Human-like coordination motion learning for a redundant dual-arm robot

The motion of a redundant dual-arm robot is subject to dual-arm coordination constraints when performing a coordination task. However, these constraints are usually fixed. To improve the ability of dual arm robots to interact effectively with human beings, it is necessary to obtain the variable dual-arm coordination constraints from the observation of human arm motion. This paper developed a novel redundant dual-arm robot motion learning method based on human-arm coordination characteristics. It can realize the human-like coordination motion of a dual-arm robot in both Cartesian space and joint space. The proposed method was implemented on a real redundant dual-arm robot platform. Experiments involving two tasks, carrying and pouring, were carried out, and the results indicate that the robot can successfully reproduce the demonstrated human-arm motion tasks, and the dual-arm robot has the characteristics of coordinated human-like motion, making robotic dual-arm manipulations more smooth and natural.