Slab and Powder-Snow avalanche animation on the GPU

Sammanfattning: Background: The video game industry has yet to achieve a physically-based real-time avalanche simulation because of the sheer complexity of modeling the behavior of snow avalanches. An avalanche is made out of snow, meaning it would require a snow simulation which itself already shows to be complex. However, a real-time snow simulation has shown up recently, making this thesis worth investigating. Objectives: The proposed method will take advantage of a real-time snow simulation framework to animate slab and powder-snow avalanches. The powder-snow avalanche is divided into two types: a loose-snow avalanche and a powder-cloud avalanche. Animating the avalanches will require tuning the parameters to get the different types and calculate the release area on a terrain. Lastly, a survey is sent out to verify the viability for use in games of the avalanches. The performance will be measured and analyzed along with the gathered data from the survey. Methods: Two particle systems were used to animate the avalanches. The Discrete Element Method was used to animate the slab and loose-snow avalanches, whereas the powder-cloud avalanche utilized the Smoothed Particle Hydrodynamics method. A procedural Voronoi pattern was used for generating the slabs and hypertexture to render the powder-cloud. In contrast, a fluid renderer was used to render the snow. Results: Measurement of the proposed avalanche animations was conducted and analyzed. The proposed slab and loose-snow avalanches reached real-time performance depending on the number of particles and the shading when rendering the scenes. However, the powder-snow avalanche did not fulfill the real-time performance criteria. Furthermore, the survey was used to verify if the proposed avalanche animations were viable for games. Both the slab and loose-snow avalanches were seen as viable for use in games, while the powder-snow avalanche was not seen as viable for games. Conclusions: The proposed slab and loose-snow avalanche animations ran in real-time with a dynamic particle count of 300k or lower, without shaded rendering, or 75K or lower with shaded rendering. Both avalanches were seen as viable for use in video games. Furthermore, the powder-snow avalanche could not reach a real-time performance of over 30 frames per second and was not seen as viable for use in video games. Further research is needed for the powder-snow avalanche.