Agent-Based Evacuation Modeling in the Context of Non-Recurring Events

Evacuation is an important measure in emergency planning, and agent-based simulation model is an effective tool to test the reasonableness of the evacuation plans. This study simulates the evacuation behavior of New York City residents based on real data from Hurricane Sandy by using the mesoscopic multi-intelligence simulation tool MATSim. It aims to estimate road capacity loss. In this study, negative binomial and log-normal distributions are applied to model accident frequency and duration. Subsequently, observations of non-recurring events are generated from the above distributions by Monte Carlo simulations, and the loss of capacity caused by non-recurring events is modeled and dynamically incorporated into the simulation network according to the observations. In addition, the NYBPM V2.1 model is also used to calibrate the road network capacity and free speed of MATSim resulting in a more realistic and reliable simulation model. The results show that the data obtained from the calibrated road network model are more consistent with Manhattan taxi travel times.