A Study Of Particle-Based Approach For Crowd Simulation

Crowd simulating is widely applied in animation, movies, games, and other industries, such as simulating birds migration, the movement of fish and the crowd, and even the racing, and so on. Recently, in other areas such as training simulation, public safety control, artificial life, the virtual life crowd simulating also has more applications. In the most typical scenarios, crowd-simulating requires crowd to move from point A to point B, therefore crowd simulating can also be interpreted as simulating movement of the crowd.At present, crowd simulating is based on three ways: 1) Kinematics, that is, "key frame" method, which requires much work to create the movement path of crowd, including planning the path of each role one by one, preventing the members collision in each frame and bearing much cost due to the changes before. 2) Crowd AI, which is a realistic but costly approach. 3) Particle simulating, in which the movement of crowd is achieved through controlling the movement of particle streaming. In this way the cost is very low, because a lot of 3D software have particle system package. However, in particle system, the particle is driven by force field, and the function of interaction between the particles is not provided, therefore particle system is unable to achieve the realistic crowd scene.According to the shortcomings in existing technologies, the thesis pays attention to the perception system and the autonomous behavior model of autonomous agent. Main contributions are as follows:First, an improvement is presented by combining particle system with the individual-based distributed model - boids mode and a virtual agent model is created in the scene of maya.Second, the virtual perception is obtained by the method of defining the radius and angle of the sense , which takes each particle as an agent in local coordinate system and applies the corresponding regional rules within the particle-centered region of perception.Third, interaction between particles in regional space is simulated by applying force field to agent to acquire the change of velocity, in which basic behaviors of agents are designed and arranged into hierarchy according to behavioral hierarchy theory. Therefore high-level tasks are achieved through the combination of the basic behaviors.Fourth, the algorithms are implemented through a series of maya nodes and MEL program, in which the natural status of flocking and automatically avoiding collision with obstacles are presented.Practice has proved the approach in this paper can achieve a series of scenes including seeking, fleeing, flocking and obstacle avoidance, at the same time, possesses high efficiency of the rendering.