Simulation Of Crowd Evacuation In Emergency Scenario Based On Event Evolutionary Graph

In recent years,with the increasing travel need of pedestrian,the volume of travel for pedestrian in various transport facilities and public places has been increasing,then dense population can be seen everywhere.Due to the unpredictability of emergencies and the various individual behaviors randomly in crowds,it often brings about more serious consequences,when emergencies occur in crowds.It is very significant that we should prevent the risks beforehand and organize the emergency evacuation afterwards for the safety of the public’s individuals and property.With understanding and master the characteristics of pedestrian evacuation,the evolution of the event state under unexpected events is the basis of relevant research.Simulation-based studies have been conducted,which are mostly based on experience and lack theoretical support in terms of element selection and parameter setting.The aim of the thesis is to use natural language processing techniques which can construct an event-based pedestrian evacuation mapping with a hierarchy of events and logical evolution patterns for knowledge inference and simulation analysis,Based on that,we propose an method for identifying and characterising the key factors Which influence pedestrian evacuation and a method for improving the simulation model.Firstly,a self-built corpus is constructed by using social network resources,media resources and journal database resources.For the event relationship inventory in unstructured texts,explicit event relationships are extracted by using a syntactic dependency-based method,as well as implicit event relationships are extracted by using the Bert+Bi-LSTM+Attention+Softmax to construct a deontological map of downstream human evacuation in emergencies.Secondly,the nodes on the whole matter mapping are classified by proposing a method for calculating the importance of nodes based on event relationships and event consequences.Then identifying four categories of key influencing factors: level of panic,level of environmental familiarity,level of conflict and level of hesitation.The superordinate nodes affecting these nodes are identified through matter mapping,fuzzy inference rules are formulated and quantitative characterisation of the influencing factors are achieved by means of fuzzy inference.Finally,a simulation model of pedestrian evacuation based on cellular automata is developed by considering the influence of key factors.A sensitivity analysis of the impact of each factor on evacuation efficiency and safety is also carried out in conjunction with case studies.Decision support is provided for the organization of pedestrian flow evacuation and risk assessment,etc.