Dynamic imitation of human motion for humanoid robot

Abstract

This study uses DARwIn-OP as the experimental robot to perform human body imitation tasks. The research integrates Kinect v2 to capture the human skeleton joints to recognize human motion. The joints data are transmitted through the wireless network to the robot to implement the imitation tasks. The joints data can classify into upper and lower body parts. In upper parts, the related angles of the robot's arms are directly calculated by the coordinates of the human body joints to control the motion of robot's arms. In the lower parts, the ratio between human leg and robot leg is measured to adjust the discrepancy between human and robot. Hence, the predicted end point position of the robot is obtained, and we apply the inverse kinematics to calculate the angles of the motors of the robot. The system measures the center of mass of the human to adjust the balance of the robot to perform human motion imitation correctly and simultaneously.