Cortical Control of a Dextrous Prosthetic Hand

描述(由申请人提供)：该项目的目标是建立和演示一种由皮质输出控制的拟人化假臂和手。人类的手臂和手大约有30个自由度(点无关的关节旋转)，是非常复杂的机械结构。手是高级进化专业化的一个例子，与双目视觉和两足运动一起，导致了工具的使用--这是人类大脑发育和行为的主要决定因素。然而，人们对自然行为过程中手的神经控制知之甚少。关于主动假手，机器人手控制方面的工作一直很少，直到最近才努力制造出真正准确的功能手复制品。灵长类的伸手抓握行为由四个部分组成--伸展、手形、方向和手指围绕物体的闭合。灵巧性的特征是通过指尖主动产生力量来保持稳定的抓握和/或操纵物体，可以被认为是手行为的一个额外组成部分。鉴于我们成功开发了拟人化的手臂假体，我们希望通过从单个神经元群体中记录的活动来提取实现灵巧假手控制所需的信号。在我们目前的仅手臂控制方案中，我们已经成功地从记录的大脑活动中提取了手臂的速度。为了达到灵巧控制的最终目标，我们还需要控制手腕方向、手形和手指力量的应用。由于这些控制类别中的每一个都是多维的，所以总体控制问题非常困难。我们将使用一些策略来解决这一难题。一个由神经生理学家、工程师、统计学家、机器人专家和心理物理学家组成的跨学科团队已经聚集在一起，他们有很强的合作历史，以开发这个项目所需的部件。该项目将由匹兹堡大学的安德鲁·施瓦茨领导，假肢控制将在那里进行。卡内基梅隆大学的Yoky Matsuoka将制造高度拟人化的机器人和行为操纵器。同样在卡内基梅隆大学工作的罗布·卡斯将开发将神经活动与运动联系起来的提取算法。亚利桑那州立大学的Marco Santello和Stephen Helms-Tillery将使用灵长类动物模型开发行为任务，然后在执行这些任务时记录大脑活动。明尼苏达大学的索赫廷博士将提供详细的心理物理数据，描述受试者使用手指力量操纵物体的方式。哥伦比亚大学的彼得·艾伦将为脑控假手开发自动化机器人抓取和手指放置算法。

DESCRIPTION (provided by applicant): The goal of this project is to build and demonstrate an anthropomorphic prosthetic arm and hand that is controlled by cortical output. The human arm and hand have approximately 30 degrees- of-freedom (dot- independent joint rotations) and are very complex mechanical structures. Hands are an example of an advanced evolutionary specialization, which along with binocular vision and bipedal locomotion, led to tool use- a major determinant of human brain development and behavior. Yet, little is known about the neural control of the hand during natural behavior. Regarding active prosthetic hands, there has been a paucity of work on robot hand control and only recently has there been an effort to make a truly accurate functioning hand replica. Primate reach-to-grasp behavior is characterized by four components- reach, hand shaping, orientation and the closing of the fingers around the object. Dexterity, characterized by the active generation of force through the fingertips to maintain stable grasp and/or to manipulate an object, can be considered as an additional component of hand behavior. Given our success in developing an anthropomorphic arm prosthesis, we expect to extract the signals necessary to achieve dexterous prosthetic hand control using activity recorded from populations of single neurons. In our present arm-only control scheme, we have successfully extracted the velocity of the arm from the recorded brain activity. To reach our ultimate goal of dexterous control, we will also need to control wrist orientation, hand shape and finger force application. Since each of these control categories is multidimensional, the overall control problem is very difficult. We will use a number of strategies to address this difficult problem. An interdisciplinary team of neurophysiologists, engineers, statisticians, robotocists and psychophysicists with a strong history of collaboration has been assembled to develop the pieces needed for this project. The project will be led by Andrew Schwartz at the University of Pittsburgh where the prosthetic control will take place. Yoky Matsuoka at Carnegie Mellon will build the highly anthropomorphic robots and behavioral manipulanda. Rob Kass, also at Carnegie Mellon, will develop the extraction algorithms relating neural activity to movement. Marco Santello and Stephen Helms-Tillery at Arizona State University will develop the behavioral tasks using a primate model and then record cortical activity as these tasks are performed. Dr. Soechting, at the University of Minnesota, will provide detailed psychophysical data describing the way subjects exert finger forces to manipulate objects. Peter Allen, at Columbia, will develop automated robotic grasp and finger placement algorithms for the brain-controlled prosthetic hand.