Study of Visual Servoing for Robot Hand Control

在本研究中，我们建立了基于视觉伺服技术的机械手控制方法，并构建了实验系统来验证其性能。作为研究结果，我们提出了一种描述动态图像与摄像机运动之间关系的总体方案。对于一个特定的物体，图像数据和摄像机位置的整体组合形成了产品空间。相机不能获得任何任意的图像，因此在这个空间内，相机姿势和所获得的图像的可能组合应该被约束在较低维度的超曲面上。我们的基本思想是将视觉伺服解释为在该曲面上找到通向给定目标图像的路径，我们的方法是分析该曲面的性质，并利用其微分或切向性质来进行通常的伺服。对于这种方法，图像信息的维度成为关键问题。我们提出用主成分分析，用K-L展开来表示具有少量“特征图像”的图像。在研究报告中，我们首先描述了该曲面的一个法矢与所谓的交互矩阵有关，并通过初步实验证实了我们的基本思想视觉伺服的可行性。然后，我们给出了机器人手臂移动到目标位置的法矢量的动态估计，在目标位置将获得给定的目标图像。我们还考虑了构造特征空间(特征图像空间)来有效地表示图像，并在控制中加速收敛。基于该方法的视觉伺服实验结果表明了该方法的可行性和适用性。

In this research, we established a methodology for robot hand control by visual servoing technique, and constructed experimetnal system to verify its performances.As the research result, we present a general scheme to represent the relation between dynamic images and camera motion. We also propose its application to visual servoing.For a specific objects, whole combination of image data and camera position form the product space. The camera cannot obtain any arbitrary image, so that within this space the possible combination of the camera pose and the obtained image should be constrained on a lower dimensionla hyper surface. Our basic idea is that the visual servoing is interpreted as to find a path on this surface leading to a given goal image.Our approach is to analyze the properties of this surface, and utilize its differential or tangential property for vusual servoing. For this approach, the dimension of the image information becomes key problem. We propose to use the principal component analysis and to represent images with a composition of small number of "eigenimages" by using K-L (Karhunen-Loeve) expansion.In the report of the research, we describe, first, that a normal vector of this surface is related to the so-called Interation Matrix, and confirm the deasibility of our basic idea visual servoing with preliminary experiments. Then, we present a dynamical estimation of the normal vectors to move robot arm mouting a camera to a goal position where a given goal image will be obtained. We also consider the construction of the eigen space (the eigenimage space) to represent images efficietly and to speed-up convergence in the control. Experimental results of visual servoing with proposed method show the feasibility and applicability of our newly proposed approach.