可動作業台をもつDDロボットの学習制御 : 制約をもつ作業空間内での位置と力の制御

概要

論文の詳細を見る
In this paper, a learning control algorithm for the motion and force of a direct-drive robot, which has a variable table and six degrees of freedom (d. o. f.), is proposed for a task in constrained work space. The five-d. o. f. relative motion of an end effector attached to the wrist of the robot arm and the internal force between the end effector and the table through an object are achieved by utilizing the redundancy to avoid interference with the object. A hybrid controller for an end effector and a position controller for a robot arm are used. Under the over-constrained conditions, which are imposed by the two tasks stated above, a learning control algorithm using neural networks (NN's)is employed to control an end effector with higher priority than the robot arm. The proposed control algorithm is applied to an experiment on the control of both the position of an end effector and the contact force on the surface of the object while avoiding interference with the object. As a result, it has been shown that the trajectories of robot joints can be generated in real-time computation and a high accuracy of the end effector is obtained after the completion of learning. The effectiveness of the proposed control algorithm is demonstrated by the experiment.
一般社団法人日本機械学会の論文
1995-07-25