisfactory services in the
future. To be able to assist end user to complete its goal.
This is done by ontology and image processing knowl-
edge base, as well as the learning capability build within
the PA. Being intelligent agent, the PA should also able
to improve itself through self-learning. The focus of the
learning module is to learn user’s preferences and pro-
files in order to provide more accurate assistance. The
learning source is provided by the user’s log, which is
the history of the user’s actions on the agent platform.
Ontology let allows the agent share the same concept
of as the domain expert. The learning mechanism will
assist agent in identifying the user’s preferences.
3.2. Service provider agent (SPA)

The service provider agent serves as a representative of
an existing web service in the agent platform. It provides
the web service to interested users and maintains a descrip-
tion of the web service expressed in WSDL and ontology,
such as Web Ontology Language (OWL). The WSDL
description is facilitated to find the necessary definitions
for Web services’ successful invocation, while OWL is used
to enhance the expressiveness of WSDL in terms of seman-
tic information.
3.3. Gateway agent

This framework presents a Gateway agent for connect-
ing intelligent agents and Web services in an automatic,
transparent manner. This Gateway agent allows Web ser-
vices to invoke agents by translating Web services’ requests
to agent communication language encodings, and enable
automatic, transparent connection between these two tech-
nologies. In other words, Gateway agent serves as Web ser-
vices and agent platform’s bridge. Integrating Web services
and intelligent agents generates the foreseeable benefits of
connecting these two application domains. Once the inter-
connection is established, intelligent agent concepts and
technologies will help enable new, advanced operational
and usage modes of Web services. Service invocation by
the Gateway agent depends on the OWL description of
the service requests. OWL is used to enhance the expres-
siveness of WSDL in terms of semantic information.
OWL is used to specify the input and output ontologies,
enabling an advanced service capability search, and find
the necessary definitions for successful Web services
invocation.
3.4. Broker agent

Service matching along with ontology and ACL assure
the correctness of matching agents and Web services
according to semantics in service register server like the
UDDI registry. Brokering is one of the most significant
discovering mechanisms among autonomous, intelligent
agents. The main task of Broker agent is to interpret que-
ries from requester and matches capabilities provided by
service provider. The brokering process can be categorized
into two steps. First, broker agent extracts required capa-
bilities from the query send by PA. Then it compares and
matches these required capabilities with what service pro-
viders can accomplish. Broker agent extendeds the tradi-
tional matching mechanism by finding service providers
by capability or functionality rather than simply by name.
Service providers register their capabilities within Broker
agent, which stores all the registries in a local knowledge
base. If capabilities change or the service providers no
longer provide the service, then broker agent maintains



Fig. 4. Service request and service provider matching.

Table 2
Comparison of imaging characteristics for contemporary auroral imagers.

Imager Frame
rate (s)

Image size
(pixels)

Imaging
method

Spatial
resolution
(Apogee)

Polar/UVI 37 200 · 228 Snapshot 40 km
Image/FUV-WIC 120 256 · 256 Time Delayed

Imaging
100 km

2024 R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028
or deletes the service registries. Fig. 4 illustrates the broker-
ing process.
4. Illustration example and discussion

4.1. Illustration example

In this section, an illustration example is provided to
demonstrate how the proposed framework operates. In a
typical scenario, an auroral physicist compares two images
from two different imagers, UVI and FUV. The detector-
specific structural details are sufficiently different between
imagers that comparison of the information content such
as the size of the oval can be difficult. The steps to accom-
plish the comparison are listed in Tables 1 and 2 summa-
rizes imaging characteristics of these two imagers.

The typical workflow can be described by Fig. 5. During
the past, physicists must be able to master all the image
databases management systems, related image processing
algorithm and be able to manipulate the image according
to the specification requirement. Image processing algo-
rithms could be implemented in different programming lan-
guages, for example, backgrounds remove and flat field
correction mentioned in the workflow were implemented
Table 1
Image comparison processes

User selects time range and spatial location

Access/find data files for Imager 1 (I1) Same for Imager 2
Select records within time range for I1 Same for Imager 2
Process I1 records to convert from stored format

to science format. Includes background
removal, calibration, flat-field correction, and
any other instrument-specific corrections

Same for Imager 2

Interpolate image to requested time Same for Imager 2
Select region of interest in interpolated image that

matches the requested spatial location
Same for Imager 2

Summarize ROI, e.g. mean and standard deviation
of all pixels within ROI

Same for Imager 2

Compare values from both cameras
in IDL, and calibration was implemented in C++. Thus,
physicists sometimes need to install or even recompile the
program and make it work in a local environment.

In our proposed framework, first of all, the workflow is
according to the following steps.

Step 1, physicist gets to select time range and spatial
location of the images to be compared through a browser.
The client system is implemented in JSP and Servlet
technology.

Step 2, if the user is the first time user, Gateway Agent
will create a new account and profile, as well as a new
PA for this specified user. Then the auroral physicist’s
PA decomposes the task ‘‘compare two images’’ into the
process mentioned above through the help of Auroral
Image Knowledge Base.

Step 3, PA sends the request to the Broker Agent for
matching process, to see if there exists any needed Data
and Functional Agents (In this illustration example, SPA
can be further categorized as data agent and functional
agent. The former provides data retrieval and the latter
provides variety of other functionalities.).

Step 4, after the matching services provided by Broker
Agent, the PA first sends a Call-For-Proposal to all the
potential Data Agent (representing which imager) to deter-
mine which Data Agents have the image that satisfies the
user’s request. Some of the Data Agents respond with Pro-
pose-essentially, detail description of the imager and its
capability to fulfill the user’s request. The PA then chooses
those imagers match user’s request.

Step 5, PA then sends the user’s query to Data Agents as
a service request. Now it is the Data Agent’s turn to inter-
pret and try to fulfill the request. After receiving the result
from Data Agents, the PA can apply image processing
algorithms, such as image background removal, calibra-
tion, flat field correction, to the images.

Step 6, in order to do so, again, the PA sends a Call-For-
Proposal to potential Functional Agent (representing
image processing algorithm) which can be helpful in satis-
fying the user’s request. Some of the Functional Agents
respond with Propose, and PA decide which Functional
Agent is most appropriate to perform the image processing
according to the user’s preferences (such as processing time
or accuracy). After receiving the response from the Func-
tional Agent, the PA then represents the result to Gateway
Agent, and Gateway Agent then sends the result to user via
JSP page. Such process can be described by the flowchart
listed in Fig. 6.



Fig. 5. Image comparison workflow.

Send request to
PA

First time
user

Create new PA
through
gateway

PA interprets 
user request

PA find
appropriate data

agent in DF

PA send fail 
message to user 

and End

PA request DA 
to perform data

retrieval

DA accepts
Pa s request or

not

DA performs 
data retrieval 

and send result 
back to PA

Yes

No

Yes

No

PA find
appropriate

functional agent
in DF 

No

FA performs
requested function

and send result back
to PA

Yes

Fig. 6. Workflow of how illustrated example is accomplished.

R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028 2025
We implement the prototype with JADE and Tomcat
web server to demonstrate the efficacy of this framework.
Currently, JADE is perhaps the most pervasive agent sys-
tem in use, especially in the context of research driven
applications. JADE is an open source, FlPA compliant
agent platform designed to be a middleware solution for
developing agent-based software. Fig. 7 illustrates how
agents in the prototype communicate with each other via
FIPA compliant ACL. We use Protégé as our ontology edi-
tor. Protégé, developed by Stanford University, is an ontol-
ogy editor, a knowledge-base editor, an open-source, Java
tool. One of the advantages of Protégé is that it provides an
extensible architecture for the creation of customized
knowledge-based applications. The reason for chosen Pro-
tégé was due to its strong user community and its ability to
support the OWL as ontology representation. The Sniffer
agent, a JADE tool which is basically a FIPA-compliant
Agent with sniffing features in order to track messages
exchanged between agents in a JADE based environment,
was employed to verify the validity of the model proposed
in this paper. Every message exchanged between agents
implies the execution of the model, can be tracked to view
in runtime. Fig. 8 drawn by Sniffer agent tool from JADE
depicts the message exchange between agents.

4.2. Discussion

Comparing the user behaviors-in both traditional means
to handle the image retrieval, image processing and via
proposed framework – in the scenario described above, it
is believed that the workload of the users can be alleviated.
Fig. 9 describes the steps to be accomplished in order to
perform analysis for two images from two imagers.

Note that in the proposed framework, the tasks to be
accomplished within the dotted rectangle are handled by
the multi-agent systems. This implies the following advan-
tages for the users.

(1) Web services solve the problem of obtaining (includ-
ing find, install or even recompile) the appropriate
software to retrieve and process the images.

(2) The auroral image process knowledge base can be
facilitated for the PA to choose appropriate Web ser-
vices to perform image retrieval and image process
for different images from different imagers and with
different image specifications. In other words, users
may invoke a set of services. By this means, users
do not have to be familiar with all the imagers’ spec-
ification to choose the appropriate tools to accom-
plish the pre-processes for the analysis. This is an
explicit advantage, since collaboration between vari-
ous image teams is common and users might not be
familiar with other imager’s specification and related
image processing algorithm. The proposed frame-
work allowed users to focus on their research and



Fig. 7. Agent communication sequence.

2026 R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028
analysis instead of mastering image processing algo-
rithm, software, and tools.

(3) Apply ontology to facilitate semantic enriched Web
services can provide better request-service matching.
Service matching performed by Broker Agent along
with ontology can assure the correctness of matching
service requester and service provider. In other words,
the capabilities offered by agents to query and invoke
semantically enriched Web services is based on
enhanced registries enriched with semantic informa-
tion that provide semantic matching to service queries
and published service descriptions.

(4) New services, agents, and users can be easily added to
the framework, thus providing an extendable and
flexible open system.
5. Conclusion

In this study, we present a framework that provides
easy access to heterogeneous data sources and image pro-
cessing algorithms through ontology and multi-agent Sys-
tems; meanwhile, we also exploit the capabilities offered
by multi-agent systems to query and invoke semantically
enriched Web services. All the users, data sources and algo-
rithms are represented by different types of agents. Service
providers register themselves to the yellow page service
provided by broker agents with semantic information. PA
represent users and send user’s requests to broker agents.
Broker agent then performs semantic matching in between
services registered within yellow page services and service
request submitted by Personal Agent. The advantages of



Fig. 8. Snapshot of agent communication sequence in JADE.

Fig. 9. Illustration example workflow.

R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028 2027
the proposed framework can be summarized as follows: (1)
develop ontology which includes standardized, commu-
nity-accepted descriptions for the auroral images. (2) Auro-
ral scientists can focus on their research instead of
mastering image processing algorithm, software, tools. (3)
Users may invoke a set of Web services with the integration
of Web services and intelligent agents. This generates the
foreseeable benefits of connecting these two application
domains, once the interconnection is established, intelligent
agent concepts and technologies will help enable new,
advanced operational and usage modes of Web services.
(4) Provide an environment which is easy to obtain and
share research data and image processing algorithms with
the assistance of ontology and multi-agent systems. (5)
New services, agents, users and service registries can be eas-
ily added to the framework, thus providing an expandable
and flexible open system. The framework could easily be
extended for similar domains, such as other satellite images
and Earth Sciences. The prototype implementation indi-
cates that the proposed framework is flexible, robust and
extendable.
References

Bailin, S. C., & Truszkowski, W. (2001) Ontology negotiation between
scientific archives. In 2001 Proceedings of thirteenth international
conference on scientific and statistical database management (pp. 245–
250). Virginia.



2028 R.-S. Chen, D.-K. Chen / Expert Systems with Applications 34 (2008) 2019–2028
Baousis, V., Zavitsanos, E., Spiliopoulos, V., Hadjiefthymiades, S.,
Merakos, L., & Veronis, G. (2006) Wireless web services using mobile
agents and ontologies. In 2006 ACS/IEEE international conference on
pervasive services (pp. 69–77). Beijing.

Bravo, M. C., Perez, J., Sosa, V. J., Montes, A., & Reyes, G. (2005)
Ontology support for communicating agents in negotiation processes.
In 2005 Fifth international conference on hybrid intelligent systems (pp.
482–487). Rio de Janeiro.

Chen, R., Chen, D., & Lin, S. (2005). ACTAM: Cooperative multi-agent
system architecture for urban traffic signal control. IEICE Transac-
tions on Information and Systems, E88-D(1), 119–126.

Day, M., Lu, C., Yang, J., Chiou, G., Ong, C., & Hsu, W. (2005)
Designing an ontology-based intelligent tutoring agent with instant
messaging. In ICALT 2005 fifth IEEE international conference on
advanced learning technologies (pp. 318–320). Kaohsiung.

Djorgovski, S. G. (2005) Virtual astronomy, information technology, and
the new scientific methodology. In Proceedings of seventh international
workshop on computer architecture for machine perception, CAMP (pp.
125–132). Terrasini-Palermo.

Fonseca, F., Egenhofer, M., Agouris, P., & Camara, C. (2002). Using
ontologies for integrated geographic information systems. Transactions
in GIS, 6(3), 231–258.

Huhns, M. N., & Singh, M. P. (1997). Ontologies for agents. Internet
Computing, IEEE(6), 81–83.

JADE, http://jade.tilab.com/.
Jennings, N. R., & Wooldridge, M. J. (1998). Applications of intelligent

agents. In Agent technologies: Foundations, applications, and markets
(pp. 3–28). Springer.

Jing, D., Bi, S., & Wu, F. (2005) Geospatial information services on the
basis of agent and OWL-S. In Proceedings of 2005 IEEE international
geoscience and remote sensing symposium, IGARSS ’05 (pp. 885–888).
Seoul.
Lee, K., Yun, J., & Jo, G. (2003). MoCAAS: Auction agent system using a
collaborative mobile agent in electronic commerce. Expert Systems
with Applications, 24, 183–187.

Malucelli, A., Palzer, D., & Oliveira, E. (2005) Combining ontologies and
agents to help in solving the heterogeneity problem in e-commerce
negotiations. In Proceedings of international workshop on data
engineering issues in e-commerce (pp. 26–35). Tokyo.

Obitko, M., & Marik, V. (2002) Ontologies for multi-agent systems in
manufacturing domain. In Proceedings of 13th international workshop
on database and expert systems applications (pp. 597–602). Aix-en-
Provence.

Ohishi, M., Mizumoto, Y., Yasuda, N., Shirasaki, Y., Tanaka, M.,
Honda, S., & Masunaga, Y. (2004) A prototype toward Japanese
virtual observatory (JVO). In 2004 International symposium on
applications and the internet workshops (pp. 591–595). Tokyo.

Schlenoff, C., Washington, R., & Barbera, T. (2005) An intelligent ground
vehicle ontology to enable multi-agent system integration. In Interna-
tional conference on integration of knowledge intensive multi-agent

systems (pp. 169–174) Waltham.
Soo, V., Lee, C., Yeh, J., & Chen, C. (2002) Using sharable ontology to

retrieve historical images. In Proceedings of the 2nd ACM/IEEE-CS
joint conference on digital libraries (pp. 197–198). Portland.

Soo, V., Lin, S. Yang, S., Lin, S., & Cheng, S. (2005) A cooperative multi-
agent platform for invention based on ontology and patent document
analysis. In Proceedings of the ninth international conference on
computer supported cooperative work in design (pp. 411–416).

W3C, Web Services Architecture, http://www.w3.org/TR/2004/NOTE-
ws-arch-20040211/, 2004.

Wu, Z., He, Y., & Jin, H. (2005) A model of intelligent distributed medical
diagnosis and therapy system based on mobile agent and ontology. In
Proceedings of eighth international conference on high-performance

computing in Asia-Pacific region (pp. 582–587). Beijing.

http://jade.tilab.com/
http://www.w3.org/TR/2004/NOTE-ws-arch-20040211/
http://www.w3.org/TR/2004/NOTE-ws-arch-20040211/

	Apply ontology and agent technology to construct virtual observatory
	Introduction
	Literature review
	Space-based auroral image, virtual observatory and Web services
	Ontology
	Agent technology

	Apply ontology and agent-based architecture to build virtual observatory
	Personal agent (PA)
	Service provider agent (SPA)
	Gateway agent
	Broker agent

	Illustration example and discussion
	Illustration example
	Discussion

	Conclusion
	References