Experiment And Simulation Of Individual And Crowd Evacuation Behaviors

Since life safety closely relies on an efficient evacuation under fire condition or stampede, to implement the research of human behavior and evacuation dynamics under emergency conditions becomes very meaningful for our public safety. But due to the dangerousness and unrepeatable property of these catastrophic disasters, there are very limited data can be retrieved from real calamitous events. The most commonly used methods for the research of evacuation dynamics should be evacuation drills, lab-scale evacuation experiments and the simulation of evacuation process and human behavior using computer technologies.In this thesis, a group of evacuation experiments under the guidance of alarm signal have been carried out in a classroom with ordered desks. During part of the experiments, the participants were asked to wear opaque masks to mimic the condition of restricted visibility because of fire smoke or power failure and this was compared with the normal condition. By investigating and analyzing the exit outflow, the individual evacuation time and the route choice behavior etc., we studied the influence of the alarm signal, the sight condition of participant and the exit width on the evacuation process and human behaviors. Several special phenomena, namely"speed varying","gender discrepancy","touch-and-attach behavior", and"Subconscious environmental information perceiving behavior and its attenuation", were observed in the video record. Corresponding simulation research has also been conducted based on these experimental results.Through the analysis of the participants'exit choice behaviors, we found that except the distance from the exits, the density of people near exits may also accounts for which exit people will choose. It seems that people are prone to choose the less crowded exit, although they may be a little further. A Cellular Automaton model was revised by considering the exit area population density to simulate evacuees'perceiving and judging of the population density of the exit area. The simulation and experiment results have shown that the exit area population density cannot be ignored and the simulation result can be more realistic by considering this parameter.The phenomenon of"subconscious environmental information perceiving behavior and its attenuation"has also been observed in the experiment. People can build a cognitive virtual map of the environment in their brain when they are familiar with it and use the"map"to look for a reasonable escape route or choose the right exit when they cannot see. But this subconscious may be attenuated with the time elapsed. According to this phenomenon, a simulation model considering this behavior was developed based on the detailed comparisons of several basic rules used in the"Floor Field"Cellular Automaton model. The simulation result under the same condition of the experiment showed that although the influence of the sound information in the"blind"environment was more significant than that in the normal condition, the sound was not the only decisive factor, because the environmental information was not totally unknown in the"blind"condition due to the subconscious information perceiving behavior.At last, we studied the gender discrepancies and the information communication behaviors in evacuation. We found that the average speed of male participants was 0.2~03m/s larger than that of the female participants and with the same experimental setup, participants with different genders at the same location need different time to escape, usually the escape time of a female participant is much longer. Then, a simulation model considering the gender discrepancies was build based on CA model to study the evacuation process. In addition, referring to the information communication behavior observed in the experiment, we used a"multi-information-map"CA model to simulate the information communication between people in complex evacuation scenarios such as"building with blocked exit"and"building with secret exit". The simulation results indicated that the model is capable of modeling the heterogeneous crowd behavior due to information transmission.