A COMPUTERIZED TUTOR FOR ARCHITECTING SOFTWARE
Supporting the Creative Aspects of Software Development
José L. Fernández-Sánchez and Javier Carracedo Pais
Industrial Enginneering School, Madrid Technical University, José Gutierrez Abascal 2, 28006 Madrid, Spain
Keywords: Intelligent agents, software architecture.
Abstract: CASE tools must be more user-oriented, and support creative problem-solving aspects of software
engineering as well as rigorous modelling based on standard notations such as UML. Knowledge based
systems and particularly intelligent agents provide the technology to implement user-oriented CASE tools.
Here we present an intelligent agent implemented as a CASE tool module. The agent guides the software
architect through the architecting process, suggesting him the actions to be performed and the methodology
rules that apply to the current problem context.
1 INTRODUCTION
Software engineering is an established discipline of
engineering where standards, notations, processes
and best practices are currently defined.
The software development process consists of a
mixture of creative and mechanistic activities some
of them performed by humans and the others well
supported by tools.
The creative activities of the software
architecting phase of a real-time system
development are mainly those related to: the
identification of system responses, the identification
and selection of the architecture components, and
the solving of the time and concurrency problems
frequent in these applications.
Current CASE (Computer Aided Software
Engineering) tools can guide software architects in
the software architecting process at the supported
methodology level, basically they can check
consistency of software models and in some cases
verify methodology rules, but they are rather
inefficient at the development process level and
software architect level.
CASE tools must be more user-oriented, and
support creative problem-solving aspects of software
architecting as well as rigorous modelling.
Knowledge based systems and particularly
intelligent agents provide the technology to
implement user-oriented CASE tools where an
intelligent agent implemented as a tool module,
guides the software architect through the
architecting process, suggesting him the actions to
be performed and the methodology rules that apply
to the current problem context.
Based on our previous experience of developing
PPOOA (Pipelines of Processes in Object Oriented
Architectures) method and tool based on UML
(Unified Modeling Language) notation for
architecting real-time systems (Fernandez 2003), we
try to demonstrate that constructing a computerized
tutor for assisting to the development of the
architecture of a real-time system is feasible.
We begin the paper describing the characteristics
of intelligent agents and their role in software
engineering activities. We then present
PPOOA_ATA (Architecting Tutor Agent), its
capabilities, the main components of its architecture
and how it is used. We then close the paper with
some conclusions and how we plan to extend the
architecting tutor agent and the CASE tool with new
features.
2 AGENTS AND SOFTWARE
ENGINEERING
Authors involved in agent research have offered a
variety of definitions for an intelligent agent. It is
not the purpose of this paper to survey the
definitions of intelligent agents but to identify what
is the essence of an intelligent agent, which are its
367
L. Fernández-Sánchez J. and Carracedo Pais J. (2007).
A COMPUTERIZED TUTOR FOR ARCHITECTING SOFTWARE - Supporting the Creative Aspects of Software Development.
In Proceedings of the Second International Conference on Software and Data Technologies - SE, pages 367-370
DOI: 10.5220/0001340003670370
Copyright
c
 SciTePress

main properties and how an intelligent agent can be
applied in the software engineering domain.
2.1 Intelligent Agents
The essence of an intelligent agent is formalized by
Franklin, when he defines it as an entity situated
within and a part of an environment, that monitors
that environment and acts on it over time, in pursuit
of its own plan and so as to change what it monitors
in the future (Franklin 1996).
Franklin also identifies the properties that may
have an agent:
 Reactive: responds in a timely fashion to
changes in the environment;
 Autonomous: exercises control over its own
actions;
 Goal-oriented: may act following a plan;
 Temporally continuous: is a continuous running
process;
 Communicative: can establish dialogs with
other agents including people;
 Adaptive: sensitive to each user´s strengths and
weaknesses;
 Mobile: able to transport itself from one
machine to another;
 Flexible: actions are not scripted;
 Character: apparent personality and emotional
state.
The intelligent agent definition given above satisfies
the first four properties. Adding other properties
may produce useful types of agents for example,
mobile learning agents.
2.2 Applying Agents to Engineer
Software
Effective intelligent agents must deal with the grey
areas of the incomplete function for which it is an
approximation. Software engineering is one of these
grey areas where the adequate output for a given
input is context-dependent. That is, there are
different solutions for the same software problem
(requirements) and the suitability of a solution
(design) depends on the context (project constraints).
The knowledge for building software is buried in
books and manuals or in the heads of software
engineering experts, and how to find and get access
to it is a challenging task. Frequently the software
engineer is blocked in one step of the development
process, having no access to the human expert that
can help and suggest what the software engineer
needs to know at this particular situation.
The availability of a personal computerized tutor
is not time restricted and can help the software
engineer in three ways: by capturing a significant
part of the developing process knowledge; by
reasoning based on this knowledge and received
events; and by automating the application of this
knowledge to the software under development.
Presenting relevant information and proposing
suggestions eases the decisions made by the
software engineer.
Figure 1 represents the interaction of the
computerized tutor, the software engineer and the
CASE tool in the scenario of generic software
development process. The computerized tutor asks
questions to the software engineer and receives
responses from him. The computerized tutor checks
the software models and receives events from the
CASE tool, for example when a new building
element is dragged and dropped in a diagram.
Following the development goals and reacting to
events, the computerized tutor sends suggestions to
the software engineer. So, the expert is available
when the software engineer has a need.
Figure 1: Computerized Tutor in Software Development.
The feasibility of the usage of computerized tutors is
also supported by recent experiences, which have
similarities and differences in goals, scope and
implementation with respect to our computerized
tutor.
Diaz Pace and Campolo report an expert system
being constructed that will support design
modifiability (Diaz Pace 2003). Bachmann, Bass,
Klein and Shelton propose an expert system that
collaborates with the architect to produce a design of
an architecture that supports an expected change
(Bachmann 2004). This rule based architecture
design assistant uses architecture modifiability
attribute models viewed as frames. The next
ICSOFT 2007 - International Conference on Software and Data Technologies
368

attribute they plan to add will be performance.
WayPointer is a commercial tool that provides an
integrated solution for implementation of best
practices for requirements engineering, modelling,
documentation and testing of software. In
WayPointer, an intelligent agent is responsible for
helping the software engineer to accomplish a well-
defined task. The knowledge of the agents is
captured by a set of rules that are defined in an
application specific rule language (Racko 2004).
3 ARCHITECTING TUTOR
AGENT
PPOOA_ATA (Architecting Tutor Agent) can be
considered a hybrid agent exhibiting two different
forms of reasoning: one reactive or event driven, and
other proactive based on action planning and the
architecting process execution for achieving the
software architecting goals.
We describe below PPOOA_ATA, its
capabilities, the main components of the agent
architecture and how the agent interacts with the
user.
3.1 PPOOA_ATA Capabilities
The current prototype of PPOOA_ATA supports the
following capabilities:
 Has knowledge about the PPOOA architecting
process and the architecting guidelines and is
able to use this knowledge to give suggestions
to the software architect;
 Can proactively take action based on the
architecting goals and the architecting process
implemented as a plan;
 Can reactively take action based on the
interoperation with the PPOOA-Visio CASE
tool;
 Can inform the software architect about the
current step of the running architecting
process or subprocess;
 Help issues regarding PPOOA are presented to
the software architect depending on the
current context;
 Provide configuration parameters to adapt the
agent to a particular software architect;
 The software development project information
is maintained independent.
3.2 Components of the PPOOA_ATA
Agent
The known agent metamodels (Henderson-Sellers
2005) and the integration of the agent with the
PPOOA-Visio tool are the main drivers of the
architecture solution selected for the PPOOA_ATA
Agent.
The agent is integrated in a Windows
environment using the Microsoft Foundation
Libraries to communicate with the software architect
through windows and dialogs.
The agent is implemented as an add-on that
communicates with PPOOA-Visio tool using the
COM interface provided by the Office suite.
The agent is coded in Visual C++ and it executes
in its own thread. The agent is autonomous but it can
open a Visio instance.
The main components of the agent are
represented in Figure 2 and described here:
 Agent controller managing the current situation
based on events and planning next steps;
 Knowledge DB keeping the knowledge
regarding the process and guidelines;
 Configuration DB keeping the configuration
parameters adapted to the user;
 Project DB keeping the information of a
particular software development project;
 Interface Visio Tool receiving events from
PPOOA-Visio tool and communicating with
it;
 Architect Interface is the man machine
interface.
These components have dependencies among them
as represented in the figure. The agent controller
centralizes the control over the rest of the
components and interoperates with the PPOOA-
Visio CASE tool and the PPOOA help files.
Figure 2: PPOOA_ATA Architecture.
A COMPUTERIZED TUTOR FOR ARCHITECTING SOFTWARE - Supporting the Creative Aspects of Software
Development
369

3.3 The Agent in Use
The PPOOA_ATA agent presents a main screen,
shown in Figure 3, which is part of the computerized
tutor and provides outputs relative to the architecting
process, next goals, help issues and methodology
guidelines to apply. This main screen acts as a real-
time library reference tuned to the exact step of the
architecting process at which the software architect
is working.
The main screen shown in Figure 3 describes the
current step, goals and substep of the architecting
process for a particular situation. In this situation the
computerized tutor recommends the software
architect to begin with step 3 of the PPOOA
architecting process. The first substep of the step 3 is
concerned with the identification of the real-time
system external events and their arrival patterns
(periodic, bounded, bursty, etc.). The buttons shown
in Figure 3 give the software architect the
opportunity to know the help issues and guidelines
applicable to this situation. These screens are not
presented here.
Figure 3: PPOOA_ATA Main screen.
4 CONCLUSIONS AND FUTURE
WORK
Our main goal was transform PPOOA-Visio
software architecting tool into a more user centered
CASE environment. The implementation of an agent
acting as a computerized tutor was one of the main
achievements.
We tested the agent with some examples of
architecture development and we believe that it
performs well regarding tutoring novice software
architects based on its interoperability with the
PPOOA-Visio CASE tool. We plan to extend the
agent and the CASE tool in the following directions:
 Extend the tool events received by the agent;
 Enhance model checking capabilities of the tool
and allow the agent to interpret model
checking information and take actions;
 The CASE tool should offer a wide range of
real-time software architecture patterns for
reuse.
REFERENCES
Bachmann, F., Bass, L., Klein, M., Shelton, C. 2004.
Experience Using an Expert System to Assist an
Architect in Designing for Modifiability. Fourth
Working IEEE/IFIP Conference on Software
Architecture. IEEE.
Diaz Pace, J.A., Campo, R.M. 2003. Design Boots:
Towards a Planning-based Approach for the
Exploration of Architecture Design Alternatives.
Argentine Symposium on Software Engineering.
SADIO.
Fernandez-Sanchez, J.L., Martínez-Charro, J.C. 2003.
Implementing a Real-Time Architecting Method in a
Commercial CASE Tool. 16
th
International
Conference on Software and Systems Engineering and
Their Applications. CNAM.
Franklin, S., Graesser, A., 1996. Is it an Agent or just a
Program? A Taxonomy for Autonomous Agents.
Third International Workshop on Agent Theories,
Architectures and Languages. Springer Verlag.
Henderson-Sellers, B., Quynh-Nhu, N.T., Debenham, J.
2005. An Etymological and Metamodel-Based
Evaluation of the Terms Goals and Tasks in Agent-
Oriented Methodologies. Journal of Object
Technology, vol 4 no 2, March-April 2005, pp 131-
150. JOT.
Racko, R. 2004. A Cool Tool Tool. Software Development
Magazine. May 2004. CMP Media LLC.
ICSOFT 2007 - International Conference on Software and Data Technologies
370