Efficient motion generation for humanoid robots using programming by demonstration based on wearable interface

Abstract

Generating a motion for a humanoid robot is very cumbersome task for robot programmers who use machine level program codes to develop new behaviors of the humanoid robot. An intuitive and easy method of generating motions for humanoid robots is needed. In this study, a method to generate motions of humanoid robots based on the user demonstration is proposed. A wearable interface is designed and a prototype of the wearable interface is implemented to capture the user’s demonstration. Kinematic analysises are applied to translate the human motion for humanoid robots. A curve simplification algorithm was applied to create an editable motion from the original motion, and GART clustering algorithm was used for grouping the similar motion frames for motion editing. Two kinds of motion generation method was proposed; motion hierarchy based generation and stochastic motion model based generation. A scenario for a task of a humanoid robot could be generated from the motion hierarchy and the generated motions variated with two transformation methods. A long motion with the repetitive patterns could be generated based on the transitions between the keyframe clusters of the stochastic motion model, and the generated motion was variated with the stochastic keyframe creation based on the Gaussian distribution of the keyframe cluster. The generated motions are validated by testing to the physical humanoid robot platforms. This work provides a tool to generate motions of humanoid robots from the user demonstration for the people who do not know anything about the robot programming. By using the two different motion models proposed in this work, the reusability of the learned motion can be growing. As the result, the robot content creating process could become simpler, so anyone can create the new contents for humanoid robots. The expansion of the robot market and the employment of humanoid robots for providing services in daily lives are expected by applying the proposed m...