Improvements to Physically Based Cloth Simulation

Tsz Ho Wong
School of Computer Science and Information Technology, Science, Engineering, and Technology Portfolio, RMIT University, Melbourne, Victoria, Australia
RMIT University, 2014


   title={Improvements to Physically Based Cloth Simulation},

   author={Wong, Tsz Ho},



Download Download (PDF)   View View   Source Source   



Physically based cloth simulation in computer graphics has come a long way since the 1980s. Although extensive methods have been developed, physically based cloth animation remains challenging in a number of aspects, including the efficient simulation of complex internal dynamics, better performance and the generation of more effects of friction in collisions, to name but a few. These opportunities motivate the work presented in this thesis to improve on current state of the art in cloth simulation by proposing methods for cloth bending deformation simulation, collision detection and friction in collision response. The structure of the thesis is as follows. A literature review of work related to physically based cloth simulation including aspects of internal dynamics, collision handling and GPU computing for cloth simulation is given in Chapter 2. In order to provide a basis for understanding of the work of the subsequent chapters of the thesis, Chapter 3 describes and discusses main components of our physically based cloth simulation framework which can be seen as the basis of our developments, as methods presented in the following chapters use this framework. Chapter 4 presents an approach that effectively models cloth non-linear features in bending behaviour, such as energy dissipation, plasticity and fatigue weakening. This is achieved by a simple mathematical approximation to an ideal hysteresis loop at a high level, while in textile research bending non-linearity is computed using complex internal friction models at the geometric structure level. Due to cloth flexibility and the large quantity of triangles, in a robust cloth system collision detection is the most time consuming task. The approach proposed in Chapter 5 improves performance of collision detection using a GPU-based approach employing spatial subdivision. It addresses a common issue, uneven triangle sizes, which can easily impair the spatial subdivision efficiency. To achieve this, a virtual subdivision scheme with a uniform grid is used to virtually subdivide large triangles, resulting in a more appropriate cell size and thus a more efficient subdivision. The other common issue that limits the subdivision efficiency is uneven triangle spatial distributions, and is difficult to tackle via uniform grids because areas with different triangle densities may require different cell sizes. In order to address this problem, Chapter 6 shows how to build an octree grid to adaptively partition space according to triangle spatial distribution on a GPU, which delivers further improvements in the performance of collision detection. Friction is an important component in collision response. Frictional effects include phenomena that are velocity dependent, such as stiction, Stribeck friction, viscous friction and the stick-slip phenomenon, which are not modelled by the classic Coulomb friction model adopted by existing cloth systems. Chapter 7 reports a more comprehensive friction model to capture these additional effects. Chapter 8 concludes this thesis and briefly discusses potential avenues for future work.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2017 hgpu.org

All rights belong to the respective authors

Contact us: