Stable And Efficient Cloth Animation

Cloth modeling and animation is a hot topic in computer graphics. In the early stage of cloth modeling, geometry approaches are usually employed, but the results are poor. Since the employment of physically based approaches, cloth animation has gained great progress. However, these methods are usually not very efficient or the physical plausibility is poor. The main problem is that physically based approaches involve numerically solving a differential equation, and the numerical instability limits the time step. Thus, the computation cost is very expensive.In this thesis, an efficient and stable method to animate virtual cloth based on physics model is proposed. The method revises both the physics model and the numerical integration method, so it is very efficient and stable. We employ the theory of column buckling to simulate the compression and bending phenomenon of cloth. To reduce computation, we simplify the compression and bending model to linear spring. To solve the linear equation system in numerical integrating, we employ relaxation method and only a few iterations are needed. This method is more efficient than conjugate gradient one, and can be computed in O(n) time complexity. Compared with other approximation implicit method, our approach has the advantage of stable, accurate and realistic. Experimental results show our method is satisfactory and is efficient enough to simulate middle-scale cloth model in real time. Because our method can be computed in O(n) time complexity, it can be used for large-scale cloth simulation.