Hybrid techniques for high-fidelity physical simulation of solids and fluids.

详细信息   

• 作者：Selle ; Andrew Paul.
• 学历：Doctor
• 年：2008
• 导师：Fedkiw, Ronald
• 毕业院校：Stanford University
• 专业：Computer Science.
• ISBN：9780549623236
• CBH：3313656
• Country：USA
• 语种：English
• FileSize：11494780
• Pages：167

文摘

This dissertation presents hybrid algorithms for the simulation of fluids and solids as well as methods for coupling between the two. Physical simulations have become prevelant in computer graphics due to their ability to predict phenomena with too many degrees of freedom for hand animation. Nevertheless, it has been challenging to obtain simulations with sufficiently high quality in a reasonable amount of time, leading to the development of a large variety of geometric representations and simulation algorithms, each suited to a particular class of problems. While this makes choosing and implementing simulation technology difficult, this diversity of techniques results from attempts to balance tradeoffs. This suggests that increased simulation fidelity and efficiency can be obtained by synthesizing hybrid techniques, and this thesis presents examples of such techniques. In addition to algorithmic improvements we also consider the use of distributed memory parallelism (MPI) to improve the tractability of highly detailed simulations.;In the first chapters, we concentrate on increasing the fidelity of fluids by reducing numerical dissipation. Truncation error in the solution of the hyperbolic advection equation is reduced by a new unconditionally stable advection method. Two semi-Lagrangian advection steps are combined to obtain second-order accuracy in space and time while also providing a constructive reinterpretation of MacCormack's advection method. To reduce dissipation in smoke, water and explosions, we also present a Lagrangian/Eulerian scheme that couples a Lagrangian vortex particle method with an Eulerian pressure-velocity method. Additionally, we consider coupling Lagrangian deformable and rigid thin shells to Eulerian fluid flows using robust ray casting, fully preventing fluid leaks across the thin boundary.;In the next chapters, we discuss simulation of solids using Lagrangian techniques. We introduce improved altitude springs for the simulation of volumetric tetrahedra, which we apply to bending in cloth and torsion in hair. We also introduce a simple mass-spring model for hair simulation that uses tetrahedral simulation techniques for simulating straight and curly strands of hair. For time integration we develop an unconditionally stable fully-implicit linear spring that can be included in a semi-implicit Newmark time integration scheme. For collisions and interactions, we develop a hybrid scheme that uses inexpensive history-based repulsions coupled with more accurate geometric collisions. We also present a method for obtaining accurate friction of deformable objects with collision bodies when a semi-implicit or fully-implicit time evolution approach is used.