Time to Leave the Louvre: A Computational Network Analysis

摘要

Increase in terror attacks has raised demand for safe emergency evacuation plans worldwide. We focus on evacuating the Louvre, the world's largest art museum. Evacuation is made difficult by the volume and variety of visitors; the Louvre management desires evacuation plans over a broad set of considerations. We partition the Louvre into sections and build an agent-based model to simulate evacuations in each section. We run simulations over each section to determine a rate by which agents exited. To connect sections, we represent the building as a graph, thereby posing a network flow problem. The strong duality property identifies bottleneck edges in the graph. Simulating blocked passages or new secret exits is simply removal or addition of edges to the graph. Bottleneck identification is our highest priority for public safety. Our model predicts that an evacuation plan using all four public exits could evacuate the Louvre in 24 minutes. Furthermore, while many bottlenecks surround the Pyramid entrance, the entrance itself is not a bottleneck. This property of the Pyramid is crucial in emergencies, since it allows access for emergency personnel. Additionally, securing the Passage Richelieu is critical to evacuation, since its safety is directly linked to the Pyramid's safety. Keeping these entrances open and useful is imperative to both speed and safety of an evacuation. Our model is powerful due to its ability to model individual human behavior, followed by an adaptable abstraction of building-flow dynamics. One weakness of our model is that we consider worst-case scenarios, but the evacuation times are an upper bound for a real evacuation.