Direct Demonstration-Based Imitation Learning and Control for Writing Task of Robot Manipulator

Abstract

In this paper, we propose an imitation learning method based on a direct demonstration of robot manipulators. To track the desired position and force, we designed an impedance controller. As a result of imitation learning, the robot can be acted as intended even if the initial position is different, and be able to perform a writing task well even if a different contact force is applied to the changing environment. We propose Long Short-Term Memory (LSTM)-based imitation learning method through the demonstration data. Finally, the proposed method was verified by applying the writing task with the actual industrial robot manipulator that acts as the expert's intention for the direct demonstration.