Adaptive anisotropic remeshing for cloth simulation

Abstract

Figure 1: A motion-captured character performs a jumping kick. His clothing is dynamically remeshed to capture detail such as wrinkles, while having larger elements in smooth areas. Here and elsewhere in the paper, large elements are shown in blue, small equilateral elements in red, and anisotropic elements in yellow. AbstractWe present a technique for cloth simulation that dynamically refines and coarsens triangle meshes so that they automatically conform to the geometric and dynamic detail of the simulated cloth. Our technique produces anisotropic meshes that adapt to surface curvature and velocity gradients, allowing efficient modeling of wrinkles and waves. By anticipating buckling and wrinkle formation, our technique preserves fine-scale dynamic behavior. Our algorithm for adaptive anisotropic remeshing is simple to implement, takes up only a small fraction of the total simulation time, and provides substantial computational speedup without compromising the fidelity of the simulation. We also introduce a novel technique for strain limiting by posing it as a nonlinear optimization problem. This formulation works for arbitrary non-uniform and anisotropic meshes, and converges more rapidly than existing solvers based on Jacobi or Gauss-Seidel iterations.

Efficient and robust strain limiting and treatment of simultaneous collisions with semidefinite programming

“…The table at top right shows the number of iterations taken by these two methods to converge. It is clear that our method converges faster (in one iteration) than Narain et al's method [3]. The disadvantage of our method is that each iteration takes longer to perform.…”

“…Anisotropic strain limiting methods [2] in cloth simulation limit strains of edges. Isotropic methods [1,3] act on the strain tensors and limit the singular values of the deformation gradient matrices of triangles. Our method also performs isotropic strain limiting for both triangular surface meshes and tetrahedral volume meshes.…”

“…Narain et al [3] posed strain limiting as a nonlinear optimization problem and used an augmented Lagrangian method to solve it. Collision handling plays a significant role in physically based simulation.…”

“…Narain et al [3] adapted an augmented Lagrangian method to solve the problem with strain constraints. However, it converges slowly and is not robust in practice, as shown in Figs.…”

“…As the picture shows, our method generates the best result. The meshes in (c) were generated by Narain et al's strain limiting method [3] using 10, 33, and 987 iterations respectively. The table at top right shows the number of iterations taken by these two methods to converge.…”