The Visual Control of Grasping

人类的手是一项非凡的工程。作为人类，我们可以伸出手，精确地抓住物体——比任何机器人都要好得多。使这一切成为可能的原因之一是我们敏锐的视觉。加拿大西安大略大学（University of Western Ontario）大脑与精神研究中心的梅尔·古德尔（Mel Goodale）博士和他的同事们正在研究我们的视觉系统用来引导手伸出并抓住物体时所使用的不同类型的信息，以及之前对物体的体验如何与当前的视觉信息相互作用。为了研究手是如何运动的，古德尔博士使用特殊的红外敏感摄像机来跟踪附着在指尖、手和手腕上的微小红外光。然后，这些相机的数据可以用来重建手和手指在抓取运动展开时的轨迹。古德尔博士实验室之前的工作表明，来自两只眼睛的信息——双目视觉——在这些运动的指导中起着重要作用。但也有其他信息来源，包括我们四处走动时世界在我们眼中的运动，以及一大堆被称为“深度图像线索”的线索。几个世纪以来，艺术家们一直利用透视、熟悉的大小、遮挡、阴影和纹理渐变等线索，在二维画布上描绘三维世界。Goodale博士和他的团队正在探索所有这些不同的线索在计划和控制手的抓取运动中的作用，以及这些线索如何与先前的经验和对世界的认识相互作用。他还研究了我们是如何应对视觉错觉的——他已经表明，欺骗我们眼睛的错觉很少能欺骗我们的手！最近，古德尔博士已经开始使用脑成像（功能性磁共振成像，简称fMRI）来探索大脑如何计算我们试图抓住的物体的大小。这项工作不仅提供了关于如何使用视觉来控制人手的基本信息，而且还有助于设计更好的控制机械手的系统。这并不是生物学第一次为工程学提供一些有用的设计经验。

The human hand is a remarkable piece of engineering. As humans, we can reach out and grasp objects with exquisite accuracy - far better than any robot. One of the things that makes this possible is our keen sense of sight. Dr. Mel Goodale and his colleagues at the Centre for Brain and Mind at the University of Western Ontario are studying the different kinds of information that our visual system uses to guide the hand as it reaches out and grasps an object - and how previous experience with the object interacts with current visual information. To study how the hand is moving, Dr. Goodale uses special infra-red sensitive cameras to track tiny infra-red lights that have been attached to the fingertips, hand, and wrist. The data from these cameras can then be used to reconstruct the trajectory of the hand and fingers as the grasping movement unfolds. Previous work in Dr. Goodale's lab has shown that information from the two eyes - binocular vision - play an important role in the guidance of these movements. But there are other sources of information as well, including the motion of the world on our eyes as we move around and a whole host of cues called 'pictorial cues to depth'. These latter cues - perspective, familiar size, occlusion, shadows, and texture gradients - have been exploited by artists for centuries to depict a three-dimensional world on a two-dimensional canvas. Dr. Goodale and his team are exploring the role of all these different cues in the planning and control of grasping movements of the hand - and how these cues interact with previous experience and knowledge about the world. He is also looking at how we cope with visual illusions - and has already shown that illusions that deceive our eyes rarely fool our grasping hand! Recently, Dr. Goodale has begun to use brain imaging (functional magnetic resonance imaging, or fMRI) to explore how the brain computes the size of objects that we are trying to grasp. This work is not only providing basic information about how vision is used to control the human hand but is also helping in the design of better systems for controlling robot hands. This is not the first time that biology has provided engineering with some useful lessons in design.