This paper proposes an agent-based simulation model to predict the crowd behaviors and evaluate the evacuation safety on passenger ships. The model focuses on the behaviors of two common types of passengers that are typically ignored during most evacuation analyses, namely, passengers who are not familiar with the ship layout and passengers who have family members or friends with them. In the proposed model, a marker concept is introduced to represent critical routing points of layout and passenger agents make a route choice based on their surroundings and characteristics instead of just following the shortest routes. The simulation model is tested by two small but targeted scenarios and one comprehensive scenario on a ship deck. For ship designers, a more realistic evacuation time is provided to better assess the evacuation performance of a ship, and a heat map of crowd density is presented to identify possible bottleneck areas.

This paper introduces a novel technique to design the level of service (LOS) for facilities or sub spaces of buildings for the purpose of evacuation planning. LOS is a standard qualitative indicator used to describe flow characteristics in a pedestrian environment. Some evacuation planners use LOS to help determine the network parameters when solving evacuation planning problems by the network flow approach. However, there is currently limited research into the optimization of the LOS parameters themselves to construct more efficient evacuation networks. In this paper the authors used a genetic algorithm optimization approach to determine LOS for facilities to improve the evacuation performance of building networks. Each individual chromosome containing a LOS design represents a fully defined evacuation network that can be solved. The fitness of each network is measured by minimum clearance time, which is calculated by the Capacity Constrained Route Planner (CCRP) approach. A comparative computational test in a hypothetical three-story building shows that the evacuation network under the optimized LOS design has a roughly 11% less minimum clearance time compared to the network under the original LOS design. Sensitivity analysis is also included, focusing on how the population size and the building layout influence the LOS design. In addition, an additional computational test for a twelve-deck cruise ship shows that the approach is scalable to solve more complex evacuation networks. The proposed approach has the potential to provide better LOS assignments for facilities for the government officials to develop effective emergency management strategies.