James Thigpen

This semester I am taking a Graduate Colloquium Class which consists of attending various seminars presented through the department and writing a short report about each one. It’s designed to give students an idea of what is going on in the research community. I have decided to include my reports in on this blog, as someone may be interested in them.

Interactive Physically Based Simulation
Dr. John Keyser
Associate Professor
Department of Computer Science
Texas A&M University
10/01/2007

Dr. Keyser presented his work on Interactive Physically Based Simulations. The goal of his work is to produce real-time simulations which are not physically accurate but are good enough estimates to be visually believable.

Dr. Keyser discussed the differences between a real-time estimation and a physically accurate simulation in the context of applications. A training simulator for example may require real-time simulation whereas aerospace simulations require physical accuracy. An aerodynamics engineer will not be overly concerned as to whether or not his simulation is real-time if the results are completely wrong even if they are visually believable. His work also has applicability in the computer graphics domain, as it is beneficial for an artist to control scriptable elements. Scriptable elements are much easier to control than initial condition simulations. It’s difficult to understand how changes in initial conditions will affect the end result, so a physically accurate simulation is not of much use to an artist creating graphics for a movie.

Real-time simulation is achieved through simplification of the underlying simulation models, and through a careful decoupling of the various models. A particular simulation may be represented as several models in the computer in order to simulate the various effects taking place. Knowing which model depends on which other models, it is possible to minimize inter-model communication as time progresses. Real-time for his case is defined as a usable “action:response” cycle frequency with the simulation system.

Dr. Keyser showed various examples of his work including a fire simulation which included fuel concentration, smoke, and deformation of objects (such as matches). He also displayed a stunning water simulation which, in certain scenarios, was relatively close to a physically correct simulation. It used wave particles to simulate the waves spreading across the surface which was then rendered using a grid based approach. The use of a particle based system allowed hardware acceleration and provided nice looking results and a real-time interaction rate when using up to 300,000 particles simultaneously.