Evacuation time estimate for a total pedestrian evacuation using queuing network model and volunteered geographic information

TL;DR

This paper develops a mean field model linking city road network features and population data to estimate evacuation times, aiding emergency planning through a novel queuing network simulation approach.

Contribution

It introduces a new mean field framework that relates catchment area attributes to evacuation time estimates using agent-based queuing network simulations.

Findings

90% of agents are within 5.4 km of their exit

Established a relationship between catchment attributes and evacuation time

Simulated evacuation scenarios to validate the model

Abstract

Estimating city evacuation time is a non-trivial problem due to the interaction between thousands of individual agents, giving rise to various collective phenomena, such as bottleneck formation, intermittent flow and stop-and-go waves. We present a mean field approach to draw relationships between road network spatial attributes, number of evacuees and resultant evacuation time estimate (ETE). We divide $50$ medium sized UK cities into a total of $697$ catchment areas which we define as an area where all agents share the same nearest exit node. In these catchment areas, 90% of agents are within $5.4$ km of their designated exit node. We establish a characteristic flow rate from catchment area attributes (population, distance to exit node and exit node width) and a mean flow rate in free-flow regime by simulating total evacuations using an agent based `queuing network' model. We use these variables to determine a relationship between catchment area attributes and resultant ETE. This relationship could enable emergency planners to make rapid appraisal of evacuation strategies and help support decisions in the run up to a crisis.