Accurate and Efficient Visual Simulation of Fiber-based Mechanical Structures

Cornell University, Ithaca NY

Investigators

Abstract

Accurate and Efficient Visual Simulation of Fiber-based Mechanical Structures Steve Marschner, Cornell University Many of the most difficult challenges in simulation for computer graphics involve materials composed of fibers. In applications where appearance matters, from designing textiles and apparel to rendering virtual actors to formulating hair cosmetics, understanding how these materials' physical properties affect their motion and appearance is crucial. Fiber-based structures behave differently from other materials. A knit fabric is much more extensible than its component yarns because the yarns can bend and slide into different macroscopic shapes. A head of long hair moves freely but smoothly, making and breaking millions of hair-hair contacts to produce a partly smooth, partly discontinuous overall motion. This research aims to improve the accuracy of cloth and hair simulation by fundamentally changing the underlying physical and computational models. The project follows an approach of measurement, modeling, simulation, and validation. Image-based measurement is used to capture complete fabric deformations. This data, together with the literature on yarn-level textile mechanics, is used to develop constitutive models based on the actual physics of fabric, which are simulated using advanced numerical methods not normally employed in textile simulation. The results are being evaluated both against the laboratory data and in the context of real problems in apparel design. By bringing together the tools of modern graphics and simulation with the detailed models of textile mechanics in a rigorous process of validation, this project is placing computer graphics cloth simulation on a firm scientific foundation for the future. Computational methods devised for fiber-fiber interactions are also being extended for efficient strand-level simulation of hair. After the simulations are validated, the project will examine approximations and algorithms to allow efficient simulation. The end result will be new, physics-based models for the mechanics of cloth and hair that are appropriate for visual applications.