Fast Simulation of Inextensible Hair and Fur

AbstractIn this paper we focus on the fast simulation of hair and fur on animated characters. While it is common in featurefilms to simulate hair and fur of computer generated actors, characters are still mostly hand-animated in computergames. A main difficulty of simulating hair is that it is perceived as inextensible by humans. Preventing an objectfrom being stretched is a global, non-linear problem. This is the reason why simulating completely inextensibleobjects in real-time remains a major challenge and an open research topic.Existing approaches typically use multiple iterations per visual frame to solve the physical equations followed bynumber of strain limiting iterations. Adjusting the number of iterations is a way to increase the accuracy of thesimulation at the expense of more computation and vice versa. In the extreme case of one solver iteration per visualframe, most existing methods break down, either by becoming unstable or by introducing a substantial amount ofstretching.In this paper, we present a robust method that guarantees inextensiblity with a single iteration per frame. Thisextreme performance comes at the price of reduced accuracy. We found that for applications in graphics, it isworth to pay this price because the inaccuracies are not visually disturbing but the speed of the method allows thesimulation of thousands of hairs in real-time.Categories and Subject Descriptors