Abstract: In order to make anthropomorphic mechanical arms generate human-like movements
accurately, a novel human-like motion planning method is proposed, which combines the trigger
conditions and hierarchical planning strategy (HPS). The method decomposes the complete arm
movements into a set of different motion processes, each of which has corresponding planning
hierarchies. The anthropomorphic mechanical arms reveal different characteristics in different
planning hierarchies. The motion models and posture prediction indicators in varying planning
hierarchies are built based on the respective characteristics to predict the postures of anthropomorphic
mechanical arms. The experiment is acted on humanoid robot NAO as the platform, and then the
prediction results of static and dynamic arm postures is performed by the proposed method and the
minimum total potential energy (MTPE) are compared. In addition, the prediction results are compared with the real arm motion data collected by motion capture system (OptiTrack). The
experimental results show that the errors of static and dynamic posture prediction of proposed method
could be reduced, and the anthropomorphic mechanical arms can generate the human-like movements
accurately through the proposed method.

Key words: anthropomorphic mechanical arms, human-like movements, hierarchical planning strategy, arm posture prediction