Map Segmentation in RoboCupRescue Using Louvain Method and Its Application to Agent Control

Abstract

In many parts of the world, large-scale disasters such as earthquakes, typhoons, tsunamis, and volcanic eruptions are frequently occurring. Responding to such disasters is one of the most critical social issues, and there is much research on applying artiflcial intelligence to disaster prevention and life-saving systems. One of them is RoboCupRescue Simulation (RCRS), which is a multi-agent system to simulate the behavior of rescue units (called agents) and citizens in a virtual city in the event of an earthquake. In the uncertain and ever-changing situation of disaster relief, it is essential for each agent to sift through the necessary information and take appropriate action. In this paper, we focus on the agent’s activity environment that is said to influence the agent’s ability to perform, and we use the Louvain method to partition the map of RCRS into regions. In addition, we examine the optimal behavior of agents using map features and information.