[This article belongs to Volume - 53, Issue - 05]
Gongcheng Kexue Yu Jishu/Advanced Engineering Science
Journal ID : AES-14-11-2021-87

Title : Human-like Motion Planning of Anthropomorphic Arm Based on Movement Primitive
WEI Yuan,

Abstract :

As the basic part of motion of humanoid robots, human-like motion planning of the anthropomorphic arm is always one of the research hotspots and difficulties. A novel human-like motion planning method based on movement primitives was proposed. This method can satisfy the feature of arm motion and improve the accuracy. Firstly, the arm structure was decoupled and the arm model was built to express different arm movements. The methods of extraction and connection about movement primitives were established. The mapping relations between the arm models and inverse kinematic (IK) solutions were established. Meanwhile, a motion framework was proposed. The joint trajectories of a certain platform can be generated to accomplish required tasks with this motion framework. Secondly, according to the motion features of different movement primitives, the associated Human Performance Measures for different movement primitives were constructed to solve the IK problem. Finally, the proposed method was verified by the similarity experiment and the human-like movement experiment for the general motion of humanoid robot NAO. In the similarity experiment, the robot NAO generated the human-like movements with the proposed method. The motion data were compared with the real data generated by humans. All the errors were less than 1 cm, which satisfied the accuracy requirements of human-like movements. In the human-like movement experiment, the robot NAO performed a human-like arm movement with the proposed method. The proposed method was also compared with the minimum total potential energy method and the last norm algorithm. Using the proposed method, 7% and 58 % increase in similarity were achieved respectively, compared with those two methods. With the proposed method, the complex motion models were decoupled into different simple sub-movements and the classification of the movements reduced the calculation amount. The experiments proved that the anthropomorphic arm can generate human-like movements accurately.