Human-like task planning for a dual-arm robot based on motion primitives

Task planning has an important effect on the anthropomorphic characteristics of the motion of the humanoid robot. In this paper, human-like task planning for a dual-arm robot is discussed. A bottom-up task planning approach is proposed based on the motion primitives which imply the features of human motion, in order to obtain a human-like motion fundamentally. Unlike the traditional methods which decompose the task into a sequence of well-designed trajectories of the end-effector in task space, we ignore the trajectory planning temporarily, and assemble the motion primitives until the task is accomplished. Therefore, the human-like posture is guaranteed all the time, and there is no need to spend much time in solving the inverse kinematics of redundant manipulators and performing an optimization for a human-like posture. Two basic and specific tasks are analyzed based on the proposed approach. Finally, simulations are performed to verify the validation and feasibility, and the characteristics of the planned motion are discussed.