Research On Data-driven Crowd Behaviors Modeling And Simulation

Crowd behavior,a key research topic,is a kind of complex social phenomenon,which has drawn increasing attention from sociologists,physicists,computer scientists and government on its effective prediction and management in the complex scenarios or events.For the efficiency,computer modeling and simulation technology has been widely used by researchers in the research field recently.The research goal of the modeling and simulation technology is to show the appearance of the crowd,pedestrian behaviors and motion actions vividly in the virtual scenes and make it possible for researchers to study the complex characteristic of the crowd and the diverse motion features in the given environments with multiple entry.It can be applied as a tool for simulation and analysis.In this paper,we present a generic data-driven approach to generate crowd behaviors that can match the video data based on the work of others.The proposed approach is a bi-layer model to simulate crowd behaviors in pedestrian traffic in terms of exclusion statistics,parallel dynamics and social psychology.The proposed approach includes two layers,the top layer and the bottom layer.The top layer models the macroscopic pedestrian behaviors,such as how pedestrians enter the scene and make their goal selection,and the perception behaviors in the given environment and the corresponding decision-making and routing process.In this layer,the fuzzy c-means algorithm(FCM)is used to obtain the distribution rules of the crowd in and out of the particular scene according to the effective pedestrian trajectories extracted from the video data and the goal selection behavior is modeled based on that obtained distribution rules.The perception behaviors and the corresponding decision-making and routing process can be also simulated by using the proposed data driven pedestrian motion model.The bottom layer focuses on the microscopic collision avoidance behaviors based on the reciprocal velocity obstacles model(RVO).To validate its effectiveness,the approach is applied to generate collective behaviors and re-create scenarios in the Informatics Forum,the main building of the School of Informatics at the University of Edinburgh.The simulation results demonstrate that the proposed approach is able to generate desirable crowd behaviors and offer promising prediction performance,which is of research value.