GPU加速的支持交互式切割的柔性体实时变形算法  

GPU-Accelerated Real-time Deformation Algorithm for Deformable Objects Supporting Interactive Cutting

作　　者：贾世宇[1] 潘振宽[1] 

机构地区：[1]青岛大学信息工程学院,青岛266071

出　　处：《系统仿真学报》2013年第9期2188-2195,共8页Journal of System Simulation

基　　金：山东省自然科学基金(ZR2009GM014)

摘　　要：使用OpenGL和GLSL实现了GPU加速的柔性体实时变形算法。变形计算使用共旋线性有限元法,可以处理大尺度旋转情况。为了能够与交互式切割操作协同运作,GPU数据结构使用带激活标志的可变长线性数组。将切割过程中需要更新的单元分为"全肮脏"和"部分肮脏"两类,以此降低GPU数据的更新量。针对GPU不能完成发散操作的缺点,设计顶点相邻四面体信息数据结构和相应的切割更新算法,将力的发散操作变为聚合操作。使用四面体单元刚度矩阵高频模式过滤方法来提高稳定积分的时间步长,以此降低切割产生的退化单元对变形计算稳定性的负面影响。最后给出若干仿真测试的结果,对仿真效果和运行效率进行了分析。GPU-accelerated real-time deformation algorithm of deformable objects was implemented using OpenGL and GLSL. The algorithm used co-rotational linear finite element method in order to deal with large-scale rotation. To co-operate with interactive cutting, GPU data structures used linear scalable arrays with active flags. Elements to be updated during cutting were classified as ＂full dirty＂ or ＂partial dirty＂ to reduce the amount of GPU update data. To overcome GPU＇s inability to do scattering, force scattering was converted to gathering through carefully designed data structure and updating algorithms for connected tetrahedron information of vertices. Filtering of high frequency modal of tetrahedron element stiffness matrices was used to increase stable integration time step, thus reducing the negative impact to deformation stability by degenerated elements created during cutting. Several simulation scenarios were tested. Their results and computational efficiencies were analyzed.

关 键 词：柔性体 实时变形 交互式切割 共旋线性有限元 GPU加速 高频模式过滤 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]