Efficient Information Assignment for Crowd Evacuation using Black Box Optimization

Abstract

Reduction of evacuation time and selection of correct exits are two crucial factors in crowd evacuations. It would be possible to increase the survival ratio of evacuees by controlling the crowd by properly providing evacuation information. However, this is challenging because of the complexity in crowd behaviors caused by cognitive biases in evacuees, such as herd behavior. In this paper, we present and formulate a crowd evacuation problem in which agents attempt to flee from a simple rectangular environment with four exits such that two of them are correct, but others are not. Thus, agents face two decision problems: selecting correct exits and rapid evacuations. We tried to maximize the survival ratio of agents by arranging the locations of evacuation information in the environment using Black Box optimization techniques and multiagent simulations. GA and SA have shown the ability to explore complex search space and discover good solutions.