A New Developmental Cognitive Architecture for the Autonomous Acquisition of Sensory-Motor Skills on Humanoid Robots
Erhard Wieser · 2019
This thesis presents a new developmental cognitive architecture that enables a robot to progressively learn increased levels of sensory-motor coordination. The contributions encompass the extension of a predictive coding method and its integration into a framework of biologically-inspired algorithms that form the architecture. The proposed methods are validated on two humanoid robots that differ in their size, sensors, and motors. Experimental results confirm the adaptability and robustness of the methods to sensory-motor deficits attributable to environmental influences or to the robot.