Physical Simulation with Impact

Brian Mirtich, EECS, UC Berkeley

Physically-based modeling is a powerful tool for engineers and 3-D content creators, with applications ranging from animation to simulation and design verification. Collisions and transient contact modes are present in many physical systems, but traditional approaches lack robustness and physical accuracy in modeling these phenomena. To address these problems, we have developed impulse-based simulation. This method of simulation replaces constraint equations with trains of collisions to affect the contact interactions between bodies. We discuss how collision detection may be performed efficiently in this context, and also describe a model for physically accurate frictional collisions.

Impulse-based simulation adeptly handles many systems which are difficult to model with constraints, but it is not a replacement for constraint-based simulation. In fact, the two paradigms are complementary. Both are necessary, for example, in modeling the complex interactions between a robot and its environment. We discuss hybrid simulation, a technique that combines impulse- and constraint-based methods, and also serves as a powerful substrate on which to design control systems for autonomous agents.