Sensorimotor Learning for an Artificial Body Schema on Humanoid Robots

The body schema in humans is a cerebral representation of the current state of the own body. Its plasticity and adaptability allows tool use and the adaption to an ever-changing body. This thesis is about representations for sensorimotor learning on humanoid robots that enable efficient learning for an artificial body schema. Two novel techniques are proposed: The Kinematic Bézier Maps for the representation of kinematic and dynamic models and a decomposition method for accelerating learning.