On the Actuator Requirements for Human-Like Execution of Retargeted Human Motion on Humanoid Robots

Building humanoid robots with properties similar to those of humans in terms of strength, agility, and motion versatility remains a challenge. To achieve human-like motion behavior, an identification of the actuation requirements is crucial. In this paper, we propose a novel approach that automatically calculates the necessary actuator requirements for a given upper-body humanoid robot kinematics performing motions retargeted from human motion data. First, we introduce a unified representation of humanoid upper limb kinematics to decouple the analysis from robot-specific features to allow the comparison between different humanoid robot kinematics. Second, we derive a novel performance index to compare actuator requirements regarding velocity, acceleration, and torque for different kinematics, which are needed for the execution of retargeted human motions. We evaluate our approach by comparing the calculated actuator requirements and performance indices of ten existing humanoid robot kinematics in addition to a new robot design, using 40 recorded human motions of different categories. The results demonstrate the impact of robot kinematics on the joint actuator requirements for achieving human-like motion.