基于GPU实现的大量刚体运动实时仿真

摘要

针对存在大量刚体运动的虚拟场景,提出一种基于在GPU上实现的实时仿真算法,利用GPU的并行计算能力,实时处理刚体交互,更新刚体状态.使用深度剥离技术,离散化刚体,并使其由一组具有相同大小的粒子表示.每一帧刚体间的运动交互则由粒子间的运动交互来实现.碰撞粒子对间的交互则用离散元法.通过统一网格方式,分割仿真域空间,以提高碰撞检测速率.实验证明,提出的新算法大幅度的提升了仿真大量刚体的速率.%In view of virtual scene where exist a mass of moving rigid bodies, we proposed a real - time rigid body simulation algorithm on CPU. This algorithm handles rigid body's interaction and status in real -time, utilizing the tremendous computational power provided by GPUs. We discretized rigid body using depth peeling, to making rigid body represented by a set of particles with the same size. We utilized particle pairs' interaction to realizing rigid body's interaction. Discrete element method was used to handle collision detection. And we discretized simulation areas across uniform grid to improve collision detection. The results show that this algorithm can improve the rigid body simulation speed largely.