Integration of Imitation Learning with Cognitive Control for a Humanoid Robot

Abstract

The aim of dissertation is to develop a cognitive robotic system by integrating imitation learning with cognitive control and evaluate its performance using object manipulation tasks. The robotic imitation learning framework is divided into five components: Behavior Acquisition, Behavior Segmentation, Behavior Dimension Reduction, Behavior Representation, and Behavior Generation. The cognitive control framework was added for robots to switch strategies (both physically and cognitively) to complete tasks. The integrated cognitive system was implemented based on the existing cognitive architecture. Three experiments were conducted to evaluate the integrated system performance. On experiment was carried out on a humanoid robot, named ISAC, and two experiments were carried out in the simulation environment. The experimental results demonstrated that the integrated system satisfies both the design and performance requirements of imitation learning and cognitive control.