Research On Related Technology Of Crowd Animation

Crowd animation was first investigated in 1980s when Reynolds introduced the boid system to model the motion of flocks of birds, schools of fishes, herds and land animals. Crowd animation gradually becomes a very important research topic in computer graphics and virtual reality. In recent years, crowd animation has been widely used in various fields, such as games, entertainment, public security, military exercise,urban planning and research simulation. Compared with the traditional computer animation, crowd animation can greatly reduce the workload of animators and create more spectacle and natural scenes.Character modeling, path planning and collision avoidance are the three key techniques of crowd animation. These techniques are mutual blend: in the process of crowd modeling,collision avoidance should be considered (such as adding a bounding box for individual); Collision avoidance problems can be regarded as the problem of local path planning in a certain extent; Individual should choose the route which is the shortest and non-congested (to avoid collision) when it plans path.The purpose of this paper is to solve the path planning, collision avoidance,intersection and shaking phenomenon in the period of crowd animation. The proposed algorithms aims at finding a balance point between the visual effect and algorithms efficiency, so that these algorithms can be applied to large and complex virtual scenes.This paper mainly includes three aspects:(1) Using DC model and density information, a crowd simulation algorithm is constructed, which discusses crowd simulation from two aspects: intra-group simulation and inter-group simulation. In intra-group simulation, the environment is divided into a series of walk-able areas. From the perspective of global planning,individuals select non-congested route and naturally smooth through the virtual environment while considering the real density information of walk-able areas, which can avoid the unnatural collision behaviors when they enter the congested route effectively. In inter-group simulation, a novel DC model is presented for path planning,which considers the global path and local information, and adopts the dynamic feedback mechanism to determine the targets of individuals. A rigorous dynamic feedback expression has been derived, thus in practical applications, impact factors will generated by the dynamic feedback mechanism for the different scenarios. As there may exist some inevitable intersection phenomenons in some virtual environments, a variability bounding box based solution was proposed. The rule of angle has been formulated by exploring the initial position information and the moving direction of the individual, then the hierarchical priority was determined, so as to solve this drawback.Compared with other similar algorithms, the experimental results show that the algorithm is more effective and robust.(2) A grid-based path planning algorithm of crowd is proposed. This algorithm realizes the dynamic path planning by propagating the change of distance information in the grid, and the priori information about crowd and obstacles will not be needed. A penalty function is used to achieve collision avoidance, which incorporates "safety margins" around obstacles using local information. This algorithm not only can improve the veracity of collision avoidance in large-scale crowd simulation effectively,but also can reduce computing cost as well. A mathematical model of crowd controlling is developed and the relevant experiment is carried out. First, the algorithm's high efficiency is proved in two-dimensional grid environment. Furthermore,the proposed algorithm has been expanded into three-dimensional terrin via the triangle subdivision mothed with height interpolation. The result confirms that the algorithm is still effective in three-dimensional terrain.(3) A crowd simulation algorithm is improved by using crowd flow information.In multi-layered three-dimensional environment,a navigation mesh is built for global path planning and uses crowd flow information to solve the collision avoidance problem among individuals. This algorithm employs a hierarchy priority based collision avoidance method to improve the precision of collision avoidance, and a"stopping rule" is proposed for individuals at original place to avoid the "shaking" and unnatural behaviors, which can effectively improve the reality of the crowd motion.The research of crowd dynamic theory indicates that the crowd flow-based algorithm can retain the substantive characteristics of human crowd and be fit for large-scale crowd simulation. Experimental results show that the algorithm can avoid individual generates unnatural behaviors, and effectively improve the reality of crowd animation.These solutions are with their own advantages and uniqueness. Different solutions are suitable for different types of virtual scenes. The first method is to solve the collision avoidance start with the individual, then works on the group to plan the optimal path,finally improves the visual effect of simulation process by exploiting the combination of local and global. The second method can realize improve-efficiency crowd animation simulation without the priori information on the crowd and obstacles. Based on the flow information,the third method focuses on the whole crowd,and can simulate large-scale crowd motion.