Human and humanoid identification from base-link dynamics

The knowledge of the inertial parameters of a dynamic systems is crucial to understand, simulate and eventually control it. Identification theory in robotics is well formulated and often used however it requires the knowledge of the actuation torque. When considering humanoid system: robots or humans identification is seldom used as the actuation torque is difficult to measure. We have developed an identification method for bipedal systems that only requires the contact forces measurements and the joint angle. It applies to humanoid robots and humans. In this paper we present the overall identification method; its efficiency is then illustrated by experimental results obtained on a robot and on a human.