Methods for Motion Generation and Interaction with a Humanoid Robot: Case Studies of Dancing and Catching

We focus on creating realistic, adaptable movement for humanoid robots and virtual characters. Here we present motion synthesis of dance movements for a humanoid robot, and interactive behavior for catching. Our approach to motion generation includes collection of example human movements, handling of marker occlusion, extraction of motion parameters, and trajectory generation, all of which must be handled in such a way as to be faithful to the style of the original movements. In our interactive behavior, we generate ball-glove impact predictions and intercept motion trajectories for a real time catching task. In this paper we present our results and discuss ideas for future improvements.