AGENCY SERVICES

An Agent-based and Services-oriented Model for Building Large Virtual

Communities

I. Lopez-Rodriguez and M. Hernandez-Tejera

Institute for Intelligent Systems (SIANI), University of Las Palmas de Gran Canaria, Las Palmas, Spain

Keywords:

Intelligent agents, Middle agents, Cloud computing, Utility computing, Matchmaking & Brokering.

Abstract:

Despite the current importance of internet and the increasing appearance of virtual societies, there is still no

widespread adoption of intelligent agents. Building solutions based on intelligent agents is not a simple task

because of the complex nature of the problems that they face, which are mainly related to planning, cooperation

and negotiation tasks. In order to overcome those difﬁculties, and in line with the recent success of the Cloud

Computing model, this paper proposes a new model in which brokers able to represent users in virtual societies

are offered as one more service of the cloud. The paper also details the technologies necessary to construct a

realistic solution which, apart from abstracting the user from all the details of the implementation, loses none

of the characteristic advantages of solutions based on agents and Cloud Computing.

1 INTRODUCTION

Although the presence of software agents has signif-

icantly increased in various ﬁelds, they have not yet

been widely adopted as a strategy. However, that is in

contrast to the success of electronic commerce and the

emergence of virtual societies, to which they can pro-

vide enormous advantages. To a great extent, the rea-

son for that absence lies in the complexity surround-

ing solutions that are based on intelligent agents and

which still represent an insurmountable technological

barrier to many users.

Moreover, the new means of access and the ubiq-

uitous character of communications are transforming

the traditional business models. The classic concep-

tion of software consumed as a product is being sub-

stituted by one of a service housed in third party in-

stallations. This is corroborated by the latest strategic

movements of companies as representative as IBM,

Google, Yahoo!, Microsoft or Amazon, all of which

are undertaking projects in which the user no longer

ﬁgures as the owner of a software license but as a con-

sumer of their services. This new way of conceiving

the software market is known as Cloud Computing.

This paper proposes the creation of Agency Ser-

vices in the cloud in order to facilitate the implemen-

tation of virtual societies based on software agents.

They are proposed as intermediary agents that are in-

stantiated in the cloud by third parties and, after being

contracted as services, represent the users in virtual

societies by conducting brokering tasks. The paper

presents a realistic solution that manages to abstract

the user from all the details of the technology without

losing any of the advantages of the agents.

The remainder of this article is structured as fol-

lows. This section is completed with both a presenta-

tion of some important challenges faced by software

agents, and a brief description of the cloud computing

model. Section 2 gives an overview of middle agents.

Section 3 describes the Agency Services model, the

technologies needed for building a realistic solution

based on it, and ﬁnally its contribution to the current

models. Section 4 comments the implementation of

an auction web site based on the new model, and Sec-

tion 5 is devoted to the conclusions.

1.1 Challenges of Software Agents

A review of the works related to the study of the re-

lationship between intelligent agents and electronic

commerce reveals that, in the quality of software

modules, agents are rarely instantiated on the client

side (He et al., 2003). On the same line, the commu-

nity devoted to the agency theory was recently invited

309

Lopez-Rodriguez I. and Hernandez-Tejera M. (2010).

AGENCY SERVICES - An Agent-based and Services-oriented Model for Building Large Virtual Communities.

In Proceedings of the 2nd International Conference on Agents and Artiﬁcial Intelligence - Agents, pages 309-314

DOI: 10.5220/0002722903090314

 SciTePress

to discuss the question “Where are all the Intelligent

Agents?” (Hendler, 2007). That author believes that

there are currently no really important systems based

on agents and noted that, despite the apparent activ-

ity in the research area, intelligent agents are far from

being the widely adopted technology predicted more

than ten years ago.

The origin of the problem lies in the fact that it

is a difﬁcult technological challenge for the users to

incorporate their own agent-based solutions to prob-

lems that would normally be related to planning, co-

operation and negotiation in distributed, shared and

complex environments.

With the aim of stimulating the creation of agent-

based solutions, some works have opted to release de-

velopment frameworks . They follow a strategy that

transfers to the users the responsibility of program-

ming their solutions and, therefore, are not aligned

with the dominant class of software that is currently

demanded by the market; precisely that which re-

quires no effort from the users. Other solutions have

also opted to offer users the opportunity to instantiate

agents on the server (Wurman et al., 1998; Sandholm,

2002). However, those agents always belong to the

virtual site, are only accessible through the web, sub-

tract autonomy from the users, do not allow to track

local resources, and do not hide the details of the tech-

nology.

For intelligent agents to become a widely accepted

technology, new models must be proposed and the

ideas to create them may be based on the Cloud Com-

puting model and service orientation, as will be de-

tailed later.

1.2 Cloud Computing

The classic conception of software consumed as a

product is being shifted by that of a service housed

in third party installations. Thus, the software is con-

tracted in accordance with needs of the user, who pays

for the speciﬁc time, quality and functionalities that

they need. This newway tomarketing computer prod-

uct and services is known as Cloud Computing (Arm-

brust et al., 2009).

The spreading of this emerging paradigm is at-

tributed to: (1) the strong ubiquitous and ﬂat character

of communications; (2) the progressive orientation to

software solution services; (3) and the maturity of vir-

tualization tools.

The experts coincide in emphasizing that the prop-

erty that best describes this new paradigm is the fact

the clients only have to concern themselves with what

they can do with the technology and not with how to

install and conﬁgure it. They also coincide in high-

lighting the following aspects as the most signiﬁcant

advantages of the Cloud Computing model (Plummer,

2008):

• It enables ﬁrms to scale up rapidly in accordance

with their requirements by eliminating the need to

add new equipment, software or personnel.

• It signiﬁcantly reduces the costs of maintenance

and energy consumptionby movingthe scalability

to the cloud.

• It allows solutions to be implemented in a much

more streamlined fashion since it eliminates much

of the investment stage.

• It expands the technological options of businesses

by smoothing the path for them to embark on

projects of greater importance without the costs

that they would normally entail.

• It makes it possible to delegate the execution of

critical tasks to specialized services.

• Informatics as a service can give way to move-

ments of cooperation in the cloud. There is not

doubt that it predisposes the net to adopt standards

of communication and interoperation.

2 AGENTS AS INTERMEDIARIES

The objective of middle agents is to assist in com-

munication tasks in order to facilitate exchanges be-

tween requesters and providers. Requesters are those

agents with objectives they wanted to be achieved by

other agents, whereas providers are those agents that

fulﬁll objectives on behalf of other agents. The pres-

ence of middle agents is especially valuable in dis-

tributed, open environments, where they constitute

a mechanism to overcome the heterogeneity of the

partners. Although there are several roles for middle

agents, three of them are mainly recognized (Klusch

and Sycara, 2001):

• Matchmaker. Its functionality corresponds to

that of the Yellow Pages. Providers register their

skills in the matchmaker agent. Requesters con-

sult it in order to identify those providers that are

capable of fulﬁlling their objectives. If the activ-

ity of the matchmaker is successful, the requester

and the provider then enter into a new dialogue.

• Blackboard. Requesters send their petitions to

the blackboard agent, and providers look up peti-

tions that they can fulﬁll in the blackboard. The

blackboard agent keeps track of the requests and

their respective answers so that other agents can

easily extract information later.

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

310

• Brokering. The broker agent registers the skills

of the providers and negotiates the requesters’ pe-

titions with them, and ﬁnally gives the results of

that process to the requesters. In the models dom-

inated by a broker, there is no direct interaction

between the requester and the provider.

For its part, FIPA, deﬁnes the protocol for the

interactions mediated by brokering agents (FIPA,

2002). In short, the agent that initiates the interac-

tion (initiator) delegates the accomplishment of a task

to a broker. After sending the request, the initiator

plays no further part in the process. In general, the

FIPA speciﬁcation proposes neither client autonomy

nor the need to relieve the client of the technical de-

tails.

3 AGENCY SERVICES

3.1 The Model

The Agency Services can be deﬁned as middle agents

that are instantiated in the cloud and, after being con-

tracted as services, represent the users in virtual soci-

eties by performing brokering and management tasks.

The objective of the Agency Services model is to

transfer the complexity of the agents’ activity to the

cloud. However, that does not mean that the presence

of an agent in the client node should be eliminated. In

this sense, the agency service agent is an abstraction

of two agents: a broker agent instantiated in the cloud

as a service, and a local agent that is light but sufﬁ-

cient to represent the interests of the users and to give

them autonomy.

In one hand, the tasks that should be delegated to

the broker agent in the cloud are all those that are

complex, such as the planning, coordination, coop-

eration and negotiation processes, and the algorithms

necessary to solve special problems. In the other

hand, the client is a software agent who is respon-

sible for the simple, easily automated tasks, which

includes: monitoring of local resources, response to

events, meeting the broker agent’s requests, and send-

ing it information such as the state of the resources or

new user’s directives.

Basically, the model proposes a functional break-

down of the tasks that a single agent usually tackles.

Now, the client will be represented by two intercom-

municating agents who together are able to accom-

plish the user’s objectives. In order not to lose the

simplicity that characterizes Cloud Computing, both

agents are supplied by a provider of Agency Services

located in the cloud. In that respect, it is particularly

important that the installation of the local agent is au-

tomated in a way that requires no intervention by the

user.

The new model (Figure 1) is necessarily made

up of providers with sufﬁcient resources to deploy

an infrastructure of Agency Services in the cloud.

These Agency Services Providers (thereafter called

ASPs) are contracted by users to participate in Busi-

ness Sites, i.e. virtual environments. Thus, the princi-

pal model is deﬁned by three elements:

• Agency Services Providers (ASPs). Technologi-

cal ﬁrms with the ability to deploy services archi-

tectures on internet. They offer clients brokering

services based on agents capable of participating

in Business Sites.

• Business Sites. A virtual environment on inter-

net in which all the ASPs that adhere to the stan-

dards of communication and its prescribed social

conventions can participate. Common examples

of Business Sites are the auction and buy-and-

sell sites on internet, grid computing organiza-

tions and applications based on Cloud Computing.

• Client. A user or company that contracts the

Agency Services in order to participate in Busi-

ness Sites from which it expects to make proﬁt.

Basically, clients do not have the necessary re-

sources to develop their own solutions, or to do so

proﬁtably. In the client node, a light agent capa-

ble of communicating with the contracted Agency

Service is instantiated.

Figure 1: Agency Services model.

This conception provides all the features of the

service orientation to the agency services model. In

general, a ﬁrm can provide Agency Services for many

types of problem and Business Sites while, due to the

deﬁnition of interfaces and social conventions, a Busi-

ness Site is able to accept the participation of agents

instantiated by different ASPs.

Figure 2 presents an example of the proposed

model. It includes three ASPs (P

, P

), two Busi-

ness Sites (B

, B

) and four clients (C

, C

ClientsC

andC

are interested in participating in Site

AGENCY SERVICES - An Agent-based and Services-oriented Model for Building Large Virtual Communities

311

Figure 2: Scenario based on Agency Services.

. To that end, the ﬁrst of those clients uses the ser-

vices of P

, and the second those of P

. Moreover,

Clients C

and C

wish to participate in Site B

and,

with that aim, C

uses the services of P

while C

uses

those of P

. In the diagram, P

represents the class

of provider able to offer services for more than one

environment. The ﬁgure also shows the fact that, as

a general rule, Business Sites must release communi-

cations interfaces based on standards to which more

than one provider can subscribe.

More speciﬁcally, the life cycle of the proposed

model comprises the following steps:

1. In order to participate in a Business Site, the client

sends a request to an ASP that is suitable for the

task and with which the client has previously reg-

istered.

2. The ASP deploys a light software agent in the

client that is able to communicate with the user

(Thanks to the technologies mentioned in Section

3.2). At the same time, it requests a broker agent

in its servers, who will subsequently be responsi-

ble for representing the user in the Business Sites.

3. If necessary, the user deﬁnes his/her directivesand

preferences on the interface provided by the local

agent, who then sends that information to the re-

mote broker agent.

4. The broker agent contacts with the Business Site

and initiates its activity on that site. Depending on

the context and the activity to be undertaken, the

broker agent may need information from the local

agent, such as the state of the local resources, or

inform it of the state of his participation in the en-

vironment. The local agent may also send new

directives to the broker agent or respond to the

events and information that the broker has com-

municated.

5. After the broker agent’s participation has ended,

it informs the local agent of the result. That infor-

mation may also be registered in the ASP and will

be useful in terms of the contract linking it to the

user.

Nothing prevents the user from having more than

one local agent assigned at the same time. Thus, a lo-

cal agent can automatically inform the broker about

the state of the resources, meanwhile the user can

use another agent installed in his/her mobile phone

for both receiving information from the broker and,

if necessary, sending it new directives. Therefore,

the broker agent, apart from participating in Business

Sites, can also work as a proxy agent able to commu-

nicate the local agents among themselves.

3.2 Software Technologies

For the proposal to be transparent, realistic and ef-

fective, the technology required by the client must be

simple. Thus, removing the brokering tasks from the

context of the user would not make sense if the client

later encounters difﬁculties in establishing or using a

module capable of interacting with the broker agent.

In that respect, by exploiting the latest technolo-

gies, it is possible to achieve solutions that, for most

scenarios, do not involve the user in the conﬁgura-

tion or installation processes. This paper emphasizes

the instantiation of remotely transmitted software in

the client. More speciﬁcally, the paper proposes the

adoption of:

• JNLP. For downloading and instantiation of the

local agent in the client node. Java Network

Launching Protocol (JNLP) permits the download

and launch of Java applications located in remote

servers (Zukowski, 2002). The following features

stand out: (1) it permits applications to be acti-

vated by using a link; (2) it ensures that the lat-

est version of the application is always run; (3)

it eliminates complex procedures of installation

and updating; (4) it only requires the Java Run-

time Environment (JRE); (5) it permits digitically

sign the application.

• JXTA or XMPP. For communicationbetween the

local agent and the broker. Both JXTA (JXTA,

2007) and XMPP (XMPP, 2009) are P2P proto-

cols. The following characteristics stand out in

both technologies: (1) they are protocolson which

light clients can be simply developed; (2) the con-

tent of the messages is coded in XML; (3) the only

infrastructure required is the installation of a mes-

senger server, which can be installed in the ASP.

JNLP and XMPP/JXTA constitute a solution that

meets the conditions of simplicity, remote instan-

tiation and bidirectional communications described

above. Moreover, they have characteristics that make

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

312

them specially attractive for open and distributed en-

vironments:

• In the Cloud Computing model, it is essential not

to ignore mobile devices, which, in the near fu-

ture, might become an important mean to access

the web. In that respect, the presence of Java is

practically indispensable nowadays.

• In an environment that is open and consequently

inhabited by heterogeneous sources, standard ver-

satile codes like XML are even more important.

• In open environments it is essential to use tech-

nologies that ensure that the communications are

secure, reliable and conﬁdential. In this regard,

XML presents mature solutions.

Moreover, the technology used in the communi-

cation between the broker agent and the Business

Site must be chosen by the latter. It is the Business

Site that determines the communications interface to

which the ASPs wishing to participate in its life cycle

must subscribe. Here it is not required to be guided

by the principal of simplicity: it is now supposed

that both parties have sufﬁcient technology to develop

elaborate and efﬁcient solutions.

3.3 Advantages of Agency Services

In general, the advantages of the concept of Agency

Services are:

• Horizontality. A solution that, with all its virtues,

can be applied to a wide range of domains.

• Participation. The transfer of the most tedious

tasks to the cloud, together with the simplicity of

the technologies instantiated in the client, makes

it easier for users to automate their participation

in virtual societies.

• Scalability. Since the most complex part of the

technology is located in the cloud, the solutions

can increase with no effort required from the

client.

• Autonomy. Bidirectional communications offer

both of them, the broker and the client, the pos-

sibility of being able to exchange dialogue while

the activity is developed in the Business Site. This

guarantees the autonomy of the users since they

can create scenarios in which they are kept in-

formed of events and, if necessary, can send new

directives.

• Flexibility. Without making any type of extra ef-

fort, the client can participate in more than one

virtual society at a time, and always with the

provider that best ﬁts its needs.

• Transparency. To participate in a virtual soci-

ety, the clients do not have to develop their own

agents.

• Reliability. The model permits criteria to be es-

tablished for the ASPs in order to guarantee that

the activity of the provided broker agents does not

endanger the stability of the system with selﬁsh or

anti-social behaviors.

• Security. Access to the Business Sites is re-

stricted to a controlled set of ASPs on whose inter-

faces all efforts related to the security of the sys-

tem can be concentrated.

• Business. The nature of the interactions, clearly

oriented to a competitivemodel, helps create busi-

ness activities: a recognized advantage of the

Cloud Computing model in which, this time, the

agency theory can play a role.

4 A PRACTICAL CASE

To illustrate the advantages of the new model and ver-

ify the viability of the proposed technology, a practi-

cal case based on an auction site similar to eBay is

implemented.

The infrastructure developed for the example

comprises two ASPs and one Business Site (the auc-

tion site). To attain an environment that is as similar

as possible to the cloud, each ASP was installed in a

different server and each server was located in a dif-

ferent physical location. Communication between the

ASPs and the Business Site took place by means of

a set of web services that were released by the latter

and basically permitted the broker agents to consult

the catalogue and make bids and offers. The web ser-

vices that the broker can invoke are:

• GetItemsList. This returns the list of items for

which the broker agents can currently bid.

• GetItemInfo. This provides the identiﬁer and de-

scription of the item, the sale price, the highest bid

made for it, and the starting time of the next round

of bidding.

• SetBid. This adds a bid for a speciﬁc item.

• SetItem. This adds a new item to the catalogue or

modiﬁes the price of an existing item.

• SetAccept. This speciﬁes that the last bid for an

item has been accepted by the vendor.

The Agent Communication Language (ACL) over

XMPP is used for communication between the agents.

More speciﬁcally, the broker sends ACL Inform mes-

sages to the local agent, whereas the latter uses ACL

AGENCY SERVICES - An Agent-based and Services-oriented Model for Building Large Virtual Communities

313

Request messages to send new directives to the bro-

ker agent. Both agents are able to initiate interactions

thanks to the bidirectional nature of XMPP.

The client uses a JNLP link for automatically in-

stantiating the local agent. The user interface de-

ployed displays the list of available items, as well as

information about the actions of the broker in the auc-

tion site. Moreover, when necessary, the client can

initiate a dialog window for sending directives to the

broker agent, which is ﬁnally responsible for bargain-

ing (Sandholm, 1999) in the auction site according to

the latest preferences speciﬁed by the user.

The speciﬁc technologies used for the example

were (Figure 3):

• Agency Services Provider:

– JRE. Runtime environment of the broker

agents instantiated in the ASP.

– OpenFire. Server for XMPP communications

between the broker and local agent.

– Jetty. Web application server from which the

local agents can be downloaded by using JNLP.

– Smack API. Java library that implements the

XMPP protocol.

• Business Site (auction site):

– Jetty. Web application server where the web

services to access the auction site are deployed.

• Client:

– JRE. Runtime environment of the local agent.

– Smack API. Java library that implements the

XMPP protocol.

The client can be any device able to run Java ap-

plications, from mobile phones to desktop computers.

Figure 3: Technologies used in the practical case.

5 CONCLUSIONS

This paper introduces the concept of Agency Ser-

vices, which is presented as an effective means of

transferring to the cloud the complexity that the par-

ticipation of intelligent agents in virtual societies cur-

rently entails. The newapproach gives the agencythe-

ory all the virtues of the Cloud Computing model and

of service orientation in general. Moreover, with the

aim of deﬁning a realistic solution, this paper has de-

tailed the technologies necessary to implement a com-

plete solution based on Agency Services, including

the remote instantiation of a light agent in the client in

order to guarantee the autonomy of the user. Also, in

order to illustrate the use of the model, an ASP based

solution for an auction web site is described.

REFERENCES

Armbrust, M., Fox, A., Grifﬁth, R., Joseph, A. D., Katz,

R. H., Konwinski, A., Lee, G., Patterson, D. A.,

Rabkin, A., Stoica, I., and Zaharia, M. (2009). Above

the Clouds: A Berkeley View of Cloud Comput-

ing. Technical Report UCB/EECS-2009-28, EECS

Department, University of California, Berkeley.

FIPA (2002). FIPA Brokering Interaction Protocol Speciﬁ-

cation.

He, M., Jennings, N., and Leung (2003). On Agent-

Mediated Electronic Commerce. IEEE Transactions

on Knowledge and Data Engineering, 15:985–1003.

Hendler, J. (2007). Where Are All the Intelligent Agents?

IEEE Intelligent Systems, 22:2–3.

JXTA (2007). JXTA Community. https://jxta.dev.java.net.

Accessed 28 August 2009.

Klusch, M. and Sycara, K. (2001). Brokering and match-

making for coordination of agent societies: a survey.

Coordination of Internet agents: models, technolo-

gies, and applications, pages 197–224.

Plummer, D. C. (2008). Cloud Computing Myths, Magic

and Mayhem. In Gartner Symposium ITxpo.

Sandholm, T. (2002). eMediator: A Next Generation

Electronic Commerce. Computational Intelligence,

18:656–676.

Sandholm, T. W. (1999). Multiagent Systems: A modern

approach to Distributed Artiﬁcial Intelligence, chap-

ter Distributed Rational Decision Making, pages 201–

258. MIT Press.

Wurman, P. R., Wellman, M. P., and Walsh, W. E. (1998).

The Michigan Internet AuctionBot: A Conﬁgurable

Auction Server for Human and Software Agents. In

Proceedings of the Second International Conference

on Autonomous Agents, pages 301–308.

XMPP (2009). XMPP Standards Foundation.

http://xmpp.org/. Accessed 28 August 2009.

Zukowski, J. (2002). Deploying Soft-

ware with JNLP and Java Web Start.

http://java.sun.com/developer/technicalArticles/Pro-

gramming/jnlp. Accessed 28 August 2009.

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

314