Turbulent Fluids for Interactive Graphics

Physics simulations are widely recognized to be crucial tools for complex special effects in feature films, and real-time simulations are often central game-play elements in modern computer games. There are, however, inherent difficulties with these simulations: we are still very far from being able to accurately simulate the complexity of nature around us. Additionally, the numerical methods that are commonly used are notoriously difficult to fine-tune and control. The central goal of my research is to address these issues with novel multi-physics solvers.

In this talk I will highlight this goal by discussing my research on controllable simulations of anisotropic turbulence. Phenomena such as the rising smoke of a chimney, a burning fire, and an exploding car are typical examples where fluid simulations are heavily used in computer graphics. As these every day phenomena exhibit complicated shapes and motions that are completely impractical to re-create manually, numerical simulations have become an invaluable tool in the CG industry to realize such effects. However, these simulations suffer from the problems mentioned above: they have very long run-times, many days in the worst case, and they are very hard to control due to the non-linear nature of the underlying equations.

This is where the wavelet turbulence approach, which I developed at ETH Zurich together with colleagues from Cornell University, comes into play. For the first time, it allows for a two-stage work-flow, such that artists can quickly iterate on a low-resolution version of a flow, and increase the amount of detail in a decoupled second step. An intuitive overview of this algorithm will be provided, together with an explanation of its physical motivation. The approach, published in 2008, has been widely successful, and has become a de-facto standard for detailed simulations of explosions and smoke in movies. Apart from its use in feature films, I will explain how these ideas can be adopted for interactive simulations. A central aspect of this direction is the use of a more sophisticated turbulence model, that allows for a more accurate prediction where visual details will develop. In addition, this model is very useful to extract information about directional preference of vortex formation from the flow. It will be demonstrated that this approach gives detailed turbulence effects with high frame rates, suitable for e.g. computer game effects.

In general terms, the goal of my research is to realize interactive, controllable solvers for a broad range of material behaviors, from splashing fluids to brittle, fracturing objects. The talk will be concluded by giving an outlook of future research projects, and the industry impact that can be expected from advances in the field. Especially recent trends, such as the pervasive nature of handheld devices, and the growing use of medical applications highlight areas where I expect interactive physics simulations to have a broad, and lasting impact.
Nils Thuerey currently has a position as research & development lead at ScanlineVFX, working on the design and implementation of large-scale physics simulators for visual effects in feature films. His research focuses on physically-based animation, with a particular emphasis on detailed fluids and turbulence. Some of his algorithms are now widely used in industry, e.g., as part of animation packages such as Houdini and Blender. He received his Ph.D in 2007 from the University of Erlangen-Nuremberg (with honours), and until 2010 worked as a post-doctoral researcher with Ageia/Nvidia and the Computer Graphics Laboratory of ETH Zurich.