Abstract

Simulated motion is central to a user's experience in many virtual environments, such as those for training or gaming. However, realistic motion is frequently expensive to compute, which has limited the size of environments and the complexity of the motion within them. Simulation culling addresses this problem by avoiding computation for motion that is out of the user's view. In this talk I will discuss various simulation culling strategies, with an emphasis on a new approach, simulation proxies. A proxy replaces an expensive, accurate simulation with a cheaper one when out of view. It is designed to be cheap to compute, yet still provide the user with a quality experience. In the environments we have explored, simulation proxies enable interactive worlds on a scale that cannot otherwise be computed in real time.
This is joint work with Okan Arikan and David Forsyth, both at UC Berkeley.