Easy humanoid motion generation from user demonstration using wearable interface

Abstract

Humanoids should have not only a similar appearance with humans but also abilities to perform human-like behaviors. However, today's humanoid robots are not smart enough to adapt to their working environments, so they need some help to learn new behaviors. Especially for an entertainment purpose, an easy motion generation method is very important for end users who are not experts on robot programming. In this paper, we propose a method to generate motion of humanoid robots from user demonstration using an intuitive wearable interface. A curve simplification algorithm and a clustering method are applied to extract motion primitives. Transition probabilities among the motion primitives are calculated to make a motion model and motions are regenerated by rearranging the motion primitives based on the motion model. A wearable interface is developed for capturing user demonstration and interacting with a partner robot. A humanoid robot, AMIO is used to test the generated motion.