Abstract
This paper clarifies robust task strategy of human by replaying the remote teaching trajectories on a robot which performs a task with environmental variations. As a case study, flexible part insertion task is selected and the tightness of insertion is intentionally varied. We proposed a novel trajectory-switching method and confirm the effectiveness of the proposed method by experiments. In the proposed method, the teaching trajectories are switched based on the sensory information. According to the experimental results, the trajectory generated by the proposed method is more robust than simply replaying the human trajectories. It is also shown that lowering the gain along with the insertion direction can realize human-like search behavior and switching the trajectory by sensory information can save spent time and minimize the force exerted on the flexible part. These results show that feedback using external sensor is effective when environmental variation is substantial.
