RI Seminar: Katsu Yamane (2009)

In this talk

1. 6:26

   Why Disney Research Pittsburgh

   Yamane describes Disney Research as a group of labs across the company, including Pittsburgh, Zurich, and Walt Disney Imagineering. He says Disney’s parks use humanoid robots as part of their business, and offer varied settings for testing them. The Pittsburgh lab also has potential connections to animation, games, toys, ESPN, and CMU.

2. 11:31

   Simulation and numerical tools

   Yamane presents tools for simulating articulated robots and handling changes in grasp, contact, and collision. His parallel dynamics method handles open and closed chains; on a 161 degree-of-freedom human model, it ran in under a millisecond using four cores. He also describes an inverse-kinematics interface that lets animators move character links while managing constraints and singularities.

3. 28:18

   Adapting motion capture

   Yamane explains why captured human motion often needs adjustment before it can drive a different character, robot, or environment. A dynamics filter corrects physically infeasible motion, such as feet that float or slip, while preserving selected joint motion. Other examples use four pick-and-place recordings to create varied manipulation motions and adapt human movement to a motorized marionette with limited mobility.

4. 40:39

   Modeling human motion control

   Yamane describes a whole-body model with 155 degrees of freedom and nearly 1,000 muscles, paired with a model of spinal reflexes. The system estimates muscle forces from motion and can incorporate EMG measurements. Its learned reflex connections showed positive weights for muscles that work together and negative weights for opposing muscles, and it reproduced the patellar tendon reflex.

5. 49:22

   Robots for parks and guest experiences

   Yamane outlines plans to use motion-capture or animation data to program park robots, improve their locomotion, and recalibrate them as parts age. He also wants robots to interact with guests instead of only replaying recorded motions. Possible settings include ride queues, grassy areas, restaurants, and stores.