International Journal of Advanced Engineering Technology
Research Article
A FORMAL APPROACH FOR AGENT BASED LARGE CONCURRENT INTELLIGENT SYSTEMS 1
Chaudhary Ankit, 2Raheja J L
Address for correspondence Dept. of Computer Science, BITS Pilani, Rajasthan, INDIA-333031 Email:
[email protected] 2 Digital Signal Processing Group, CEERI, Pilani, Rajasthan, INDIA-333031 Email:
[email protected] 1
ABSTRACT Large Intelligent Systems are so complex these days that an urgent need for designing such systems in best available way is evolving. Modeling is the useful technique to show a complex real world system into the form of abstraction, so that analysis and implementation of the intelligent system become easy and is useful in gathering the prior knowledge of system that is not possible to experiment with the real world complex systems. This paper discusses a formal approach of agent-based large systems modeling for intelligent systems, which describes design level precautions, challenges and techniques using autonomous agents, as its fundamental modeling abstraction. We are discussing Ad-Hoc Network System as a case study in which we are using mobile agents where nodes are free to relocate, as they form an Intelligent Systems. The designing is very critical in this scenario and it can reduce the whole cost, time duration and risk involved in the project. KEY WORDS: Intelligent Systems, Autonomous Agent, Design Level Priorities, Ad-hoc Networks INTRODUCTION
these problems. One of the new approaches
Development teams during the system
that have been proposed is agent-based
development
uses
modeling. The fundamental notion on which
functional analysis and data flow, or object-
Agent based engineering is autonomous
oriented modeling, which are not sufficient
agent. One Key reason to consider an agent
in many cases in themselves, to capture
as an autonomous system is capable of
today’s dynamic and flexible requirements
interacting with other agents in order to
of some of the current complex projects that
satisfy its design objectives, and a naturally
are undertaken. Researchers are now seeking
appealing one for system designers [1].
new methods and approaches that can help
Agent modeling in system engineering is a
System designers to grapple with some of
relatively young area, and there are, as yet,
life
cycle,
mostly
IJAET/Vol. I/ Issue I/April-June, 2010/95-103
International Journal of Advanced Engineering Technology
no standard methodologies, development
places. In systems also if human interaction
tools, or system architectures. System can be
is not available then also intelligent agent
defined using multi agents as they can work
can perform the functions according to the
concurrently to increase the performance of
environment changes. It may be predefined
system. A design level strategy is needed to
or can be perform by agent by self learning.
secure the fortune of designed system and to
It can be called about the agent that
care all the coming problems in advance.
intelligent behavior is the selection of
Agents also have their several different
actions based on knowledge. Agent can be
kinds of problems that need a different kind
destructive also like computer virus [4], if
of treatment. Our work focus is on the
modeled badly. Consequently the promoters
design phase for large intelligent systems.
of Agents argue that agent-based approaches can significantly enhance our ability to
AGENT-BASED MODELING
model, design and build complex systems,
The idea of agents involves from artificial
as they provide different flexibilities. Agent-
intelligence and neural networks. To show
based techniques can be implemented during
intelligence and self learning mode in agents
the
for self learning Intelligent Systems, these
techniques can be used to model the
approaches are highly needed. Those behind
problem domain and the system design, and
this movement assert that key techniques for
during the implementation phase, where
managing
as
agent-based development tools could be
decomposition, abstraction, and organization
used. The use of the agent based approach in
can be comfortably accommodated within
both phases is a natural fit, but not required.
the agent-based modeling. A little work has
The use of agent-based methodology is
been done in this area by Luck [2], who uses
increasing rapidly in industrial sector, for
Z formal language for different level agent
simulated
designing frameworks. Fisher [3] used
Sensor specific applications and other
different logics to check and model for
critical systems, this is beneficial. However,
agents
like
this is not necessary to use both design
METATEM. An agent can be treated as a
approaches (Object based Vs Agent Based)
human agent that works on behalf of a
together; it would fail to take advantage of
person and can represent the person at some
the natural mapping from one phase to the
complexity,
and
their
such
executions
IJAET/Vol. I/ Issue I/April-June, 2010/95-103
design
phase,
where
applications,
agent-based
motor
control,
International Journal of Advanced Engineering Technology
other. People like Hilaire [5] have tried to
is against the nature of Intelligent
link formal approaches and agent modeling.
systems.
Here we will make reference to agent-based
2.
In object orient methods action
design and our work will concentrate on
choice is not defined, any member
agent-based development frameworks for
can invoke any publicly available
Intelligent Systems.
object [4], while in Intelligent
AGENT VS OBJECT MODELED
Systems, system have to take
DESIGN
action
An intelligent system which makes the life
environmental changes.
simpler like intelligent washing machine or
Agent
according
based
systems
to
the
are
very
microware oven, are harder to build than
complicated in nature. On the other hand
other simple same kind of system. In certain
object-based does not provide a good set of
domain of problems they are very critical
methodology to form a model of these
and time bounded, so that they become very
systems. So to make easy understandability
complicated.
and strong high level design the component
Domain
like
architecture
is
manufacturing systems, intelligent systems
computer
software
have different perspectives , so their
autonomously without human intervention.
designing is very critical and all system
Agent-based programming is the extension
success is dependent on the designing of
of object-based programming with some
system. As the easiness of the system
improvement. The idea behind agent-based
increases,
the
systems, are that they are capable to
implementation increases. The design itself
reconfigure or operate themselves whenever
is very complicated. People think about
they needed. OO methodologies are not
object-based
certain
directly applicable to agent systems. An
difference between object- based and agent-
Agent based Models of G-Nets are shown in
based design.
Fig 1. Agents are usually significantly more
telecommunications,
1.
the
robotics
complexity
design,
there
control,
of
are
used.
Agents that
are
the
functions
Objects are passive in nature,
complex than typical objects, both in their
without
internal structure and in the behaviors they
invocation
message,
mostly they never active [4], that
IJAET/Vol. I/ Issue I/April-June, 2010/95-103
exhibit [6].
International Journal of Advanced Engineering Technology
Figure 1.A Generic Agent Based Modeling for G-Net [1]
Figure 2:.Issue in Agent Based Modeling Software System IJAET/Vol. I/ Issue I/April-June, 2010/95-103
International Journal of Advanced Engineering Technology
designers to use modular and abstraction ISSUES IN AGENT DESIGN
approach to reduce complexity. Hierarchical
The agent paradigm is based upon the notion of
reactive,
autonomous,
internally-
motivated entities embedded in changing, uncertain environments which they perceive and in which they act [6]. Traditional system engineering approaches offer limited support for the development of intelligent systems. To handle the tremendous complexity and
decomposition is also possible that depend on system. Agents are entities that have predefined
developers
properties of Agents are•
•
many
action
and
Agents are defined in the system to
their internal sub goals that they attempt to fulfill though their actions
of the system are tedious jobs to do.
in the process. These sub goals are
Although we know many techniques from
predefined in the system to access it
software engineering but still a better one is
or to do function according to them.
needed. There can be several issues in the •
Agents have the goals that are part of system goal. Agent can start some
will be system dependent that is applicable
action to fulfill the goals, called pro-
to only that particular application domain.
activity. This is the action in
Here we are discussing main issues or say
advance, based on some information.
properties of agents that should maintain
IJAET/Vol. I/ Issue I/April-June, 2010/95-103
so
fulfill the system goal. They have
among the agents and different functionality
the design techniques [9]. We recommended
perform
Agents
that agent.
(called as meet in agent modeling [8])
the implementation. These issues are above
operation,
with a known state. That is state of
The Complexity of the system is obvious
during the design and should continue till
the
from outside the world they come up
higher-level development constructs [7].
agent based system design, but most of them
During
reactions but when they get accessed
need
because synchronization and interaction
they
change and work according to it. Few
intelligence, self learning, adaptive ness and integration,
mechanically
triggered either from internal or external
the new engineering challenges presented by
seamless
goals,
•
Agents are an intelligent entity, so it can operate itself without the user interrupt. This property is autonomy.
International Journal of Advanced Engineering Technology
•
Agents operate without any direct
all agents should use common
input and
interface system to communicate, or
have
control
over
their actions and internal state.
use a protocol so that they can
Agents have to interact to each other
understand each other clearly.
for different purposes, several times
limitation. It should not be work
times to fulfill the
the
outside the systems boundaries. It
system. Meets have a great role in
should check all system limitation
agent based system modeling. Proper
and try to fulfill its goal that is
communication behavior of agents
system dependence.
goal
of
Agents should be able to work in group of one or more. The group
communicate should be well defined.
behavior
Agents
computational
because all intelligent agents have
elements. They do computation, then
own view on each decision, so it
do some action, reaction, and show
should be defined properly.
are
the
behaviors,
so
their
is
critical
to
perform
Above it there are many other issues that can
complexity will be high.
be application specific to a particular
At design level, it should consider
domain and designers want to add them.
that
These issues can be added as a sub issue or
user
should
unaware
of
complexity.
as independent part. Other issues can be
Agents have to communicate each
mobility,
other and
behavior, security, privacy, performance
it should check that
knowledge
level,
run
time
should
etc., as shown in Fig 2. Mobility is with
not affect the other’s action. Side
respect to location so we have to consider
effects should remove carefully, if
the mobility issue here. Knowledge level for
not possible, put explicit condition to
different agents can be different or same or
have system stable. This is known as
it can be a part of autonomy. This property
cooperation.
defines the system intelligence and help to
Agents communicate each other to
achieve the goal. Security is very important
pass information, so it should be that
aspect and very critical issue that helps to
the action of one
•
•
system through which they will
intelligent
•
Agents should work in the system
for it fulfilling the goal and several
and interface, connectors in the
•
•
agent
make systems invulnerable. IJAET/Vol. I/ Issue I/April-June, 2010/95-103
International Journal of Advanced Engineering Technology
Privacy is another important issue that gives
intelligent agent and have self learning
the privacy to agents but in some systems it
capability. On the other network the other
can create problems, it should be application
Agent (which will be foreign agent (FA) for
specific. Performance has the scalability and
the other network nodes) is working on
time response issue that can be measure
FOREIGN
though
or
Now when the node from home network
modeling. Performance is very important in
visits to the other network then FA detects
every system but critical in Real Time
the node that there is a remote node is in my
systems design.
network. It broadcast the address of this
simulation
of
the
system
ALLOCATION
REGISTER.
node to all and the home agent comes to know that my node is in the network of that particular
agent.
Now
FA
sends
the
information of network address to HA. HA comes to know where its node is and on what address it is operating. Now, when any nodes from outside these two networks want to communicate with that node, so as it, say Figure 3.Mobile Agent Working Scenario
in its home network, it sends messages to
CASE STUDY
the home address. The HA gets the message
ROUTING IN AD-HOC NETWORKS In this section we are taking the routing scenario on nodes where they are free to relocate. These computing nodes are visiting to
other
networks
correspondent node, knows that it should be
and
still
able
to
communicate with their local identity [10]. A mobile node can be anything, from a laptop, a mobile phone, a PDA, a smart phone to an iPod also. It have its home configuration and it is pre registered to its HOME ALLOCATION REGISTER where its home agent(HA) is working, that is an
at home network, read it and check the status of the node in its register. Now as it is in the remote network and it have the new address, it wraps the message with the new address and send it to FA of that network where node resides. FA reads the message and sees the internal address for the node which is residing in its network, so it sends this message to it. The message has sent to its receiver. If this remote node want to reply for the message to the sender, then it need not to be to go all the way back, as it
IJAET/Vol. I/ Issue I/April-June, 2010/95-103
International Journal of Advanced Engineering Technology
knows the address of the node which sent it
The reactivity of an agent can not be known
that message, it directly sends it the message
priory,
with its original address as shown in Fig 3.
intelligence that what it decide at that time.
If the correspondent node gets the message,
At design time may be its not possible to
it will get the address of home network
address all these issue related to reactivity
because the mobile node sent the message
before implementation, But the Proactivity
directly, without interference of FA. So it
of a system should be clear and well
will remain close to correspondent node that
defined.
where the node is. The correspondent node
CONCLUSION AND FUTURE WORK
will think that the node is in its home
In this paper we discuss the issues that have
network only. Here the intelligence of the
very important roles in the designing of
agents has a great role in all operations. The
agent based large concurrent intelligent
agents at the design level should be modeled
systems. This paper show why agent-based
as the interactive, cooperative, autonomous
approach is better than other and what are
and able to learn. Agents interact to each other and also to other nodes and they do different kind of functions that is very important and difficult to think at design level. The cooperation of agent with each
it’s
a
function
of
artificial
the issues related to it at design level. The Proposed approach is applicable to all domains and is not dedicated to any particular system or implementation details yet it is very clear to apply. The data represent is based on analysis of agent-based
other should be defined priory and simulate
systems and propose the difficulty that
them. As it can be that agents are working
generally comes after the design. This paper
well alone but in groups they are showing
gives
different
and
consideration to system designer so that the
Proactivity are the other important issues
problems and critical points could be
that are very critical in this example.
managed and designed carefully, and it
Reactivity is the phenomenon that it
helps in designing better Intelligent Systems.
perceives something from the environment
We present mobile node example that have
and reacts to it. The reaction can be
many
behavior.
Reactivity
predefined or it can perform on its own. The reactivity should be according of the Proactivity. Proactivity is main goal of the system for which, all agents are working. IJAET/Vol. I/ Issue I/April-June, 2010/95-103
the
view
issues
of
involved.
whole
system
Agent-based
techniques have many advantages over other design techniques and give better results. This kind of systems would be much helpful for
military
and
other
places
where
International Journal of Advanced Engineering Technology
intelligent alert is continuously required.
for Systems of BDI Agents”, Technical
Our future work will be based on test and
notes,
resolve these issue in different domains and
Melbourne, Australia, 1996.
MAAMAW’96,
Springer,
7. L. Sterling, T. Juan: “The Software
refine their specific problems.
Engineering of Agent-Based Intelligent Adaptive Systems”, ICSE’05, St. Louis,
REFERENCES
Missouri, USA, May15-21, 2005.
1. H. Xu, S. M. Shatz.”A Framework forModel-Based
Design
Oriented Software”,
of
Agent-
Design
M.d`Inverno,”A
Formal
Engineering”, 10th IEEE Asia-Pacific Software
Procedings
(APSE’03), 2003.
conference
first
International
on
Agent
9. K. Chan and L. Sterling: “Specifying
Framework for Agency and Autonomy”, of
of
Roles within Agent-Oriented Software
Vol. 29, No.1, USA, Jan 2003. Luck,
Goal-Oriented
Systems”, QSIC’05, IEEE, 2005.
IEEE
Transactions on software engineering, 2. M.
8. J. Khallouf, M. Winikoff:” Towards
Multi-Agent
Engineering
10. Kurose
James
Conference
F.,”
Computer
Systems(ICMAS), pp-254-260,1995.
Networking: A Top-Down Approach
3. M. Fisher, “Representing and executing
Featuring the Internet”, third Edition, pp
Agent
Based
Systems”,
Intelligent
Agents- Proceddings of International workshop
on
Agent
Theories,
566-570, Pearson Education, 2005. 11. P. Bresciani, et. al.:”A Knowledge Level Software Engineering Methodology for
Arhcitetures and languages, pp 307-
Agent
323,1995.
AGENTS’01, Quebec, Canada, 28 May-
4. Amund Tveit:”A survey of AgentOriented Software Engineering”, First NTNU CSGSC, May, 2001. 5. V. Hilaire, A. Koukam, P. Gruer, and J.-
Oriented
1 June, 2001. 12. K. Chan, L. Sterling, S. Karunasekera:” Agent-Oriented Australian
Software
Software
P. Mu¨ ller, “Formal Specification and
Conference
Prototyping of Multi-Agent Systems,”
Australia, 2004.
Engineering Societies in the Agent
Programming”,
13. T.
Juan,
A.
Pearce,
“ROADMAP:
Extending
(ESAW), Aug.2000.
Methodology
for
Methodology and Modelling Technique
IJAET/Vol. I/ Issue I/April-June, 2010/95-103
Engineering
(ASWEC’04),
World, First International Workshop 6. D, Kinny, M, Georgeff and A. Rao:”A
Analysis”,
L.
IEEE, Sterling:
the
Complex
Gaia Open
Systems”, AAMAS’02, Bologna, Italy, July
2002.
