Thesis Presentation

Highly interactive characters that behave in situationally appropriate ways are an important goal of researchers, film makers, and game developers. For example, synthesizing believable boxers would involve representing the stylistic bobbing and weaving motions while generating realistic dynamic responses to blows that are given and received, including the resulting interactions with the boxing ring. Realizing that no single motion synthesis technique is perfect for every situation, we propose a hybrid system that favors the one most suited to the current objectives.

Motion graphs are an effective tool for synthesizing realistic and easily directable characters that accurately reproduce the stylistic nuances of human motion. However, because motion graphs rely on splicing data obtained before runtime, they are not adequate for applications where the external forces or detailed interactions with the environment cannot be predicted. Simulation allows the physical interactions between a character and its environment to be modeled realistically, but does not provide a wide range of behaviors because of the difficulty in constructing control systems for complex behaviors.

This thesis combines the complementary strengths of these two techniques so that complex animation tasks in novel environments may be synthesized interactively. Our system attempts to reasonably resolve when either technique is most appropriate and provide the facilities to transition between them as the character's goals and interaction with the environment evolve. To ensure these transitions are smooth, a fast variant of the Approximate Nearest Neighbors search algorithm is developed to locate a good correspondence between the simulation and motion database. Joints are actuated by physical controllers to guide the simulation near existing motion found using the correspondence search. These proportional derivative controllers are also used to add human-like subtlety to the simulated motion in a biomechanically inspired way for behaviors such as protective falling. We demonstrate and evaluate the power of this approach by switching between simulated and data-driven tasks in a context dependent way, triggered by physical interactions with the virtual human. As simulation techniques improve, such an architecture can support the future goal of fully autonomous simulated characters, while still being able to fall back on motion data for hard to simulate behaviors.

Thesis Committee:
Jessica Hodgins, Chair
Nancy Pollard

Close this window