Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 02, 2018 - June 05, 2018
Visual servo control uses images obtained by a camera to control robots. This paper focuses on a problem of positioning a object using a robotic manipulator with visual servo control. To perform the positioning task, image-based visual servoing requires visual features which can be extracted from appearance of a target object. Therefore, positioning error is remained for textureless objects, such as industrial parts. To solve this problem, this paper proposes a novel concept of visual servoing named “Active Visual Servoing”. Active visual servoing projects pattern light on the target object using a projector. Thus designing of the projection pattern affects on the positioning error. We theoretically derive the optimal projection pattern which maximizes a cost function for image-based visual servoing. Experimental results show that the proposed active visual servoing achieves less positioning error compared to conventional image-based visual servoing.
