Autonomic Multi-agent Systems

Nadir K.Salih 1 # G.K.Viju 2 *

1. Doctorate Student, Harbin Institute of Technology, Computer Science and Engineering, ICES
2. rofessor, Department of Computer Science, !arary "ni#ersity, !hartoum, Sudan
# nadircom2!"gmail.com * #ijug$2%"gmail.com
Abstract-
The important of autonomic computing is to
create computing systems capable of managing
themsel#es to a far greater e$tent than they do
today. This paper mi$ed multi%agent system &ith
autonomic feature that completely hides its
comple$ity from users'ser#ices. (entioned )*DE as
platform e$ample of this en#ironment, could
applied to &eb ser#ices as front end to users. +ith
multi agent support it also pro#ides adaptability,
intelligence, collaboration, goal oriented
interactions, fle$ibility, mobility and persistence in
soft&are systems.

!ey&ords, *utonomic, multi%agent system, &eb
ser#ices
&. &ntroduction
Agent-Oriented Programming (AOP) is a
relatively new software paradigm that brings
concepts from the theories of artificial intelligence
into the mainstream realm of distributed systems.
AOP essentially models an application as a collection
of components called agents that are characterized by,
among other things, autonomy, proactivity and an
ability to communicate. eing autonomous they can
independently carry out comple!, and often long-
term, tas"s #$%. &ntelligent autonomic agent must
build and maintain a model of the e!ternal
environment and of its own components. Atop-level
e!ecutive component ma"es decisions based on the
models and its current emotional state. A planner
component is used to create multiple step scripts or
se'uences of actions necessary to achieve the high-
level goals being pursued by the e!ecutive#(%. Agents
have the capability to move from one environment to
another see fig.$. )he agent design, using *ra+a,
-advanced behavior. wrappers (li"e autonomously
search according to the agent "nowledge of the user
or planning strategy to arrive to the target point)#/%.
)he autonomy of each agent and the messaging
interface are useful in most of fle!ible and e!tensible
systems. ecause agents are not directly lin"ed to
others, then it is easy to ta"e one out of operation or
add a new one while the others are running #0%.
+ulti-agent development has emerged as a viable
approach to meet the autonomic system
re'uirements-autonomy, adaptability, intelligence,
goal-oriented interaction, collaboration, and
fle!ibility. 1sing multiagent approach, real-world
problems can be modeled in the form of autonomous,
interacting agent components #2%.
An autonomic system is an autonomous
computing environment that completely hides its
comple!ity. 3omple!ity hiding from users4services
means that autonomic computing will provide users
with a computing environment that allows them to
concentrate on what they want to do without
worrying about how it has to be done #$(%. )he
characteristics of +ultiagent ,ystems(+A,s) are
that ($) each agent has incomplete information or
capabilities for solving the problem and, thus, has a
limited viewpoint5 (() there is no system global
control5 (/) data are decentralized5 and (0)
computation is asynchronous#$/%.
)he paper is organized as follows6 ,ection &&
reviews 7elated 8or" ,ection &&& focuses on the
Programming 9anguage and )ools. ,ection &:
describes ;A<= And the Agents Paradigm. *inally,
,ection : ta"e )he 1tility of Agent and 8=
,ervice &ntegration before concluding the paper.
&&. 'elated (or$
)he 1nity system components are implemented
as autonomic elements> individual agents that
control resources and deliver services to humans and
to other autonomic elements. =very 1nity component
is an autonomic element. )his includes6 computing
resources (e.g. databases, storage systems, servers),
higher-level elements with some management
authority (e.g. wor"load managers or provisioners),
and elements that assist other elements in doing their
tas"s (e.g. policy repositories, sentinels, bro"ers, or
registries)#?%. +AA3= emphasis on self-organization
and self-healing of application services and it is an
open and e!tensible computing environment to allow
heterogeneous agent to @oin it. y the cooperation of
agent federation system, agent mediate system and
agent monitor system, +AA3= lead to automated
$
control and management of a wide range of networ"
centric applications and services #$A%.)he ean
Benerator is a tool that supports agent engineers in
creating message content ontologies compliant with
the ;A<= support. )he tool is a plug-in for ProtCgC,
which is a commonly used ontology editor that
enables engineers to graphically model ontologies.
*urthermore, additional functionality and storage
formats can be Dplugged inE to the system. Another
advantage of the ProtCgC tool is that other ontologies
can be imported. 7epositories of e!isting ontologies
ranging from biological domains to mar"et place
product and service descriptions can be found at the
ProtCgC community page and at the <A+9 site$. )he
languages used to represent these ontologies can be
F+9, 7<*, <A+9-O&9, F+&, ,G9 or 1+9#H%. )he
jadem) ;A<= add-on, which provides two ma@or
capabilities6 the ability to interface ;A<= agents with
;ava ;+F (;ava +anagement =!tensions) and the
ability to unit test ;A<= agents using ;1nit. ;adem!
is available for download from the third-party
software area of the ;A<= website. =veryday, useful
software systems rarely e!ist in isolation. &ndeed, one
of the strengths of ;A<= is that the full capabilities
of the ;ava environment are available when creating a
software agent application. ;+F is the ;ava
technology for management and monitoring of
software systems5 it was originally part of the ;ava
== enterprise platform (formerly "nown as ;(==),
but as of ;ava H it is available as part of the standard
;(,= environment. *urthermore, unit testing is an
important techni'ue for the development of robust
software and ;unit is a standard methodology for the
unit testing of applications written in ;ava. *adem)
was developed for an industrial software agent effort
re'uiring management using ;ava == and to be unit-
testable. A jadem) agent can be configured either
programmatically or by using F+9 #$%. )he ;ava
Sni++er is a stand-alone ;ava application, developed
by 7oc"well Automation, &nc., that can remotely
connect to running ;A<= systems and is intended as
an alternative to ;A<=Es built-in sniffer. )he tool
receives messages from all agents in the system,
reasons about the information, and presents it from
different points of view (see *ig.()#I%. )he *ade)
elief <esire &ntention <& reasoning engine that
allows development of rational agents using
mentalistic notions in the implementation layer. &n
other words, it enables the construction of rational
agents following the <& model. &n contrast to all
other available <& engines, ;ade! fully supports the
two-step practical reasoning process (goal
deliberation and meansJend reasoning) instead of
operationalizing only the meansJend process. )his
means that ;ade! allows the construction of agents
with e!plicit representation of mental attitudes
(beliefs, goals and plans) and that automatically
deliberate about their goals and subse'uently pursue
them by applying appropriate plans. )he reasoning
engine is clearly separated from its underlying
infrastructure, which provides basic platform services
such as life-cycle management and communication.
Kence, running ;ade! over ;A<= combines the
strength of a well-tested agent middleware with the
abstract <& e!ecution model. *or the programming
of agents, the engine relies on established techni'ues,
such as ;ava and F+9 and, to further simplify the
development tas", ;ade! includes a rich suite of run-
time tools that are based upon the ;A<=
administration and debugging tools. &t also includes a
library of ready-to-use generic functionalities
provided by several agent modules (capabilities)#L%.
O+A3, is a metamodel for agent organizations. &t
defines the re'uired organizational structure that
allows multiagent teams to autonomously reconfigure
at runtime, thus enabling them to cope with
unpredictable situations in a dynamic
environment#$$%. +AB=, an agent-oriented
programming environment, with complete tools to
support agent-based re'uirement analysis, design,
development and deployment, is a powerful
development environment for autonomous computing
#$0%.
,igure 2 JavaSnifer user interface
&&&. -rogramming .anguage and /ools
+ulti-agent systems programming languages,
platforms and development tools are important
components that can affect the diffusion and use of
agent technologies across different application
domains. &n fact, the success of multi-agent systems
is largely dependent on the availability of appropriate
technology (i.e. programming languages, software
libraries and development tools) that allows relatively
straightforward implementation of the concepts and
techni'ues that form the basis of multi-agent systems.
(
+ulti-agent systems can be realized by using any
"ind of programming language. &n particular, ob@ect-
oriented languages are considered a suitable means
because the concept of agent is not too distant from
the concept of ob@ect. &n fact, agents share many
properties with ob@ects such as encapsulation, and
fre'uently, inheritance and message passing.
Kowever, agents also differ from ob@ects in several
"ey ways5 they are autonomous (i.e. they decide for
themselves whether or not to perform an action on
re'uest from another agent)5 they are capable of a
fle!ible behavior5 and each agent of a system has its
own thread of control. An important characteristic
that multi-agent systems should provide is the
capability to support interoperability among legacy
software systems. )herefore, the availability of
software tools for their integration with other
common technologies can be a "ey to their success.
)he &nternet is one the most important application
domains and the most important communication
means that multi-agent systems can use to provide
interoperability among legacy software systems5
therefore, a lot of current research and development
wor" is oriented towards providing suitable
techni'ues and software tools for the integration of
multi-agent systems with 8eb technologies such as,
for e!ample, 8eb services and ,emantic 8eb
technologies.
&V. *A01 And the Agents -aradigm
;A<= is a software platform that provides basic
middleware-layer functionalities which are
independent of the specific application and which
simplify the realization of distributed applications
that e!ploit the software agent abstraction. A
significant merit of ;A<= is that it implements this
abstraction over a well-"nown ob@ect-oriented
language, ;ava, providing a simple and friendly AP&.
)he following simple design choices were influenced
by the agent abstraction. An Agent is Autonomous
and Proactive6 An agent cannot provide call-bac"s or
its own ob@ect reference to other agents in order to
mitigate any chance of other entities coopting control
of its services. An agent must have its own thread of
e!ecution, using it to control its life cycle and decide
autonomously when to perform which actions. )he
,ystem is Peer-to-Peer =ach agent is identified by a
globally uni'ue name (the Agent &dentifier, or A&<,
as defined by *&PA). &t can @oin and leave a host
platform at any time and can discover other agents
through both white-page and yellow-page services
(provided in ;A<= by A+, and the <* agents as
defined also by the *&PA specifications). An agent
can initiate communication with any other agent at
any time it wishes and can e'ually be the ob@ect of an
incoming communication at any time. On the basis of
these design choices, ;A<= was implemented to
provide programmers with the following ready-to-use
and easy-to-customize core functionalities6-
M A fully distributed system inhabited by agents,
each running as a separate thread, potentially on
different remote machines, and capable of
transparently communicating with one another, i.e.
the platform provides a uni'ue location-independent
AP& that abstracts the underlying communication
infrastructure.
M =fficient transport of asynchronous messages via
a location-transparent AP&. )he platform selects the
best available means of communication and, when
possible, avoids marshalling4unmarshalling @ava
ob@ects. 8hen crossing platform boundaries,
messages are automatically transformed from ;A<=Es
own internal ;ava representation into proper *&PA-
compliant synta!es, encodings and transport
protocols.
M ,upport for agent mobility. oth agent code and,
under certain restrictions, agent state can migrate
between processes and machines. Agent migration is
made transparent to communicating agents that can
continue to interact even during the migration
process.
M A set of graphical tools to support programmers
when debugging and monitoring. )hese are
particularly important and comple! in multi-threaded,
multi-process, multi-machine systems such as a
typical ;A<= application.
M &ntegration with various 8eb-based technologies
including ;,P, servlets, applets and 8eb service
technology. )he platform can also be easily
configured to cross firewalls and use NA) systems.
M An in-process interface for launching4controlling
a platform and its distributed components from an
e!ternal application.
V. /he 2tility o+ Agent and (13 Ser#ice
&ntegration
&ntegrating 8eb services and software agents
brings about an obvious benefit6 connecting
application domains by enabling a 8eb service to
invo"e an agent service and vice versa. Kowever, this
interconnection is more than simply cross-domain
discovery and invocation5 it will also allow comple!
compositions of agent services and 8eb services to
be created, managed and administered by controller
agents. )o the users of 8eb services, whether human
or computational, agents can be a powerful means of
indirection by mas"ing the 8eb service for purposes
of, for e!ample, redirection, aggregation, integration
and administration. 7edirection describes the case
where a 8eb service may no longer be available for
some reason, or the owner of the 8eb service wishes
to temporarily redirect invocations to another 8eb
service without removing the original
/
implementation. Aggregation allows several 8eb
services to be composed into logically interconnected
clusters, providing patterned abstractions of behavior
that can be invo"ed through a single service interface.
&ntegration describes the means of simply ma"ing
8eb services available to consumers already using,
or planning to use, agent platforms for their business
applications, and administration covers aspects of
automated 8eb service management where the agent
autonomously administers one or more 8eb services
without necessary intervention from a human user.
V&. 4onclusion
Autonomic 3omputing is providing new vistas
in reducing the comple!ity incurred in todayEs
distributed systems. &t minimizes human intervention
and reduces the administration cost of enterprise &)
systems. 8ith multi agent support it also provides
adaptability, intelligence, collaboration, goal oriented
interactions, fle!ibility, mobility and persistence in
software systems.
&n this paper, we mentioned ;A<= was
implemented to provide programmers with the ready-
to-use and easy-to-customize core functionalities .An
Agent is Autonomous and Proactive. &n addition we
have recommended an agent-8eb service that has the
features of both the agent technology as well as the
8eb services technology and is managed by an
autonomic system based on multi-agent support. )his
can help to develop enterprise &) systems that are
optimal, highly available.
'e+erences
#$% ;ohn 8iley, ,ons 9td, (AAL, <eveloping +ulti-
Agent ,ystems with ;A<=
#(% ;. P. igus <. A. ,chlosnagle, A tool"it
for building multiagent autonomic systems, &+
,ystems ;ournal, :ole 0$, NO /, (AA(
#/% Kenri Avancini, AnalOa Amandi. A ;ava
*ramewor" for +ulti-agent ,ystems, ,A<&O
=lectronic ;ournal of &nformatics and Operations
7esearch, vol. /, no. $, pp. $-$( ((AAA).
#0% *atemeh <aneshfar, Kassan evrani. +ulti-Agent
,ystems in 3ontrol =ngineering6 A ,urvey, Kindawi
Publishing 3orporation ;ournal of 3ontrol ,cience
and =ngineering :olume (AA?, Article &< H/$A2A, $(
pages
#H% Alessio Bosca, Dario Bonino. Ontology
=!ploration through 9ogical :iews in ProtCgC, $2
th
&nternational 8or"shop on <atabase and =!pert
,ystems Applications
&===, (AAL.
#I% Pavel :rba, Pavel )ich. 7oc"well AutomationEs
Kolonic and +ultiagent 3ontrol ,ystems
3ompendium, &=== )ransactions On ,ystems, +AN,
And 3ybernetics>PA7) 36 Applications And
7eviews, :ole. 0$, NO. $, ;anuary (A$$
#L% *ran" 3hiang, 7obin raun, Autonomic ,ervice
3onfiguration for )elec-munication +A,, with
=!tended 7ole-ased BA&A and ;A<=!. (AAH &===
#2% Bilda Pour, +ulti-Agent Autonomic Architectures
for Guality 3ontrol ,ystems, ,an ;ose ,tate
1niversity ,an ;ose, 3A, 1.,.A.
#?% Gerald Tesauro, David M. Chess. A +ulti-
Agent ,ystems Approach to Autonomic 3omputing,
AA+A,EA0, ;uly $?-(/, (AA0, New Por", New Por",
1,A
#$A% ;un 8, ;& Bao, ei-,hui 9iao, ;iu-;un 3hen.
+ulti -Agent ,ystem ased Autonomic 3omputing
=nvironment, Proceedings of the )hird &nternational
3onference on +achine 9aming and 3ybemetics,
,hanghai, (I-(? August (AA0.
#$$% 8alamitien K. Oyenan and ,cott A. <e9oach.
<esign and =valuation of a +ultiagent Autonomic
&nformation ,ystem. &nternational 3onference on
&ntelligent Agent )echnology (AAL &===48&34A3+.
#$(%Kuaglory )ianfield. +ulti-Agent Autonomic
Architecture and &ts Application in =- +edicine,
Proceedings of the &===48&3 &nternational
3onference on &ntelligent Agent )echnology, (AA/.
#$/% Qatia P. ,ycara. +ultiagent ,ystems, )his &s a
Publication of )he American Association for
Artificial &ntelligence, $??2
#$0% Rhongzhi ,hi, Kai@un Rhang, Pong 3heng.
+AB=6 An Agent-Oriented ,oftware =ngineering
=nvironment, Proceedings of the )hird &===
&nternational 3onference on 3ognitive
&nformatics,(AA0
0