Solving Inverse Kinematics of Humanoid Robot Using A Redundant Tree-shaped Manipulator Model

Abstract

This paper presents a concept for solving inverse kinematics in humanoid robots using a tree-shaped manipulator model. Robot trajectory is given as a set of characteristic points (feet, hands, center of mass) trajectories in a discrete time domain. Next, the motion is described in a local frame related to the robot's right foot. This allows for representing the robot as a tree of serial open-loop redundant manipulators with base in the supporting foot. Stability during motion is provided by the trajectory of center of mass of the entire system that fulfills the zero moment point criterion. Inverse kinematics are solved off-line using first order inverse differential kinematics method. Proposed algorithm uses the robot's redundancy to avoid joint limits and minimize joint torques due to gravity force. The results of the presented method have been tested on half meter tall robot prototype designed by the authors.