Real time human motion imitation of anthropomorphic dual arm robot based on Cartesian impedance control

This paper presented a real-time human motion imitation approach to control an anthropomorphic dual arm robot by human demonstration. We use the processed positions of human skeleton joints from Kinect sensor as commands directly to control the robot arms by using Cartesian impedance control to follow the human motion without solving inverse kinematics problem. In order to avoid a jerky robot arm motion, we apply an on-line trajectory generator algorithm to obtain a smooth movement trajectory by imposing the limit of velocity and acceleration. Moreover, the self-collision problem has also been considered. When the distance between two parts of body is close enough, a repulsive force will automatically generate to prevent collision. Taking the robot capability and safe issue into account, the output force is restricted to ensure that the action of robot is stable. We demonstrate the feasibility of the approach by implementing the human motion imitation system on a humanoid dual arm robot developed in our lab. The experimental results show that the system is in good practice and flexible enough to imitate various human motions.