GPU-Based Rigid Body Dynamics

The simulation of rigid body dynamics is an area of interest both for scientific research and engineering applications. The accurate numerical solution of the underlying complementarity problems is complex. The use of the computational power of current graphics processing units (GPUs) to solve these systems seems promising. Using NVIDIA ’s CUDA technology for General-Purpose computation on Graphics Processing Units (GPGPU), an object-oriented iterative solver using a Jacobi method for the efficient simulation of rigid body dynamics is developed. The solver is then integrated into an existing physics engine as one of multiple possible methods for the collision response. Special care is taken to provide an easy to use interface for the synchronization between the host system and the graphics card. The final part of the thesis is focused on the performance evaluation of the implemented solver.