Monday May 3, 2004
9h30 - 10h30
Methodological
impacts of User Interaction in Agent-Based Simulation
Yves
Demazeau
The interaction between an agent-based simulation system
and a user is usually considered as being secondary behind the design and
development of the agent system itself. We believe that this is largely a
mistake and that the interaction with a multi-agent system has to be taken
into account from the very beginning of the analysis process itself. We will
first discuss the three types of usage one may find within multi-agent
systems, from the most immersive ones where the user is more an actor to the
most indirect ones where the user is more an observer. In every case we will
study the particular expectations of the user and see which characteristics
this implies for the design and development of the system. We will also
explore how this also affects the evaluation of the system itself. We will
illustrate the presentation with examples coming from the domain of
interactive games which hopefully covers the three categories of usage one
may find in multi-agent systems. As a conclusion, we will show that giving
back to the user the more important place she should have, we hope to
generate more useful agent-based simulation systems.
|
Wednesday May 5, 2004
9h30 - 10h30
Simulation
as Part of Agent-Oriented Software Engineering
Adelinde
Uhrmacher
The
more complex software systems become that are intended to work
autonomously in open dynamic environments, the more important a systematic
experimentation will be. However, for modeling and simulation to play an
as important and accepted role in designing software agents as they do
today in other areas, e.g. designing manufacturing systems, the testing of
agents has to be comfortable, efficient, and effective, and to be firmly
rooted within the agent-oriented software engineering process. Challenges
have to be faced at the level of theoretical foundations of modeling and
simulation, modeling and simulation methodologies, and implementations,
equally. The software to be tested, the objective of the simulation study,
and the stage of the agent software development influence the
environmental model used for testing and the mechanism that synchronize
the execution of agents and simulation. A clear distinction between model
and simulation layer, and a modular component-based design of the
simulation system support the required flexibility. It will be explored,
how suitable mechanisms for execution might be chosen on demand in setting
up simulation experiments.
|