Multi-agent simulation of civilian's evacuation

Civilian's evacuation has been simulated with a multi-agent system to propose an evacuation plan in a megalopolis. In the simulation, floating-point number coordinates are employed to minimize calculation errors. A megalopolis is comparted into small areas to be distributively computed by numerous computers. Behavior of human in evacuation is characterized as "an agent", which is a robot program that acts autonomously. A global city structure is modeled as topography data of a "map". The topography data, which consist of "nodes" and "edges", are input into virtual space on a computer in real number coordinates. Nodes and edges construct routes for escape, which correspond to a crossing of a road and to a straight-line road between nodes, respectively. The agents evacuate with circumventing obstacles and another agent in human-like behavior on the designed simulation primitive, which is a unit of distributed processing. The multi-Agent simulation system might be effective to simulate civilian's evacuation in a megalopolis.