Real Time Large Scale Crowd Simulation Based On Fluid Framework

Crowd Simulation is a kind of technology that simulates the characteristics and behaviors of the real world crowd movement based on computer graphics.By establishing a virtual world,we can visually and intuitively simulate the crowd movement.And accompanying with the development of the computer graphics in the past few years,the use of crowd simulation is becoming increasingly wider and deeper.We can easily find the application of crowd simulation in many fields such as games,movies,animation,virtual reality and so on.In recent years,although the development of the urban brings much convenience to the residents,the boom growth of population,heavy traffic and the complexity of the terrain all bring a lot of dangers.Especially for those places with a large scale of crowd such as subway stations,some unexpected events including fire will definitely lead to congestion.In some extreme cases,it will lead to a big tragedy that can be a great threat to life and security.As a result,it is of highly essential to establish a realistic model for large-scale crowd simulation and provide us with a real-time detection method for collision avoidance assessment to prevent the occurrence of such incidents.Agent-based method is one of the most mature approaches in this field.It defines each agent as individual and simulates the movement of the entire crowd by calculating the interaction force among each other.Despite the fact that the differences among each individual are taken into consideration,lack of the reaction to the external environment will result in a lot of unrealistic phenomenon such as gathering.Meanwhile,it involves too much calculation during simulation.Therefore,it doesn't perform well for large-scale simulation in real time.To solve this problem,this paper designs a real time simulation framework for large scale of crowd based on continuum crowds.This framework integrates the advantages of the continuum model and agent-based model.Both reaction to the external environment and personal decision are preserved in the simulation.Meanwhile,we add some flock-based features into our framework to make the simulation of small groups or movement joint more realistic.Also,we provide a virtual node algorithm based on KNN method to accelerate the speed performance.We evaluate our framework through a series of experiments and results strongly prove that our framework works well in real-time scenes.