The Control of Tandem Robot Based on the OpenPose Human Key Points

Abstract

Aiming at the trajectory control of the tandem robots, this paper proposes a scheme for real-time guidance of the joint position of the end of the tandem robots by using the 3D human body skeleton key points. The scheme is based on the algorithm of the OpenPose which can estimate the depth information of the image, generate the 3D human skeleton from the 2D human bone shape by using a monocular camera. And we use the human wrist coordinates in Cartesian coordinates as the control source to guide the end joint position of the tandem robots, guiding the actions of the tandem robots in real time. Due to the use of monocular cameras, our scheme has a higher cost performance than multi-eye vision systems and complex sensor systems. Our scheme can make the robot imitate human behavior in a human-like manner, such as the task of writing, which is full of human intelligence. At the same time we propose a new solutions for intelligent control of robots. In this paper, we use the OpenPose to extract the coordinates of human body key points, analyze and optimize the performance of it in controlling the tandem robot. Finally, we verify the results combined with the robot virtual simulation experiment platform.