ENGINEERING MOBILE AGENTS

Arnon Sturm

Department of Information Systems Engineering, Ben Gurion University of the Negev, Beer Sheva, 84105, Israel

Dov Dori

Faculty of Industrial Engineering and Management, Technion – Israel Institute of Technology, Haifa, 32000, Israel

Onn Shehory

IBM Haifa Research Lab, Haifa University Campus, Haifa, 31905, Israel

Keywords: Agent-based system, Agent-Oriented Software Engineering, Mobility.

Abstract: As the mobile agent paradigm becomes of interest to many researchers and industries, it is essential to

introduce an engineering approach for designing such systems. Recent studies on agent-oriented modeling

languages have recognized the need for modeling mobility aspects such as why a mobile agent moves,

where the agent moves to, when it moves and how it reaches its target. These studies extend existing

languages to support the modeling of agent mobility. However, these fall short in addressing some modeling

needs. They lack in their expressiveness: some of them ignore the notion of location (i.e., the "where")

while others do not handle all types of mobility (the "how"). Additionally, they lack in their accessibility, as

the handling of the mobility aspects is separated into multiple views and occasionally the mobility aspect is

tightly coupled with the functional behavior specification. View multiplicity reduces the comprehensibility

and the ease of specification, whereas the coupling with the functional behavior specification reduces the

flexibility of deploying a multi-agent systems in different configurations (i.e., without mobility). In this

paper, we address these problems by enhancing an expressive and accessible modeling language with

capabilities for specifying mobile agents. We provide the details of the extension, then illustrate the use of

the extended modeling language, and demonstrate the way in which it overcomes existing problems.

1 INTRODUCTION

Agent mobility has been intensively studied in

recent years. Although mobility is not a primary

property of an agent, it may enhance agent

autonomy (which is definitely a primary attribute of

agents (Wooldridge et al., 2000)). Gray et al. (2001)

identified several advantages of using mobile agents:

conservation of bandwidth, reduction in total

processing time, reduced latency, connected

operation, mobile computing, load balancing, and

dynamic deployment of software. Due to these

advantages, mobile agents are sometimes the most

suitable paradigm for agent-based systems

engineering (Cabri et al., 2001). This paradigm has

been demonstrated beneficial in several application

areas, e.g., distributed information retrieval,

workflow management, and network management

(Bellavista et al., 1999; Brewington et al., 1999;

Gray et al., 2001).

Because of the increased industrial and research

interest in mobile agents, agent-oriented

methodologies should support the mobility aspect. In

particular, it is necessary to integrate agent mobility

into the modeling languages of these methodologies,

to provide a comprehensive engineering approach

for building agent-based systems. This will help

designers of agent-based systems to engineer mobile

agents.

OMG (2000) and FIPA (2001; 2003) carried out

many activities in order to standardize agent

mobility aspects. Those activities led to an

agreement on the required infrastructure for agent

mobility in terms of general mobility concepts (such

as places and regions) within a multi-agent system

Sturm A., Dori D. and Shehory O. (2008).

ENGINEERING MOBILE AGENTS.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - SAIC, pages 79-84

DOI: 10.5220/0001680300790084

 SciTePress

(MAS) and the required functionality of mobile

agents (such as agent migration, agent cloning, and

agent invocation). To support the standards (to be),

it is necessary that agent-oriented modeling

languages facilitate the specification of these agent

mobility concepts. In this paper, we enhance an

existing agent-oriented modeling language with the

capabilities of specifying the mobility aspects within

an agent-based system. Note that our study is not

first in addressing this need. Some of the agent-

oriented modeling languages recognized the need for

modeling mobility aspects such as why a mobile

agent moves, where the agent moves to, when it

moves and how it reaches its target (Mouratidis et

al., 2002).

In this paper, we address the problem of

integrating the mobility aspects of agent-based

systems into a modeling language by introducing

two key elements that should be supported by a

modeling language with respect to mobility: the

definition of places and environments and the

definition of agent mobility (i.e., migration, cloning

and invocation) (FIPA, 2003). We our approach

leverage on an exiting method – Object-Process

Methodology (OPM) and its MAS extension and

suggest a modeling language for specifying mobility

aspects of agent-based system. The choice of OPM

for modeling agent-based system is discussed in

Sturm et al. (2003), in which the authors present the

motivation for adopting OPM. They mainly discuss

the advantages of OPM with respect to accessibility

and expressiveness. We found it useful as well since

it provides a single unified model for capturing the

various system aspects and enables to view the

system as a whole, as we believe, required for

modeling mobile agents.

The contribution of this paper is the introduction

of a comprehensive modeling language for

specifying multi-agent systems including the

mobility.

The rest of this paper is organized as follows.

Section 2 discusses related work. The Object-

Process Methodology for Multi Agent Systems

(OPM/MAS) is shortly described in Section 3.

Section 4 introduces the enhancements we propose

to support the modeling of the mobility aspects of

MAS, and Section 5 concludes with a discussion on

the advantages of the OPM/MAS approach for

specifying mobile agents and with future research

directions.

2 RELATED WORK

In the last decade many studies address the notion of

modeling mobile MAS. The Multi-agent Software

Engineering (MaSE) which is a general-purpose

methodology for developing heterogeneous MASs

(DeLoach et al., 2001) was extended to enable the

specification of mobile agents (Self and DeLoach,

2003). GAIA which is a methodology for agent-

oriented analysis and design (Wooldridge et al.,

2000) has been enhanced by m-GAIA to model the

mobility aspect (Sutandiyo et al., 2003). Several

extensions for UML (OMG, 2007), which is the

standard de-facto for modeling systems were

suggested including Park et al. (2000), Klein et al.

(2001), Mouratidis et al. (2002), Poggi et al. (2003),

and Kang and Taguchi (2004). The various

extensions mainly lacks in their expressiveness and

their accessibility. With respect of expressiveness

this are lacking in specify the locations related to the

mobility operation and the integration of the

mobility notion into the system functionality. The

accessibility limitation is mainly stems from the

model multiplicity problem discussed in Kabeli and

Shoval (2001) and Peleg and Dori (2000). The

problem is characterized by a lack of integration

between the models and the need to maintain

consistency across them and the need to gather

information from various models for understanding

the system specifications.

3 OPM/MAS IN A NUTSHELL

OPM/MAS is based on the Object-Process

Methodology (Dori, 2002; OPM, 2007), which is an

integrated approach to the study and development of

systems in general and information systems in

particular.

OPM is a general-purpose methodology, thus

using its core symbol set may be too low-level for

modeling a domain-specific application. This is so

because the pertinent domain uses specialized

concepts and building blocks that are at a higher

level of abstraction than the basic OPM entities

(objects, processes, and states).

OPM extension for MAS, as discussed by Sturm

et al. (2003), follows principles of the Meta Object

Facility (MOF) concept (OMG, 2002), which

enables its flexibility. In that work the authors

suggest a three layers architecture as follows: (1) a

meta-model, which is the OPM itself; (2) an

intermediate meta-model, which describes the

specific building blocks within the MAS domain

ICEIS 2008 - International Conference on Enterprise Information Systems

using OPM; and (3) a model, which is based on both

the meta-model and the intermediate meta-model.

That intermediate meta-model, which is the core

element of OPM/MAS, was designed based on

previous research. The designers of OPM/MAS

divide the set of MAS building blocks into two

groups. The first group consists of static, declarative

building blocks, while the second group consists of

building blocks with behavioral, dynamic nature.

In figures 1 and 2 the intermediate meta-model

for multi-agent systems using OPM is presented. It

also includes mobility-based concepts which were

not discussed in Sturm et al. (2003). This

intermediate meta-model provides guidelines and

constraints for modeling multi-agent systems,

including mobile agent systems.

OPM extension for MAS, as discussed by Sturm

et al. (2003), follows principles of the Meta Object

Facility (MOF) concept (OMG, 2002), which

enables its flexibility. In that work the authors

suggest a three layers architecture as follows: (1) a

meta-model, which is the OPM itself; (2) an

intermediate meta-model, which describes the

specific building blocks within the MAS domain

using OPM; and (3) a model, which is based on both

the meta-model and the intermediate meta-model.

That intermediate meta-model, which is the core

element of OPM/MAS, was designed based on

previous research. The designers of OPM/MAS

divide the set of MAS building blocks into two

groups. The first group consists of static, declarative

building blocks, while the second group consists of

building blocks with behavioral, dynamic nature.

In figures 1 and 2 the intermediate meta-model

for multi-agent systems using OPM is presented. It

also includes mobility-based concepts which were

not discussed in Sturm et al. (2003). This

intermediate meta-model provides guidelines and

constraints for modeling multi-agent systems,

including mobile agent systems.

Figure 1: System Diagram of the OPM/MAS meta-model.

4 DESIGNING MOBILE AGENTS

USING OPM/MAS

In this section we present the OPM/MAS

enhancements to support agent mobility

specification. The first sub-section introduces the

new building blocks, whereas the second sub-section

demonstrates the use of the intermediate metamodel

for specifying agent mobility within the context of

MAS.

4.1 Mobility in OPM/MAS

In this section we introduce the proposed changes to

OPM/MAS in order to make it suitable for

specifying agent mobility. As stated in Sturm et al.

(2003), the OPM/MAS meta-model can be changed

according to application needs. Following that

principle, we propose an enhancement to the

suggested set of building blocks.

The new the building blocks are the following:

Figure 2: OPM/MAS meta-model - Agent in-zoomed.

ENGINEERING MOBILE AGENTS

• Mobilizing: A set of methods by which an

agent performs a mobility activity such as

migration, cloning, and agent invocation.

• Type: The mobility type - migration,

cloning, or invocation.

• Location: The location to which the agent

should migrate.

• Priority: The importance level of the

mobility. This parameter is defined in order

to define priorities in case of several

possibilities for mobility within an agent.

• Result: The indication of the mobilizing

process result.

The enhancement of the set of building blocks

proposed in this paper refers to the system structure

and to the system flow. The system structure is

separated into the logical architecture, which is

represented by the environment building block and

the physical architecture, which is represented by the

platform building block. Further, the relationships

between the architectures are also specified. Since

we are dealing with structural aspects of the system,

the above building blocks were chosen to be OPM

objects. The system flow enhancement consists of

adding the mobilizing process and its parameters.

Since we view mobility as part of the agent

functionality (as in other studies, e.g., Self and

DeLoach (2003) and Baumeister et al. (2003)), we

define it as an OPM process in order to integrate the

agent mobility into the regular flow of the agent

process and yet differentiate it from regular tasks by

its role (i.e., mobilizing). However, in contrast to

existing languages the mobility specification can be

easily removed by filtering out all mobility building

blocks mentioned above, allowing the specification

of a system without introducing mobility constraints.

4.2 The Technical Report Searcher

Case Study

In the following, we present an example of using

OPM/MAS for modeling mobile agents related to

distributed information retrieval via the technical

report (TR) searcher case study (Gray et al., 2001).

In this case study technical reports are distributed

across several machines, each executing a stationary

information retrieval agent. When these agents are

executed, they register with a virtual yellow pages

agent. In search for technical reports, a client agent

queries the yellow pages agent for the location of the

stationary information retrieval agents. It then sends

its "child" agents to these locations. These child

agents interact with the stationary agents, which in

turn reply to the child agents with the search results.

In addition, the client agent checks for the network

quality and determines whether it requires migrating

to a proxy site in order to communicate with the

machines hosting the stationary agents.

Figure 3: Technical Report Searching System – System

Diagram.

Figure 4: Technical Report Searching System – TR

Searching Agent in-zoomed.

Figure 3 presents the system diagram of the

Technical Report Searcher system. It consists of four

platform types: Client Platform, which hosts the TR

Searching Agent; Administrator Platform, which

hosts the Yellow Pages Agent; the Information

Resource Platform, which hosts the Stationary IR

(Information Retrieval) Agents; and the Proxy

Platform, which is capable of hosting the TR

Searching Agent in case of faulty communication

between the Client Platform and the Information

Resource Platform. The OPD in Figure 3 also

describes the communication paths (which are the

logical routes among the agents) and messages.

ICEIS 2008 - International Conference on Enterprise Information Systems

Figure 5: Technical Report Searching System – Searching

Task in-zoomed.

In Figure 4, the TR Searching Agent is in-

zoomed. The agent is activated by the Client User,

as shown by the agent link from the object Client

User (a human) to the process TR Searching Agent.

The TR Searching Agent performs its tasks

sequentially, as determined by the vertical order

within the OPD. The Querying Task accepts the

user's input and yields a Query object. The

Administrating Task follows the Querying Task

and yields an object indicating whether mobility is

required and an object representing the required

Agent Location. If mobility is required, the Proxy

Mobilizing process occurs and causes the TR

Searching Agent to migrate to the Proxy Platform.

If the Proxy Mobilizing process fails, then it is

triggered again. This is indicated by the “e”, for

"event", attached to the arrowhead of the link

connecting the failure state within the Mobility

Result object and the Proxy Mobilizing process.

When the mobility result is a success (or if it was not

required in the first place), the Searching Task is

performed, followed by the Results Receiving Task.

In Figure 5, the Searching Task from the OPD

of Figure 4 is specified by zoom into its

specification. It begins with the Agent Location

Mapping Task, which determines Mobility

Location, in which the Child Agent has to be

invoked. This task is followed by launching the

Child Agent in the appropriate platform. In case the

invocation fails, the agent tries to re-launch the

Child Agent until it succeeds. When the Child

Agent is running, the TR Searching Agent sends a

Request Query Message and waits for a reply.

Upon receiving the query results, Reply Query

Message, the results received from its child agents

are merged to obtain the Results Set, which the

Client User ultimately receives, as Figure 4 shows.

5 CONCLUSIONS

In this paper we leverage on the object-process

methodology to facilitate the modeling of agent

mobility. We show how the OPM/MAS intermediate

meta-model can be enhanced in order to support

agent mobility. Following that enhancement, we

demonstrate the use of that intermediate meta-model

to specify a MAS application. In particular, we

exemplify the way according to which the

OPM/MAS addresses the four questions of mobility:

1. Why a mobile agent performs a mobility action?

In OPM/MAS the reason of the agent mobility is

encapsulated within the task flow. This means

that the agent mobility is determined according

to the task flow and decisions that are made

during its process. The mobility is represented as

an OPM process (Mobilizing) thus easily

integrated within the task flow. In the TR case

study, the Client agent moves in order to

improve its communication, where as a Child

agent is invoked in order to search information in

another location.

2. When the agent performs a mobility action?

The timing in which the agent moves is specified

in OPM/MAS via the process sequence. It may

move due to other task termination, it may move

due to new information, or it may move due to a

user request. In the TR case study, the Client

agent moves upon determining Administrating

problems.

3. Where the agent moves to?

The destination of the agent in OPM/MAS is

determined by the mobilizing process parameter

– Location. The locations within a system

according to OPM/MAS could be platforms

(could be referred to as places) or environments

(could be referred to as regions). These are

usually specified within the top level OPDs.

4. How the agent reaches its target?

The path according to which the agent reaches its

target is specified by the order of the mobilizing

processes. In TR case study, the path is a straight

forward way, from the Client platform to the

Proxy platform.

The weaknesses of the existing agent-oriented

methods with regards to mobility are addressed

within the proposed solution. We refer to the

location notion with its various level of abstraction

ENGINEERING MOBILE AGENTS

by providing the environment building block and its

relationships with the platform building block. The

proposed solution refers to all mobility types by

defining the mobility type object. We integrate all of

the mobility aspects within a single-unified

framework, whereas in the other methods the

integration of some of the mobility aspects is not

clear, difficult to understand, or non-exist.

Further research is required to examine the

accessibility and adherence of the OPM/MAS

approach to build agent-based systems and in

particular, mobile agents.

ACKNOWLEDGEMENTS

This work was partially funded by Deutsche

Telekom Laboratories at Ben-Gurion University.

REFERENCES

Bellavista, P., Corradi, A., and Stefanelli, C., 1999. An

Open Secure Mobile Agent Framework for Systems

Management. Journal of Network and Systems

Management (JNSM), Special Issue on Mobile Agent-

based Network and Service Management 7(3), 323-

339.

Brewington, B., Gray, R., Moizumi, K., Kotz, D.,

Cybenko, G., and Rus, D., 1999. Mobile Agents for

Distributed Information Retrieval. In Intelligent

Information Agents, chapter 15.

Cabri, G., Leonardi, L., Mamei, M., and Zambonelli, F.,

2001. Mobile Agent Organizations. In WOA 2001.

DeLoach, S. A., Wood, M. F., and Sparkman, C. H., 2001.

Multiagent Systems Engineering.The International

Journal of Software Engineering and Knowledge

Engineering, 11 (3), 231-258.

DeLoach, S. A. and Wood, M., 2001. Developing

Multiagent Systems with agentTool. In Proceedings of

ATAL 2000, LNCS 1986.

Dori, D., 2002. Object-Process Methodology - A Holistic

Systems Paradigm, Springer, Heidelberg.

FIPA, 2001. FIPA Agent Management Support for

Mobility Specification.

http://www.fipa.org/specs/fipa00087/PC00087B.pdf.

FIPA, 2003. FIPA Modeling Area: Deployment and

Mobility.

http://www.auml.org/auml/documents/DeploymentMo

bility.zip.

Gray, R.S., Cybenko, G., Kotz, D., and Rus, D., 2001.

Mobile Agents: Motivations and State of the Art.

Handbook of Agent Technology.

Kabeli, J., and Shoval, P., 2001. FOOM: functional- and

object-oriented analysis & design of information

systems - an integrated methodology. Journal of

Database Management, 12 (1), 15-25.

Kang, M. and Taguchi, K., (2004) Modelling Mobile

Agent Applications by Extended UML Activity

Diagram. ICEIS 2004, 519-522.

Klein, C., Rausch, A., Shiling, M., and Wen, Z., 2001.

Extension of the Unified Modeling Language for

Mobile Agents. In The Unified Modeling Language:

Systems Analysis, Design and Development Issues, 1-

25, Idea Group Publishing.

Mouratidis, H., Odell, J., and Manson, G., 2002.

Extending the Unified Modeling Language to Model

Mobile Agents. In Proceedings of the Agent Oriented

Methodologies Workshop (at the OOPSLA 2002).

OMG, 2000. Mobile Agent Facility Specification.

http://www.omg.org/docs/formal/00-01-02.pdf.

OMG, 2002. Meta Object Facility (MOF) v1.4.

http://www.omg.org/technology/documents/formal/mo

f.htm.

OMG, 2007. Unified Modeling Language (UML) 2.0,

http://www.omg.org/technology/documents/formal/um

l.htm,.

OPM, 2007. http://www.objectprocess.org.

Park, S., Kim, J., and Lee, S., 2000. Agent-Oriented

Software Modeling with UML Approach. IEICE

Transaction on Information & Systems, E83-D (8),

1631-1641.

Peleg, M. and Dori, D., 2000. The Model Multiplicity

Problem: Experimenting with Real-Time Specification

Methods. IEEE Transaction on Software Engineering,

26(8), 742-759.

Poggi, A., Rimassa, G. , Turci, P., Odell, J., Mouratidis,

H., and Manson, G., 2003. Modeling Deployment and

Mobility Issues in Multiagent Systems Using AUML

In Proceeding of Agent-Oriented Software

Engineering (AOSE), 69-84.

Self, A. and DeLoach, S. A., 2003. Designing and

Specifying Mobility within the Multiagent Systems

Engineering Methodology. Special Track on Agents,

Interactions, In Mobility, and Systems (AIMS) at The

18th ACM Symposium on Applied Computing.

Sturm, A., Dori, D., and Shehory, O., 2003. Single-Model

Method for Specifying Multi-Agent Systems. In

Proceedings of the second international joint

conference on Autonomous agents and multiagent

systems, 121-128.

Sutandiyo, W., Chhetri, M. B., Krishnaswamy, S., and

Loke, S. W., 2003. From m-GAIA to Grasshopper:

Engineering Mobile Agent Applications. In

Proceedings of iiWAS2003.

Wooldridge, M., Jennings, N. R., and Kinny, D., 2000.

The Gaia Methodology for Agent-Oriented Analysis

and Design. Journal of Autonomous Agents and Multi-

Agent Systems 3 (3), 285-312.

ICEIS 2008 - International Conference on Enterprise Information Systems