Motion generation of redundant musculoskeletal humanoid based on robot-model error compensation by muscle load sharing and interactive control device

Musculoskeletal humanoid has robot-model errors because of muscle elasticity or elongation. This leads to low accuracy and repeatability in joint position control. In particular, this is a serious problem during high-load motions. In this paper, we propose and implement its compensation methods for generating motions in the presence of robot-model errors. One of the control methods presented is a muscle-load sharing method utilizing redundant muscle arrangement which is specific in musculoskeletal humanoid. The other is an interactive control system enabling interactive joint position modification by using human physical or visual information. Finally, we tested several motions by integrating the two systems.