基于可编程图形处理器的实时烟雾控制模拟  

Real-Time Control of Smoke Simulation Based on Programmable GPU

作　　者：湛永松[1] 杨明浩[2] 石民勇[1] 费广正[1] 

机构地区：[1]中国传媒大学数字技术与艺术研发中心,北京100024 [2]中科院自动化所数字互动媒体实验室,北京100080

出　　处：《工程图学学报》2008年第4期60-64,共5页Journal of Engineering Graphics

基　　金：国家自然科学基金资助项目(60403037);教育部科技研究重点资助项目(JK2G02035)

摘　　要：提出一种基于可编程图形处理器(GPU)有效控制烟雾模拟过程的实时算法。对于给定的烟雾当前状态和目标状态,该算法以自然的烟雾流动方式实现状态之间的转换。根据Navier-Stokes方程定义烟雾流体场的物理模型,通过调整方程中的风力项以达到控制目的。风力项由驱动力和聚合力组成,驱动力促使烟朝着目标密度分布运动,聚合力则用于抵消烟的扩散现象。烟雾的速度和密度参量被整合为纹理颜色通道,并传送到像素程序中由GPU完成计算。该算法无需保证两状态之间转移为最优,从而消除了非线性优化所带来的计算成本。实验表明该算法能实时高效地模拟两种烟雾状态之间的转换。A real-time algorithm for programmable GPU is presented. Given the efficiently controlling animated smoke based on target smoke state, the method generates a smoke simulation in which the current smoke state is driven towards the target in a natural-looking way. The smoke is modeled by the Navier-Stokes equation, and two terms are added to the standard flow equation： one is the driving force term that causes the fluid to carry the smoke towards a particular target, the other is the smoke gathering term that prevents the smoke from diffusing too much. The smoke velocity and density are integrated as color channels of texture and then passed to fragment program processed by GPU. The method need not to guarantee the optimization of state transfer, thus no expensive cost is required, and the whole simulation is accelerated by GPU. Experiments show that the algorithm is effective to simulate smoke state transfer in real time.

关 键 词：计算机应用 烟雾控制模拟 流体力学 纳维-斯托克斯方程 图形处理单元 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]