基於混合式模型簡化法的形變模擬

Abstract

In computer graphics, model reduction method that utilizes a low-dimensional subspace to approximate the original, high-dimensional deformation space can simulate deformation well in force-free conditions. However, when external forces are applied to the simulated objects, obvious differences between low-dimensional simulation and full-coordinate simulation can be observed. Therefore, to improve the simulation accuracy of reduced deformable models when the external forces are applied and to retain its advantage of fast run-time performance, we present a hybrid framework that utilizes bases constructed from forced and force-free deformations. The forced deformations are precomputed from data of full-coordinate simulation by applying external forces to different parts of the deformable object. This problem is formulated as a force sampling problem and solved by space partition and surface sampling. In the run-time stage, if there are external forces, we simulate deformation in the low-dimensional subspace of forced deformations. When the external forces are released, we simulate in another subspace by adopting the modal derivative bases because its computation is automatic and does not need any presimulation. To reduce the artifact in the transition, we linearly blend forced and force-free deformations. Our results show improved accuracy compared to the results of using only the modal derivative bases while the speedup over full-coordinate simulation is still significant.

In computer graphics, model reduction method that utilizes a low-dimensional subspace to approximate the original, high-dimensional deformation space can simulate deformation well in force-free conditions. However, when external forces are applied to the simulated objects, obvious differences between low-dimensional simulation and full-coordinate simulation can be observed. Therefore, to improve the simulation accuracy of reduced deformable models when the external forces are applied and to retain its advantage of fast run-time performance, we present a hybrid framework that utilizes bases constructed from forced and force-free deformations. The forced deformations are precomputed from data of full-coordinate simulation by applying external forces to different parts of the deformable object. This problem is formulated as a force sampling problem and solved by space partition and surface sampling. In the run-time stage, if there are external forces, we simulate deformation in the low-dimensional subspace of forced deformations. When the external forces are released, we simulate in another subspace by adopting the modal derivative bases because its computation is automatic and does not need any presimulation. To reduce the artifact in the transition, we linearly blend forced and force-free deformations. Our results show improved accuracy compared to the results of using only the modal derivative bases while the speedup over full-coordinate simulation is still significant.