Physically-Based Simulation

Simulating the macroscopic dynamic behavior of nature is very important for many applications, such as in computer graphics, virtual/augmented reality, computational design and fabrication, medical diagnosis, as well as robot training. Our lab aims at creating efficient and accurate solver algorithm and computing system to simulate different kinds of dynamic physical phenomena, with a particular focus on fluid simulation, and with coupling to different kinds of solids. We have developed particular statistical kinetic formulation with local dynamic rules for efficient fluid simulation, both for air and liquid. We have also developed novel accelerating algorithm based on machine learning for fluid simulation. By coupling all types of materials and simulating their behaviors, we are creating a virtual world on computer systems, with the future goal of creating intelligent agents from such a virtual world.