Figure 2. Schematic diagram of the teaching sequence generation process.

As shown on the left, the assistant selected one action randomly while guiding the hands of the robot to move the cubic object. For each action, the assistant started from a posture in which the hands of the robot were open and then the assistant grasped the object by using the hands of the robot. Whenever the object was moved to another position, the assistant released the object and returned the robot hands to the starting posture for a brief period of time. The graph on the right depicts the trajectory of the center of the object captured by the vision sensor.