International Workshop on Massively Multi-Agent Systems (2004.12. 10-11)

A Massively Multiagent System for Disaster Mitigation

Ikuo Takeuchi

Computer Science Department,　The University of Electro-Communications

One of the most outstanding features of the Integrated Disaster mitigation Simulation System (IDSS) now under the development　in the DaiDaiToku project funded by the Japanese government is that　this disaster simulation system incorporates massively multiagent simulation essentially unlike other systems that merely estimate and predict physical aspects of a disaster. At a disaster, say, a big earthquake, human individual and social activities are important to estimate and mitigate the disaster damage effectively as well as complicated physical phenomena. Especially, an earthquake disaster in urban area involves a huge number of different kinds of people, suffering civilians, volunteers, disaster responsive professionals, local (and central) government officials and so on.

In the talk, we will discuss the following issues:

(1) How did we design the IDSS architecture so that it can integrate many different kinds of simulators and a massively multiagent system on top of them? As can be easily seen, the whole system will be so large that it should be implemented in a large-scale distributed computation framework. It raises some technical problems by various reasons, for example, because agents can interact with others non-locally, say, be telecommunications.

(2) How can we exploit a massively multiagent simulation to mitigate a disaster in reality? We experienced that a simple simulation like RoboCupRescue gave a deep impression to many citizens in the simulated area; therefore a good multiagent simulation is very useful to enlighten civilians on how to act at a big disaster. This will be readily able to be extended to professional training. But it is a difficult problem to make the simulation system useful just after a big disaster happens. That is, it is a big challenge to associate virtual massively multiagent system in the simulation with a huge number of real humans in the disaster field in order to make the real-time simulation fully exploitable to mitigate the disaster.