Approximate Data Driven Parallel Shape Matching for Soft Body Physics Simulations

We propose an algorithm that is fully parallel and has linear time complexity for soft body simulation that addresses three principal issues; Visual Quality, Performance and Ease of use. It works using the power of precomputed collision result look-up data and basic approach of shape matching. Since data driven shape matching approach only uses user generated precomputed collision results, deformation results cannot be unexpected. This creates visual quality and improves ease of use. Also, usage of these look-up data opens ways to improve Performance. In our tests, we achieved direct linear speed up depending on the processor's core count.