Simulating crowd movement in agent-based model of large-scale flood

Abstract

Crowd movement during natural disasters and major accidents can affect the success of evacuation procedure. Thus to develop an effective evacuation plan, a wide-range of scenarios causing different routes and patterns of crowd movement should be considered. Recent researches in agent-based simulation have achieved techniques to simulate crowd movement in emergency scenarios at city scale. However simulating crowd movement at a macroscopic level for disasters which might affect several cities, like large-scale flood, is still a challenge. This paper addresses this issue and makes three contributions. First, the development of an agent-based layered model to simulate large-scale flood using GIS is demonstrated. Second, the simulation of crowd agents' movement on available roads is presented. Third, the preliminary experiments running on private Cloud server is reported. The experiments cover case studies illustrated the movement of crowd agents around Thailand while several parts of the country were inundated. The 2D animation depict the movement of crowd; and the simulation results show the status of the agents and the amount of individuals which required shelters. To simulate one day events, the simulator took 4–9 hours execution time depending on the severity of floods and available facilities.