Research On Collision Detection Between Cloth And Rigid Body Using Adaptive Enclosing Ellipsoids

Cloth simulation has always been an important and difficult item in computergraphics and virtual reality. With the improvement of computer hardware performance, thecloth simulation technology tends to be more practical in daily life. The application refersto many fields such as virtual exhibition, game animation, film effects, and garment designand so on. The result and efficiency of cloth simulation depend on cloth modeling andcollision detecting, so this paper makes discussion and research on them for the sake ofachieving a realistic and real-time simulation.Firstly, in order to model the arbitrary shape and physical properties of cloth, weconstruct the mass-spring model based on triangular grid. Newton Second Law is appliedto analyze force and get motivation equation of the mass. Via comparison among commonintegration methods, we choose Verlet method which is fast and has good stability. In themass-spring model, high-strength springs cause wobbling phenomena, and low-strengthsprings cause super-elasticity phenomena in cloth simulation. To avoid the aboveproblems, we propose the nonlinear spring whose elasticity is not constant but changingwith its deformation length. As a result, the super-elasticity phenomenon is successfullyrestrained.Secondly, for the purpose of accelerating collision detecting, we adopt the enclosingellipsoid to replace the complex geometric model. We use the surface error as thenumerical value of the approximate error between the model and the ellipsoid. The surfaceerror is calculated by the weighted average of the normal distance and the radial distance.Based on the surface error, a segmentation of the model is adaptively computed with theoptimized K-mean clustering method for generating a union of enclosing ellipsoids. Theseellipsoids enclose the model tightly and avoid the complex hierarchical structure ofgeneral bounding volume.Thirdly, we propose a new method of classifying bounding volumes. According to thesurface error, enclosing ellipsoids are classified into two classes. While the surface error ofthe ellipsoid is small enough, cloth grid points are detected with the ellipsoid instead of the model. Otherwise, cloth grid points should be detected with triangles of the modelafter being detected with the ellipsoid.Finally, we design and implement a real-time dynamic cloth simulation system tovalidate the effectiveness of our algorithm proposed in this paper. Experiments show thatthis system can satisfy requisitions of real-time and realistic cloth simulation.