| There are three major problems in modeling the natural motion of virtual clothing in a 3D environment: (1) a large number of contact points, (2) the computations of collision information in the presence of numerical errors, and (3) interactions with objects having sharp features. This thesis proposes new techniques that successfully address these three problems. The results show that these new techniques can be applied on virtual garments with complex geometry and on unique interactions between continuous and non-smooth surfaces. Our techniques are classified into four categories: (1) culling non-colliding subsurfaces in collision detection, (2) intrinsic collision detection, (3) collision response, and (4) the editing and simulation of special virtual garments.; For culling non-colliding subsurfaces in collision detection, we propose an image-based method for interference test, a method of decomposition of a surface into a new type of (pi, beta, I)-surfaces for exact collision detection in the time domain and an adaptive backward voxel-based hierarchical structure for dealing with highly compressed deformable surfaces. For intrinsic collision detection, we propose a new system architecture. Robust treatments of numerical errors are devised. For collision response, a penetration-free motion space is proposed for handling features involved in multiple collision events, and a static analysis method is suggested for handling friction and stiction. Thus, interactions with objects having sharp features are handled efficiently. For the editing and simulation of special virtual garments, we propose techniques to handle multi-layered surfaces, surfaces with sharp features and sewn surfaces. A multi-layered surface is constructed by gluing several surfaces together either along lines or over regions. A surface with sharp features is represented by two meshes. A sewn surface is obtained by systematically performing a stable sewing process.; These techniques have been integrated into a high performance system for virtual clothing dynamics. |
