HCC: Small: Rethinking Simulation in Computer Graphics

Texas A&M Engineering Experiment Station, College Station TX

Investigators

Abstract

Abstract ? Keyser (0917286) Simulations play an important role in many graphical applications, ranging from entertainment to virtual environments used for training. These applications demand greater and greater realism, and this in turn creates a need for more believable and more efficient simulations. While there have been major improvements in simulation technology, many of these techniques can be quite slow, and simply applying greater computing power is not sufficient to meet the demands of real-time systems. The focus of this research is on finding ways to create real-time simulations by taking different approaches to the simulation framework used in computer graphics. The researchers explore ways to replace simulation by statistical data, simplify the theory of effects-based simulations, and incorporate these approaches into a system that trades off accuracy and speed to meet the requirements of a given problem. This work can change the manner in which simulation is performed in graphics and other applications. This research deals with developing methods for real-time simulation by taking fundamentally different approaches to the simulation problem. First, the researchers investigate ways of replacing full physics-based simulations with statistically-based capture of simulation effects. Rather than simulating at run-time, statistics are gathered regarding simulation behavior across several samples, from either full simulations or other sources. At run-time, the simulation is replaced by a result generated according to the statistical distribution. Second, the researchers develop ways of simplifying the theory of certain simulation systems to allow simplified simulations that still capture the important effects, while leaving the less important details to be handled by other (less computationally-involved) means. Third, the researchers investigate ways to develop level-of-detail simulations that allow a tradeoff between fidelity and efficiency, supporting both a highly accurate simulation and a faster simulation with guaranteed performance. The statistical and simplified approaches are incorporated into this level-of-detail simulation.