In collaboration with Eitan Grinspun and the Columbia Computer Graphics Group at Columbia University
Computational simulations of complex natural phenomenon have become useful not only for scientific research but also for visual design and entertainment. Key examples are the complex physics engines developed to simulate the behavior of multi-particle dynamic systems such as smoke, hair, and cloth in digital platforms like Autodesk Maya.
These tools operate by generating massive amounts of data at the particle level and then calculating individual physical interactions among the particles. In this case, limiting the particle sampling rate and resolution becomes key to optimizing the simulations. The goal of the ADVP is to work with the data generated by a dynamic hair simulation, using statistical and visualization tools to discover how micro relationships among particles predict macro results. This has the potential to reveal new patterns not predicted by the researchers, and guide the future optimization of the simulation engine. Finally, it can generate new insight on the functioning of the dynamic systems themselves.