The electrifying stage of SIGGRAPH Asia 2023 witnessed Style3D Research’s triumphant unveiling of two ground-breaking research papers through captivating oral presentations. This pinnacle gathering, renowned as a global hub for computer graphics, provided the perfect canvas for Style3D’s innovation.
The first paper focuses on improving the stability of the dihedral-angle-based bending model (DAB) by using an analytic eigensystem and adaptive geometric stiffness.
We present novel analytic expressions for the eigensystem of a DABenergy Hessian, revealing that DAB models typically have positive, negative, and zero eigenvalues, with four of each, respectively. By using these expressions, we can efficiently project an indefinite DAB energy Hessian as positive semi-definite analytically. We also propose rectifying DAB’s indefinite geometric stiffness matrix by using orthotropic geometric stiffness matrices with adaptive parameters calculated from our analytic eigensystem.
Additionally, we suggest adjusting the compression stiffness according to the Kirchhoff-Love thin plate theory to avoid over-compression. Our method not only ensures the positive semi-defi niceness but also avoids instability caused by large bending forces at degenerate geometries. To demonstrate the benefit of our approaches, we have shown in the paper comparisons against existing methods on the simulation of cloth and thin plates in challenging examples. In the near future, we plan to integrate this method into our Style3D Physics simulation engine, delivering a more realistic and improved experience for our users with enhanced bending effects on clothes.
The second paper introduces a method to improve the efficiency of cloth simulation on GPUs. We propose an efficient cloth simulation method that combines the merits of subspace integration and parallelizable iterative relaxation within the framework of projective dynamics (PD). Our core ingredient centres around the utilization of subspace to expedite the convergence of Jacobi-PD. This involves solving the reduced global system and smartly employing its precomputed factorization.
The prefactorized subspace system matrix is exploited in a reduced-space LBFGS. TheLBFGS method starts with the reduced system matrix of the rest shape as the initial Hessian approximation, while the full-space Jacobi iteration captures high-frequency details. Our method can be efficiently executed on modern GPUs.Experiments show significant performance improvements for high-resolution cloth simulation. In the near future, we aim to implement this method in our Style3D Physics simulation engine, enhancing the overall user experience by offering faster simulation speed.