Adaptation of human motion capture data to humanoid robots for motion imitation using optimization

Abstract

The interactions between a humanoid robot and human are important, when the humanoid robot is requested to serve people with human-friendly services. For such interactions, the imitation of human arm motions by a humanoid robot is discussed as the first step for imitating the full body motion of a human. The human motions captured by a motion capture system may not be applied directly to the humanoid robot because of the differences in geometric aspect, dynamics and kinematic capabilities between the robot and human. To overcome this difficulty, a method to adapt captured motions to the humanoid robot is developed. The geometric difference in the arm length is resolved by scaling the arm length of the robot with a constant based on a length ratio. The imitation of human arm motion is then realized by solving an inverse kinematics problem which is formulated as an optimization task. The errors between the captured trajectories of human arms and the approximated trajectories of robot's arms are minimized. The dynamics capabilities of the joint motors such as limits of joint position and velocity, are imposed on the optimization problem. Several human motions are imitated by the humanoid robots developed in our institute.