2004 年 70 巻 691 号 p. 751-758
A flexible tool held by the end joint of a robot will be deformed when the robot is used to complete a contact task. Without establishing a deformation model of the flexible tool, the end position of the flexible tool can not be obtained from computing robot kinematics. In this paper, a position and force hybrid control method using the visual information has been proposed for a 6DOF robot to complete a contact task. To detect the end position of the flexible tool, two CCD cameras are used and a real time image processing algorithm is developed. An advantage of the proposed method is that establishing a deformation model of the flexible tool is not necessary. An online learning neural network is also applied to a position/force hybrid control system to decrease tracing error of position. Using the proposed method, a flexible tool held by a 6 DOF robot hand was used in experiments of tracing a given curve with a specified force. The experiments have been made successfully and the results demonstrate the effectiveness of the proposed method.