Fracturable Surface Model for Particle-based Simulations

Abstract: Providing destruction in games is usually achieved by having pre-calculated fracturingpoints, swapping models at appropriate times while hiding the crimes witha puff of smoke or an explosion. An area of continued research is procedural destructionwhere an object will fracture in a realistic way depending on appliedforces such as gravity, explosions or load. This thesis proposes and begins the implementation for a triangle based surfacerepresentation capable of supporting procedural destruction in real-time for anunderlying point-based simulation; deriving the methodology from the paper byM. Pauly et al [12]. Too wide project scoping prevented the implementation from fully realising theinitial goals; where the surface and physics simulations was never married intoa single simulation. It is one half of a larger project on procedural destruction,focusing primarily on the surface representation where the second half is detailedin the report by C. Stegmayr [14]. Even without a complete simulation, performance is an evidently limiting factor.For more detailed simulations, with a simple test mesh and a small step size whenpropagating a fracture, frame times quickly raise up to almost 247 ms/frame.There are multiple areas of improvement for the implementation to reduce frametimes; however, scalability and performance are still major points of concern dueto inherent challenges with running multiple fractures in parallel. Unless scalingcan be improved, it is worth pursuing alternative approaches.