MODELING AND EXECUTING SOFTWARE PROCESSES
BASED ON INTELLIGENT AGENTS
M
elle
F. Aoussat
Computer Science Department/ Faculty of Sciences
SAAD DAHLAB University, BLIDA
BP 270, Blida, Algeria.
M
r
M. Ahmed Nacer
Computer Science Department/ Faculty of Electrical Engineering,
University of Science and Technology Houari Boumediene,
BP 32, El-Alia 16123 Bab-Ezzouar Algeria
Keywords: Modeling software process, in
telligent agents, software process centered environment.
Abstract: This paper presents a new approach for modeling and executing
software processes based on the concept of
multi-agent systems. We introduce the modeling process as one of the most important goal of the agent, and
we use the concept of “intelligent agent” to give more flexibility when adapting software processes to
unexpected changes. This is possible thanks to the multiple capacities of the agent like autonomy and
reactivity.
1 INTRODUCTION
Multiple platforms (CORBA, COM, EJB…) have
been used for distributed software process
development. The distribution requires 1) a
performent interaction models that can be adapted to
all kinds of software process interactions and 2) a
dynamic software process model to support and to
handle all changes and unexpected events that can
happen in the environment.
The concept of the agent was used for the
devel
opment of environments that focus their efforts
on the execution aspects of software process models
using the agent characteristics of interactions
(communications, negotiations, cooperations …) and
actions. However, few works was focused on the
modeling aspect of software processes based agents
and unfortunately, the representative capacities of
the agents were less exploited.
This paper presents our approach for modeling
and e
xecuting software processes based on the
concept of multi-agent system. This approach allows
to modelize dynamic distributed software processes.
Thanks to the agent concept that has features of
autonomy , reactivity and proactiveness, we use the
concept of “intelligent agent” to give more
flexibility when adapting software processes to
unexpected changes, handling unexpected events,
managing exception, etc.
Section 2 presents a brief state of the art in the
dom
ain of process software development based
agents. Section 3 describes our approach for
modeling and executing software processes based
agents. We describe the architecture of our system
through the multiple interactions between the
software process agents and their internal structures.
We show how, in one hand, it is possible to use the
representative capacities for modeling the software
processes, in parallel with action and interaction
capacities; and, in an other hand, how to give more
autonomy to supervise software process execution.
2 SOFTWARE PROCESS
CENTERED ENVIRONMENT
BASED AGENTS
This study is based on the most important software
processes environments based agents: (Wang,
2000a), (Wang, 2001), (Wang, 2002), PEACE+
(Alloui, 1996) and ALLIANCE (Alloui, 2001a).
441
F. Aoussat M. and M. Ahmed Nacer M. (2005).
MODELING AND EXECUTING SOFTWARE PROCESSES BASED ON INTELLIGENT AGENTS.
In Proceedings of the Seventh International Conference on Enterprise Information Systems - DISI, pages 441-444
Copyright
c
SciTePress
+
All the
Agents are
in charge
to manage
the software
process
and data
consistency
+
4 OUR APPROACH FOR
MODELLING AND EXECUTING
SOFTWARE PROCESS BASED
INTELLIGENT AGENTS
Our system is defined in terms of hierarchic
software process agents. We describe below their
roles, their multiple interactions and their internal
architectures. We consider three kinds of agents:
Task Agents: the role is to execute the activities
defined in the software process model. An execution
report is sent to a fragment agent.
Fragment Agents: Each fragment agent is localized
in one workspace and manages its associated
software process fragments. Its goals are to 1)
divide the software process fragments into
sequential tasks, 2) to initialize its task agents with
heir associated sequential tasks and 3) to synthesize
the execution results for the supervisor agent.
Supervisor agent: It is considered at the top of the
agent hierarchy and has the global vision of the
software process execution in all the workspaces;
this is to ensure the global software proxcess
consistency.
The supervisor agent, considered in the top of
the agent hierarchy, has a global vision of the
software process execution in all the workspaces;
this is to ensure the consistency of the global
software process.
The tasks agents, considered at the low
level, are in charge of executing the
different tasks regarding the software
process model. The decisions taken by a
task agent are locally and prior to its
workspace.
4.1 Internal architecture of
the software process agent
After the study of the well-known internal
architecture of the agents, we choose a modular
architecture (Ferber, 1997), (Jennings, 2000). This
architecture is suitable to our software process agent,
where many parallel functions are executed. An
agent architecture is modular and horizontal. Each
agent is composed of six modules that work in
parallel and in cooperation. Each module is
responsible of many internal tasks of the agent. This
is illustrated in fig 2.
Figure 2: Internal architecture of the software process agent
Communication
-Manage the heterogeinity of the
messages
Planning and decision’s
- Take decisions about negociations
and planing of the execution .
Knowledge base manager
-Manage the knowledge consistency
Realization and evaluation
-Execute the planified tasks
-Evaluate the results.
Resources and accointance
Update the state of ressources and
accointances
Software process manager
-Model the softawre process
-Introduce the changes
External Environment.
Interaction between agents of the same level (between neighboor).
Interaction master/slave (between level).
Interaction agent/human.
Task Agents :
Execute the tasks
of the software
process model
Fragment Agent :
Interprete and affect
the tasks to tasks agents.
Supervisor Agent :
Modelize and affect
the software process
fragments.
Responsible1
Redundancy.
Figure 1: Distribution of the software process
Responsible2
Work s
p
ace1 Work s
p
ace2
Human
resources
More
geographical
distribution.
Specialization.
Multiple
interaction flow
MODELING AND EXECUTING SOFTWARE PROCESSES BASED ON INTELLIGENT AGENTS
443
The modules of the agent software process can
be classified into two categories:
1- Modules to manage the internal function
of the agent: These modules are the
communication module, the knowledge base
management module, the resources and
accointance module.
The functions of these modules are the same
for the three kinds of the software process agent
(tasks, fragments and supervisor).
2- Modules for modeling and executing the
software process: The main goal is to manage the
software process modeling, planning, executing
and introducing process changes. These modules
are: software process manager module, planning
and decision module, the realization and evaluating
module.
The functions of these modules are various
according to the kind of the agent. For instance, the
main function of the fragment agent module is to
evaluate the execution results and to decide about
the needed changes if necessary. Each kind of
agent is responsible of a given fragment of
software process and has the responsibility to
manage its representation and the evolution of its
execution ( at the different levels of course).
5 CONCLUSION
This paper presents a new approach for modeling
and executing the software process using the multi-
agent concepts. A state of art, first done, on three
existing environments (CAGIS, PEACE+ and
ALLIANCE) shows that the priority of these
systems is at the execution level of the software
processes and not at the modeling level. The
capacities of the agents that can handle the
modeling level is being unexploited.
Our contribution is the development of a
software process environment based on the concept
of multi-agent system. An agent is, thus,
considered as a part of a software process. In that,
an agent allows not only the execution of the
assigned tasks, but also, a permanent control of the
modeling and the executing of the software
processes. The “intelligence” aspect in our system
reside in the possibility to force (anticipate) the
events that can be happened and, thus, to take a
pertinent decisions at a time.
The hierarchical structure designed in our
system seems beneficial thanks to the cooperative
aspects of the agent. The current work tends to the
development of process fragmentation techniques,
the evaluation techniques and the mobility of the
agents.
REFERENCES
Alloui, I. and Oquendo, F., 1996. PEACE+: A Multi-
Agent System for Computer-Supported Cooperative
Work in Software Process Centred Environments. In
8th International Conference on Software
Engineering and Knowledge Engineering, , Nevada,
USA, IEEE Computer Society Press.
Alloui, I., et al., 2001a. A software framwork for
modeling, executing and flou control of intensive
software processes. LLP/CESALP Laboratory ESIA
Engineering School University of Savoie, France.
Alloui, I., et al., 2001b. Advanced Services for Process
Evolution, Monitoring and Decision Support. Society
for Design and Process Science Printed in the
United States of America Transactions of the
SDPS,Vol. 5, No. 4, pp. 39-53
Ferber, J., 1997. the book, Les systèmes multi-
agents :vers une intelligence collective. Laforia,
université de Pierre et Marie Curie, interEditions.
Jennings, N., R., et al., 2000. Autonomious Agent for
busness process managment. Dept. Electronic
Engineering, Queen Mary & Westfield College,
University of London, London E1 4NS,UK.
Wang, A., 2002. A Process Centred Environment for
Cooperative Software. Seke’02 july 15-19 Ischia,
Italy. ACM 1-58-113-556-4/02-070.
Wang, A., 2001. Using an environment based agent to
support a cooperative software processes .
Norvegian PhD thesis, Trondheim, Norway.
Wang, A., 2000a. Support for Mobile Software
Processes in CAGIS, In seventh europeen workshop
on softwareSprocess technologie, Kaprun near
Salzbourg, Austria.
Wang, A., 2000b. Using Software Agents to Support
Evolution of distributed Workflow Models. In Proc.
International ICSC Symposium on Interactive and
Collaborative Computing (ICC’2000), page7,
Wollongon (near Sydney), Australia.
Wang, A., 2000c. Implementing a multi-agent
architecture for cooperative software engineering. In
twelfth international conference on software
ingineering and knowledge ingineering seke2000,
chicago, USA.
ICEIS 2005 - DATABASES AND INFORMATION SYSTEMS INTEGRATION
444
