This thesis analyzes how parallel processing with Graphics Processing Unit (GPU) could be used for massive crowd simulation, not only in terms of rendering but also the computational power that is required for realistic simulation. The extreme population in massive crowd simulation introduces an extra computational load, which is quite difficult to meet by using Central Processing Unit (CPU) resources only. The thesis shows the specific methods and approaches that maximize the throughput of GPU parallel computing, while using GPU as the main processor for massive crowd simulation. The methodology introduced in this thesis makes it possible to simulate and visualize hundreds of thousands of virtual characters in real-time. In order to achieve two orders of magnitude speedups by using GPU parallel processing, various stream compaction and effective memory access approaches were employed. To simulate crowd behavior, fuzzy logic functionality on the GPU has been implemented from scratch. This implementation is capable of computing more than half billion fuzzy inferences per second.

Background. Several methods have been used to simulate soft body deformation, such as mass-spring systems and position-based dynamics. This has been done using tetrahedral mesh models for preservation of shape and volume. In real-time applications however, there is a limitation to how high resolution the model can be, creating the need for optimizations. Objectives. To achieve better performance for high resolution models, tetrahedral deformation is used, making it possible for the tetrahedral mesh and triangle mesh to use different resolutions. In combination with this, the GPU is used to execute the simulation in parallel, improving performance further. Methods. For evaluation of performance and accuracy, an implementation was created to simulate soft body deformation using extended position-based dynamics and the Vulkan graphics API, with the option to use tetrahedral deformation. By experimentation, comparisons are made between using different resolutions on the tetrahedral mesh to the full resolution in terms of performance and accuracy. Results. The results show that performance and accuracy are altered when using tetrahedral deformation on lower resolution tetrahedral mesh. The performance is improved based on the decrease in workload, such as with higher base resolution models or multiple soft bodies. The accuracy is however not correlated to the reduction of resolution, but instead dependant on the rest shape of the model used. Conclusions. The implementation created demonstrates a new optimization that can be used to simulate soft body deformation in parallel on the GPU, with a smaller change in accuracy. Improvements exist in areas of usability, features and other optimizations that can be further explored in future research.

No description available.