Matthias Müller received his Ph.D. from ETH Zürich for
his work on the atomistic simulation of dense polymer systems. During
a two year post-doc with the computer graphics group at MIT he changed
his research focus from atomistic offline simulations to macroscopic
real time simulation in computer graphics. In 2002 he co-founded
Novodex, a company that developed a simulation engine for computer
games. In 2004 Novodex was acquired by AGEIA which, in turn, was
acquired by NVIDIA in 2008.
Physics in Games
Physical simulations have a long history in engineering and have been successfully used to complement real world experiments. Main advantages computer simulations have over real experiments are the ability to study extreme conditions and the analysis of very small time intervals. With this in mind, the accuracy of the models and the results are central to engineering applications.
For more than three decades, physical simulations have also been used in computer graphics in order to increase the realism of animations and to free artists from animating secondary motion by hand. The two main applications are special effects in movies and physical effects in computer games. Here, accuracy is important to the extent that plausible behavior is generated. There are, however, additional requirements not present in the engineering world that are more important than accuracy. One such requirement is controllability: movie directors and game developers want to be able to control how a building collapses or what path a flood wave takes in order to create the desired effect or to make sure game play does not get blocked. Another aspect that plays a major role, especially in games, is stability. The simulations need to be unconditionally stable even in un-physical situations such as characters turning 180 degrees in a single time step.
These new requirements are the reason why physically based simulation in computer graphics has become an important research field separate from scientific computing. In my talk I will present a variety of simulation methods we have developed to meet these requirements, while still producing plausible physical behavior. Examples are approaches to simulate soft bodies, clothing, destruction and liquids.