Dmp accuracy improvement to facilitate learning from demonstration for industrial cooperative robots

An industrial manipulator with a high degree of freedom is useful in many tasks. However, to control such a manipulator, one requires an understanding of structures, such as kinematics and robot control, and knowledge of various programming techniques. To operate and control the manipulator conveniently, a control method based on direct teaching is widely used to transmit the intention of a person to the manipulator and operate it. In this paper, we propose a strategy to increase the reference trajectory tracking accuracy of DMP (Dynamic Movement Primitives) based on a reference angular acceleration as a Direct Teaching strategy. A simulation is conducted using two trajectories that are unexpected to expect accurate tracking performance by using DMP, and the feasibility of the proposed strategy is verified by comparing the errors.