Progressive Automation with DMP Synchronization and Variable Stiffness Control

Progressive automation is a method that allows the operator to demonstrate a task a few times until the robot gradually learns to execute it autonomously. In this letter, we combine a novel structure of dynamic movement primitives (DMP) with variable stiffness control, to allow synchronization with the demonstrated motion during the operator's intervention. This structure enables the DMP to speed up or slow down depending on the human demonstration. In addition, we present a variable stiffness controller to change the role of the robot between the follower and the leader based on the tracking error of the robot from the proposed DMP and on the guidance forces. The proposed variable stiffness controller is proved to be passive using energy tanks. The effectiveness of the proposed method is demonstrated experimentally.