学習に基づいた2自由度マニピュレータの接触動作制御の実験

概要

論文の詳細を見る
In recent years, as robotic manipulators have become more widely used in real tasks such as grinding, deburring and assembly, the contact motion control of robotic manipulators has received considerable attention and is now regarded as an essential capability of dextrous manipulators. Previously, we proposed a contact-motion control method based on an iterative learning process for one-degree-of-freedom manipulators, and extended it to the case of two-degree-of-freedom ones. It is a kind of feedforward control scheme that is intended to improve the transient part of the contact force response by iteratively modifying the feedforward input and the system initial conditions, including the approaching velocity of the manipulator to the environment. In this study firstly, we experimentally verify the effectiveness of the previously proposed algorithm for two-degree-of-freedom manipulator cases and discuss some implementation issues. Secondly, we verify the relation-ship between the learning convergence speed and the learning robustness. Thirdly, it is shown that the approaching velocity to the environment, which is adequately adjusted by the learning process, can be increased by choosing an appropriate contact configuration which reduces the impact force. Finally, we address a frequency-domain interpretation of this time-domain learning method.
社団法人日本機械学会の論文
1996-04-25