Simulation Research On Self-organization Behavior In Crowd Evacuation

In recent years,with the rapid development of economy and society,large crowd can frequently be seen in public places.In order to prevent people from crowd congestion and protect people's life and property,simulation methods are used to simulate the behavior of crowd evacuation,which provides a good solution for the study of pedestrian evacuation in emergency situations.In the studies of crowd behavior,the group behavior of the crowd is the significant issue that can not be ignored.In the process of real-life evacuation,not all pedestrians exactly know the export information in the scene.Thus they can not fully master the evacuation path,as a result,the behavior of the individual pedestrians can be easily affected by external behavior during the evacuation process,their self-judgment and behavior are consistent with the public driven by the herd mentality,only a few people keep moving independence.The group-psychology results in the group self-organization phenomenon.At the same time,the research shows that pedestrians can not freely move as they want due to road congestion,they prefer to choose smart group motion so that achieve a variety of collective behaviors.Therefore in the crowd evacuation simulation,not only the individual's autonomous behavior should be considered,we also need to understand self-organization behavior characteristic and mechanism driven by herd mentality,which can provide some theoretical support for simulation of crowd evacuation behavior.Most of the existing researches simulated crowd aggregation phenomenon by adding attraction into dynamic model,which results in large computation and the slow aggregation process.At the same time,the group leader and the rest of the group members were treated equally in the group.They failed to show their particularity during their movement process,and the role of the leader was not maximized.To solve the above problems,firstly,the paper studies the real motion video of pedestrian evacuation's cluster process,and the results show that there are leaders in self-organization groups.Secondly,based on social force model,the paper presents a two-layer social force model.At the same time,the two-layer social force model also provides an opportunity for the combination of evacuation simulation and large data analysis.In this paper,an instance-based learning method is used to extract the existing evacuation parameters and find the path that can not be fully used during the evacuation process.Finally,a complete self-organizing group simulation platform is constructed,and the model is fully verified by comparing with the self-organizing behavior of the crowd in the evacuation process of real earthquake evacuation.The main work and innovation are summarized as follows:(1)In view of crowd movement process,on the basis of social force model,the paper proposes a two-layer social force model.This model takes leaders as a link between exports and group members,and achieves self-organizing group behaviors such as pedestrian's fast gathering and local tight evacuation.At the same time,considering the influence of the disorganized pedestrian's herd psychology,the process of a disorganized pedestrians' dynamically gatheringinto groups is realized.At the same time,in the multi-room scene,a temporary leader is added into the improved model,so that the model can solve the simulation in a more complex scenarios.(2)A method for the dynamic gathering of disorganized pedestrians is proposed.Through the analysis of the real evacuation process,a group partition method considering disorganized pedestrians is proposed.At the same time,taking full consideration of the factors such as the crowd of discrete pedestrians during the evacuation process,a method of disorganized pedestrians' dynamic gathering is proposed,which simulates the effect of group aggregation and following process of disorganized pedestrian dynamics(3)Two-layer social force model provides the possibility for the combination of social force model and big data analysis.The combination of instance-based learning and two-layer social force model,reduces the possibility of the retention phenomenon due to the line of sight obstruction,improves pedestrians flow rate.Finally,on the basis of the above work,the simulation system is further perfected,and several simulation experiments are carried out in the paper.The simulation results show that the model can truly reproduce the process of aggregation,following and extinction of self organized groups.At the same time compared to the true characteristics of sports population,self organization behavior to validate this method proposed can realistically simulate the real people,for the analysis of crowd behavior characteristics for different scenarios of future and provide efficient evacuation scheme has high reference value and application value.