Kinematics-based approach for robot programming via human arm motion

Abstract

Programming by demonstration approach is becoming an interesting option for programming industrial robots due to the costs involved in the conventional methods for development and maintenance of robot programs. In this approach, the implicit knowledge of human operator is demonstrated through examples that allow robot to learn the skill without being explicitly programmed for every detail. In this paper, a novel approach based on fusion of human arm kinematics with the Kinect-based motion sensing system is proposed for control of robot motion by demonstrating human arm motion. The demonstrated configurations from human workspace are mapped to robot workspace blending concepts of incremental inverse kinematics and Kinect motion sensing system. The proposed approach is implemented for real-time motion control of an industrial robot SCORBOT ER-4u. The performance of the devised strategy is assessed in terms extent of human hand path imitation by the robot end effector in real time. Further, reproducibility and repeatability in path imitation are estimated to ensure the effectiveness of the developed method. The developed kinematic model is validated by comparing its performance with a priori approach. The experimental results indicate that the proposed method is reliable, quick and efficient to teach robot from task demonstrations.