Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 02, 2018 - June 05, 2018
Requirement for robots has been changed from the work that performed by operators via robots, to the autonomous action took by themselves. Visual servoing is a method to control robots' motion by installing the visual information into the feedback loop. Therefore, it is expected to enable robots to work in the constantly changing environment or unknown environment. However, to recognize target object from images, some degree of predefined knowledge is needed. Therefore, our research group proposed a hand-eye visual servoing system without using them. Models are created firstly from the image projected on the left camera, then after performing inverse projection to 3D space to make search models. Projection of the search models in 3D space are projected to the right camera to determine the position and rotation of target object. In this study, the accuracy of pose estimation of arbitrary-shaped target object by this system is verified.
