Meshfree physically-based animation of natural phenomena in computer graphics

点击翻译按钮获取中文摘要

Title: Meshfree physically-based animation of natural phenomena in computer graphics My research program for the next five years will focus on a mesh-free method for physically-based animations of natural phenomena. We recently developed a new and simple kernel-based method for solving partial differential equations. My plan is to further extend the scheme in order to apply it to simulating natural phenomena dynamics in computer graphics such as fluids, solids and cloths. Smoothed Particle Hydrodynamics (SPH) is popular for simulating liquids due to its simplicity when considering the simulation and the visualization of the fluid. SPH is also considered as being well suited when dealing with irregular domains and topological changes. But using SPH when dealing with sharp domain edges does not give as accurate results as it does with mesh-based variational methods. Hence particle-based methods for simulating fluids were enhanced with meshes to deal more accurately with sharp boundaries. Our Mesh-free Kernel-based method is more accurate than the SPH approximation, yet its use offers a similar simplicity. In the proposal, I emphasize simplicity and ease of use, believing that it is a key ingredient in having a true impact especially in the field of interactive applications. In the short term, we will work on more accurate computation near the boundaries, without having to resort to a mesh. This is possible with the method due to a promising feature of the method: its flexibility and in particular, the possibility to consider an upwinding of the kernels. Finally, I will generalize the method to encompass other matter than fluid. Particle-based methods have been used for simulating unified models, which facilitates the interactions between different types of matter (Fluid-Solid, Fluid-Cloths, etc.). It is also important to mention that, regarding interactive applications, this method can be used to develop animation algorithms, and will adequately map to the Graphics Processing Unit (GPU).