SCSC2003 Abstract S91361

Multi-agent system design - The ANTIC training simulation project

Multi-agent system design - The ANTIC training simulation project

Submitting Author: Mr. Samuel Morisset

Multi agent system (MAS) methodologies (GAIA, Cassiopeia, etc…) take as a starting point a list of goals, functionalities, behaviours, entities, structures related to the final application goals. We propose an approach where modelling those prerequisites is integrated into the methodological process. For this approach our needs are to conceive a model of the contextual domain, the objectives and the dynamics of a training session, and to develop a methodology to extract a multi-agent architecture.

As training simulation is one of our main research axes, we have two goals. One is to introduce a scenario language allowing the description of events and links between them. Second is to build user interface mechanisms between a user and an agent of the MAS.

This approach responds to the needs expressed by firemen officers in the Antic 2 project. The problematic of the antic 2 project was to conceive a simulator functioning like a role playing game, based on their training methods. This project would have the following characteristics:
– Possibility to induce predefined events into the simulation (play scenario).
– Possibility to add new scenarios to the simulator.
– Possibility to configure a training session.
– Management Multimedia briefing and debriefing.
– Future integration of new actors and actions (next projects).

This project requires high level of specification and representation, and interoperability between the MAS and its users. It is a perfect case study to experiment the methods of representation by ontology, of scenario events specification, and of specification of agents and their actions (MAS).

Back to SCSC2003 Abstracts