ABSTRACT
Future autonomous robots or cognitive agents in complex systems can be achieved by making human–robot interaction more intuitive, which will utilize adopting techniques like imitation learning. This chapter presents a novel cognitive learning framework for human–robot skills transfer, which simultaneously considers the motion and the contact force during the demonstration. We use the DMPs to model the motion and the force to achieve skills generalization. The adaptive admittance model is proposed to simplify the teaching process. To reproduce the motion and the contact force, the hybrid force/motion controller is developed based on the original position controller of the Baxter robot. The NN-based controller is designed to overcome the impact of the unknown payload so that the manipulator is able to track the given motions more accurately. Experiments with the Baxter Robot have verified that the success rate of the task performance is improved, and the robot can perform the force-related task better than the motion-only method by employing the proposed robot learning framework.
