A data-driven, scenario-based human evacuation model for passenger ships addressing hybrid uncertainty

Abstract

In a disaster at sea, the safe and timely removal of passengers from the ship is paramount. Here, a human evacuation plan enables passengers to be swiftly displaced from a risk area to one less so. Despite existing research on evacuation planning, there is a need for a more comprehensive model that considers various uncertainties and factors. This paper proposes an optimization evacuation model that balances uncertain variables, including passenger walking speed and travel distance, and deterministic factors like deck layout, door capacity, initial density, and corridor traffic flow. The model also accounts for varying starting locations and two levels of awareness—alert and non-alert. The model utilizes a data-driven technique, i.e., the k-means algorithm, to cluster historical data on speeds and generate scenarios. An adjustment scheme is applied to account for the ship's rolling motion, affecting passenger speeds during the evacuation planning period. Travel distance scenarios are produced to capture the impact of different route choices per passenger. A risk-neutral two-stage programming model is constructed to handle uncertainties. The model is tested on multiple problems for a passenger ship deck in day and night modes, revealing valuable managerial insights for the maritime safety sector.