Cloth simulation has been a problem of interest to the communities of graphics and engineering for decades. Although many effects have been done in this area, generating efficient and realistic cloth simulation results is still a challenging problem. In the thesis, we present efficient algorithms for cloth simulation on two issues: collision detection and cloth modeling. The core of our collision detection algorithms is based on the uniform spatial subdivision technique. In our method, a fast voxelization approach is presented, the efficiency of self-collision detection is significantly improved by making use of the surface curvature, and a triangle mapping method is proposed to quickly verify the locality of the cloth surface. Based on the voxel method, we present a fast collision dete...[ Read more ]

Cloth simulation has been a problem of interest to the communities of graphics and engineering for decades. Although many effects have been done in this area, generating efficient and realistic cloth simulation results is still a challenging problem. In the thesis, we present efficient algorithms for cloth simulation on two issues: collision detection and cloth modeling. The core of our collision detection algorithms is based on the uniform spatial subdivision technique. In our method, a fast voxelization approach is presented, the efficiency of self-collision detection is significantly improved by making use of the surface curvature, and a triangle mapping method is proposed to quickly verify the locality of the cloth surface. Based on the voxel method, we present a fast collision detection algorithm for dressed human simulation. We accelerate cloth/human collision detection by exploiting spatial and temporal coherence. In the algorithm of cloth/human collision detection, the voxel technique is used to rapidly identify the potential collision region, and a collision tracing method is presented to quickly search collisions.

To speed up cloth simulation, we present an adaptive modeling method. In the method, the cloth surface is separated into regions with small deformations and regions with large deformations at each time step. Our strategy is to make the cloth vertices within the region with small deformations quickly reach their positions of the next time step, and refine the regions with large deformations. Therefore, we can produce realistic simulation results at a reduced computational cost.

Our algorithms have been implemented and tested. Experimental results show that our algorithms are efficient. Further, they are relatively simple to implement.