Project : alcove
Section: Scientific Foundations
Autonomous physical simulation
In order to provide an efficient software framework, it is important for further global developments, that tools would be provided so that most recent software techniques, such as agent-based ones, would be usable, which is so far not the case for physical simulation. Some tests have been done this year about this, and first results have been proposed on physical simulation within agent-based software structure: the "one agent per object" principle has been fully applied, letting as much freedom as possible to each object, and the first result we get seems to be convincing.
Jérémie DEQUIDT, PhD student, aims at building a simulation framework which does not impose a shared discrete timeline for the simulated objects . The framework may be seen as a multi-agent system where each autonomous agent is a simulated object that can maintain its simulation state and maintain its behavior by checking its environment. Decision schemes are provided to agents allowing them to control their computation time, in order to maintain global synchronization and simulation realism. This thesis also focuses on collision detection frameworks (usually centralized processes) to handle collision detection for autonomous simulated objects.