Motion control based on modification of the Jacobian map between the muscle space and work space with musculoskeletal humanoid

Abstract

Many robots need body calibrations when their physical parameters are changed. This is because their motions are created with absolute variables such as links position and orientation. In this paper, we propose two methods using the musculoskeletal humanoid Kojiro. First, we apply a body control method with a Jacobian map between muscle space and work space, which is made with relative variables. Second, we apply a pose estimation method by making robots observe its own motion using its vision to get relations between the pose and the environment. By these methods, we achieved the motion control without calibrations even when the physical parameters of the robots are changed.