作者: S. Kawamura , F. Miyazaki , S. Arimoto
DOI: 10.1109/21.87060
关键词:
摘要: To make a robot track given desired motion trajectory, learning control scheme is proposed which based on the repeatability of motion. In this obtains by repeating trials (test motion). A merit that input torque pattern generates can be formed without estimating physical parameters dynamics. practice, to allow approach one in each trial, at present trial modified only velocity signal real previous and one. The convergence theoretically proved for linear time-varying mechanical system, an approximate representation nonlinear dynamics vicinity effectiveness demonstrated through actual experiments revolute-type manipulator with three degrees freedom used, trajectory not terms joint-angle coordinates but also task-oriented coordinates. >