Example Based Soft-Body Simulation on Graphics Processing Units

Abstract

Nowadays Soft-body Physics Simulations are used in lots of different areas such as movies, videos and video games. Depending on the application, use of simulation improves realism, fun and teaching factor thus creates more quality in the product. Generally, Soft-body Physics Simulations are very resource consuming operations. In this thesis we proposed a Soft-body Physics Simulation algorithm. Aim of this proposed algorithm is to address three of the big problems of Soft-body Physics Simulations; performance, visual quality and ease of use. Proposed algorithm uses precomputed soft-body physics simulation results and creates its outputs by using precomputed simulation results as examples for a potential solution. Method of using examples simplifies and reduces expensive computations thus improves performance. Visual quality depends on given examples. By simply improving visual quality of examples, visual quality of simulation can be increased. Algorithm only creates results by using examples so there cannot be any unexpected outputs, thus this improves ease of use. Proposed Soft-body Physics Simulation algorithm includes independent tasks and can be implemented fully in parallel. As a result, it can use full potential of graphical processing units, which became exceedingly popular in parallel computation for very promising performance. Proposed algorithm has a time complexity of O(n) where n being the number of vertices. Even in cases that contains three dimensional models that have millions of triangles proposed algorithm computes result in couple of milliseconds. It can easily be used in real-time applications.