The Study On The Data-driven Simulation Modeling Of Pedestrian And Evacuation Dynamics

With the rapid development of urbanization in China,people in public places often suffer from instability accidents.In the era of big data,the acquisition of pedestrian movement and crowd evacuation data is unprecedented convenient.How to use these data to enhance the effectiveness of modeling and simulation technology in the whole process of crowd emergency management in public places faces great opportunities and challenges.In order to cope with these opportunities and challenges,this paper proposes a data-driven approach to study the modeling and simulation of pedestrian movement and crowd evacuation.Firstly,the application framework of pedestrian movement and crowd evacuation modeling based on parallel simulation is established,among which data-driven simulation modeling is the key technology to realize the framework.Then the generalized model of pedestrian movement and crowd evacuation is studied from two aspects of model structure space and parameter space.Finally,a more reasonable risk assessment method is proposed by integrating the above research results with the risk assessment of the evacuation of people affected by the diffusion of toxic gas.The main results are as follows:(1)The application framework of pedestrian movement and crowd evacuation modeling simulation was established based on the parallel simulation method,and three components of data-driven simulation modeling were proposed from the perspective of statistical learning.The application framework of pedestrian movement and crowd evacuation simulation based on parallel simulation includes five parts: emergency evacuation crowd,crowd simulation model,data observation,data-driven modeling and decision support based on simulation.This framework innovates the application form of modeling and simulation technology in emergency evacuation research and helps to form a feedback loop from emergency evacuation scenario to simulation model and then to auxiliary decision-making.Then,three elements of data-driven simulation modeling--model,strategy and algorithm--are proposed from the perspective of statistical learning,which lays a theoretical foundation for the research of data-driven simulation modeling.Secondly,focusing on the relationship and difference between the statistical learning model and the simulation model,this paper discusses the problems existing in the three elements of data-driven simulation modeling from the four aspects of model decoupling learning.Finally,the model,strategy and algorithm of data-driven pedestrian movement and crowd evacuation simulation modeling are given.(2)A model space correction and optimization method based on the optimal performance of the simulation model is proposed,and the optimal setting method of the social force model and the expected rate sub-model based on the differential equation model is given.In this paper,a model comparison framework based on the optimal representation of the model is proposed to modify and optimize the structure space of the model so as to improve the adaptability of the model to the actual application scenarios.The model comparison framework includes four parts: simulation model,reference data,model evaluation index and optimization algorithm.Applying this framework to social force and the pedestrian movement and the crowd evacuation model based on differential equation of the expected rate of sub models of optimization process,the simulation model based on literature review and empirical data analysis summary of nine expected rate sub models,reference data from controlled laboratory experiment,the model of evaluation index based on population density,the optimization algorithm using differential evolution algorithm.Through the simulation experiment,the selection method of the optimal expected rate submodel is given,and the conclusion with practical guiding significance is obtained.(3)A robust crowd evacuation path selection model based on potential energy field is established.This paper analyzes two problems in the robustness of the current crowd evacuation path selection model based on potential energy field: the algorithm is biased against the gray area in the environment layout and the algorithm parameter allocation interval is sensitive to the change of the environment size,and proposes a method to improve the robustness of the model from the perspective of modifying the definition of path capacity.The simulation results show that the proposed definition of route correlation capacity and euler average route correlation capacity can improve the robustness of the model.The experiment also proves that the modified crowd routing model can still reflect the real crowd routing behavior.The above work verifies the effectiveness of the modification of model structure space and parameter space on the improvement of model robustness.(4)A more reasonable method for assessing the risk of evacuation caused by gas diffusion is proposed.Through field investigation,relevant data of railway station square are collected,and a more accurate environment model of railway station square is established.The interaction between crowd and gas diffusion is modeled with multi-agent modeling method,in which the decision-making process of personnel is in line with the BDI framework,the movement models of personnel action layer and tactical layer adopt the social force model of recommended expected rate configuration and the path selection model based on potential energy field proposed in this paper.Simulation results show that the proposed method is significantly improved compared with the original method,and it complements the simulation-based method of crowd evacuation risk assessment under the influence of toxic gas.In conclusion,from the perspective of using big data to improve the efficiency of modeling and simulation in practical application,this paper studies the basic theory and key technologies of data-driven simulation modeling,and solves the problem of unclear theoretical framework of data-driven simulation modeling.This paper applies the data-driven simulation modeling theory to the simulation of pedestrian movement and crowd evacuation,proposes a method to improve the universality of the model from the perspective of model structure space and parameter space,and adopts the data-driven simulation modeling method to improve the risk assessment method of crowd evacuation affected by gas diffusion.The research results in this paper have important theoretical significance and application demonstration value for modeling simulation and research on pedestrian movement and crowd evacuation.