大规模群体运动的实时模拟研究与实现

【作者】 周珊珊；

【导师】 张建中；

【作者基本信息】 电子科技大学 ， 计算机系统结构， 2010， 硕士

【摘要】 鸟群、鱼群和人群等集群运动是自然界最美丽也最常见的景象之一。但是这些复杂的集群运动在计算机动画领域却不多见。近几十年来,由于影视动画、虚拟现实和计算机游戏等领域的不断发展,群体动画也成为人们研究的热点方向。尤其是对大量人群聚集的实时交互性行为的研究,在虚拟训练、安全预演、数字游戏等领域都得到了广泛的应用。它通过探索真实世界中群体行为的物理本质和生物学、社会学特性,利用计算机为生物体建立行为模型,从而更加逼真的表现生物体的群体行为。与典型的计算机动画建模只关注角色外形和物理性质相比,基于智能体角色的行为动画专注于角色行为的建模,处理角色的行为和运动细节。这些行为包括了从低级的路径规划到高级的情感交互等行为。但是由于群体中个体行为的复杂性和多样性,大规模群体中对每个个体的行为控制带来的计算量巨大,难以满足实时模拟的应用需求。本文以万人级大规模群体为研究对象,试图在群体行为模拟的真实感和效率之间寻求一个良好的平衡点。本文展示了具有代表性的群体行为模型,包括粒子系统、集群系统、行为系统、混合系统、混乱模型等行为模型和认知模型;介绍了角色动画系统的架构,并说明了群体动画子系统的框架以及在整个系统中所处的地位;提出了群体动画建模方法,并在此基础上实现了万人级群体运动实时模拟,展现了真实感的群体行为,包括路径规划、碰撞避免、运动跟随等。一课题的研究着重集中在以下两个方面:第一,对群体动画模拟方法作出了全面的研究和透彻的分析,在此基础上提出了一种基于行为模型的大规模群体实时模拟框架。这种框架具有分层结构,信息在层级之间共享具有较高的效率。并且其固有的有限逻辑顺序,大大减少了行为之间的矛盾。第二,针对万人级群体动画的核心问题,提出了一种基于动态场景的局部路径规划改进算法。这种改进的享有全局知识的动态局部路径规划算法,既能够提供类似全局规划的最优路径,又具有基于局部知识动态更新的高效性,且统一规划,免去为每个个体单独计算的开销,有效地减少了计算量,提高了计算效率。更多还原

【Abstract】 The aggregate motion of a flock of birds, a herd of land animals, or a school of fish is a beautiful and familiar part of the natural world. But this type of complex motion is rarely seen in computer animation. In recent decades, due to the continuous development of the field such as film, television, animation, virtual reality and computer games, crowd animation has become a hot research field. Especially the research about behaviors while large crowds interactive in real-time, have been widely used in the field of training systems, safety science, computer games and so on. It does this by exploring the physical, biology and sociology nature of crowd behavior in the real world, and the use of computer to establish models, and thus making the behavior of the crowds more realistic. Agent-based behavioral animation focus on the behavior modeling, handing details of behaviors and movements, Compared with typical computer animation modeling which only concerned with the appearance of characters and physical properties. These behaviors include low-level path planning, high-level emotional interaction and so on. However, due to the complexity and diversity of individual behavior in crowds, control of each individual in crowds brings expensive computation, so it’s difficult to meet the real-time simulation applications. This paper, as study thousands of individuals, is attempted to find a good balance between realism and efficiency of crowd behavior simulation.It shows some representative crowd behavior models, including particle system, flocking system, behavior system, hybrid system, chaos system, cognitive model and so on; describes the architecture of character animation system, the framework of crowd animation sub-system and its position in the whole system; presents a method of crowd animation modeling, and on the basis to implement a large crowd simulation in real-time, which displays the realistic crowd behavior, including path planning, collision avoidance, and path following.The study focuses on the following two aspects:First, making a comprehensive research and thorough analysis with the method of crowd simulation, and proposing a framework of large-scale crowd real-time simulation base on behavior model. Such a framework has a hierarchical structure, and information sharing between levels with high efficiency. And its inherent limited logical sequence has greatly reduces the contradiction between behaviors.Second, for the core issue of large crowd animation, this paper presents an improved algorithm of local path planning based on dynamic environment, improves dynamic local path planning algorithm with global knowledge, provides the optimal path similar to global planning, but also has efficiency according to dynamically updated based on local knowledge, and unified planning, replacing the cost of calculating separately for each individual, effectively reducing the computation and improve the computation efficiency.更多还原