Simulating Safety Evacuation for Emergency Situation in a Building

Abstract

Crowd behavior simulation in a building or in-door space is an important goal when considering emergency situations. If a real threat like a fire breaks out, it can induce panic in the crowd resulting in much more dangerous situations like crushing or trampling. These actions can introduce large number of injuries and even for deaths. Proper planning is needed to save valuable lives and resources. Therefore the main goal of this research is the development of methods to simulate crowd behavior within in-door spaces during emergency situations. Placing the exit location of a building is considered as one of the most important decision when it comes to design a building. Computer based evacuation model can be used for testing safety of the building and to evaluate human behavior during emergency evacuation. This will, prospectively, help in training people to e ciently and e ectively act in emergency situations. Another important aspect of this research is considering disabled persons for emergency evacuation simulation. This is because very little work has been done on evacuation modeling techniques for disabled persons. Children are also neglected in the previous researches. This research will therefore attempt to modify the existing models to also accommodate the behaviors of disabled persons as well as behavior of people according to their age level. An evacuation simulation model which evaluates the optimal door location for a specified room in a building with emergency fire has been developed. A genetic algorithm is used to evolve the optimal door location in order to optimize overall evacuation time. A* algorithm is used for path panning of agents with some improvements. Performance evaluation is done to evaluate the implemented evacuation simulation model. NetBeans profiler is used for profiling the model for analyzing the performance and the memory usage. Performance of evacuation simulation model is also evaluated using machines with di erent configurations. It is evident that performance of the model is not dependent on configuration of the machine for small number of agents. Evacuation behavior of agents would not be very realistic by applying A* algorithm, because all the agents select the shortest path to the door. Therefore path planning of agents is done using A* algorithm with improvements. Optimal positioning of exits around a room is explored in this research using genetic algorithm. Among many other important results, the research shows that genetic algorithm seems quite suitable for discovering optimal door location for a given room in a building for evacuation planning.