The human shoulder joint is constrained by the plural number of muscles and ligaments which surround ball joint, and is driven by the balance of those forces. It is thought that the construction of human shoulder could realize the motion with 3 degrees of freedom as a joint of compact size and light weight in comparison with industrial robot arm which was composed of usual pin joints. Therefore, the authors thought that the new joint mechanism of the robot arm with 3 degrees of freedom in compact size and light weight should be realized in ball joint mechanism driven by wires. In the new joint mechanism, anatomical skeltal structure and muscle arrangement were introduced. The movability of the new mechanism was evaluated by the moment arm produced by wires. And, to improve the deficiency of moment arm, in some flexion-extension and rotation, the corresponding wire position was adjusted on the basis of the evaluation of the moment arm. Thus, the effective movability was ascertained in scheduled movable area on 6 wire's model. Though inverse kinematics should be solved when the new shoulder joint is driven as a robot arm, in this paper this was solved by making it learn on the inverse kinematics in artificial neural network (ANN). To improve ANN learning inverse kinematics of the mechanism precisely, the receptive layer was incorporated in ANN. The joint was properly driven by learned ANN, and the capability as a joint mechanism was demonstrated.
|Number of pages||16|
|Journal||Memoirs of the Faculty of Engineering, Kyushu University|
|Publication status||Published - Dec 2001|
All Science Journal Classification (ASJC) codes
- Atmospheric Science
- Earth and Planetary Sciences(all)
- Management of Technology and Innovation