Capture-Point Based Controllers for Robot Bipedal Locomotion: Analysis and Implementation on the iCub Platform.

Bipedal locomotion is still an open problem in the humanoid community. This thesis proposes a complete controller architecture that allows a humanoid robot to walk on a rigid and flat terrain. First, two controllers are developed to ensure the tracking of the desired divergent component of motion and the zero momentum point. Then, an inverse kinematics algorithm is used to evaluate the desired joint positions. As an additional contribution, a task-based velocity controller has been developed. This can be used, instead of the inverse kinematics algorithm, to reduce the computational effort. Furthermore, since the task-based velocity controller retrieves the robot signal to evaluate the desired joint values it ensures a better tracking of the desired Cartesian tasks, i.e. the desired CoM and feet position and orientation. Experiments in the Gazebo simulator and on the iCub humanoid robot validate the proposed architecture