Success-Failure Learning for Humanoid: study on bipedal walking
John Nassour · mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) · 2014
This thesis draws on neuroscience and neurophysiology findings towards robotics. Specifically, accounting for the role of the Anterior Cingulate Cortex in human brain in error detection and the involvement of the Orbitofrontal Cortex in coding rewards adaptively, a success-failure learning framework has been proposed. Furthermore, a new multi-layered multi-pattern Central Pattern Generator model that enables a wide range of motor patterns has been realised. Bringing them together, we validated our models on a humanoid robot, enabling it to learn walking under different conditions while dealing with external and unknown disturbances.