手指不够用？可穿戴式机器人帮你忙

作者：R. Colin Johnson 你曾经尝试以单手打开瓶盖或信封封口吗？或是其他需要用两只手来做的事？现在，如果你穿戴上美国麻省理工学院(MIT)开发的新式机器手腕，这些工作要以单手来做，可说是轻而易举。尽管这款可穿戴式机器人仍处于原型阶段，但已为具备“协同作用”(synergism)的辅助机器人技术建立新的典范。 “众所周知，人类的手部动作是由机体间的协同所控制──这就是肌肉群经由单一控制讯号启动的概念，”MIT博士候选人Faye Wu指出，“我们想把这种协同控制的概念延伸至可穿戴式机器人研究上。”

MIT博士候选人Faye Wu表示，在靠近拇指和小指旁的两个机械手指可依照人类手指的动作，协助用户以单手完成通常需要两只手才能进行的任务。
Source：MITVkNesmc

为此，可穿戴的手腕机器人配备一副手套，可测量每个手指的角度和方向，然后将控制讯号传送至其他手指以实现协同反应。例如，在抓取一个大型对象时，一般的手指只需抓住顶部，让较长的机器人手指抓握住底部。 “当我们用餐时手拿刀叉或是驾驶汽车，如果长时间使用这些工具，就会感觉他们像是身体的一部份。因此，这就是我们试图用在机器人的想法。如果人们有了额外的手 指或手臂，而且能与其良好互动，就会感觉他们就像是身体的一部份延伸，”MIT机械工程教授Harry Asada说。 Asada 的研究小组正开发新的协同算法，让额外的机器人附件协助难以用单手完成的任务，未来，还可能进一步协助人类顺利完成以双手或双臂难以完成的工作。此外， 并不需要给这些机器手人清楚的指令，它就能依照与人类手指协同作业的方式，配合手指的动作与需要共同抓取对象。

如果有机器手指帮忙握住杯子的话，人们只要以单手就能轻松搅拌饮料了。
Source：MITVkNesmc

MIT 将这种额外的机器手指称为“赘生”手指，因为它们是由连接在一起的致动器组成，与人类的手指头共同发挥同等强度的力量。研究人员们经由探索人类抓取对象的方式发现，手指的这种协同作用只需利用两种基本算法──使手指并拢，以及向内弯曲取物。大多数的对象抓取动作都只是这两种基本模式的组合。 研究人员们希望，这些基本模式也适用于其他肢体发挥“生物－机器”协同效应，不过，研究人员的下一步在于为这项组合添加动力。例如，如何让机器手指知道要施 加更多力量，才能拾取沉重或光滑的小对象？其次，不同的人在抓取对象的施力与方式也有所不同，因而必须为不同的技巧编译工具库，使机器人可针对协助对象从 工具库中挑选正确的算法，学习辨识以及更有效地进行协助。 该研究小组还致力于缩小机器手指机制，使其能够内建于手环中，在需要时才弹出多余手指来帮忙，在不用时又能随即内折妥善收藏。 本文授权编译自EE Times，版权所有，谢绝转载 编译：Susan Hong 参考英文原文：MIT Wrist-Robot Adds Extra Fingers，by R. Colin Johnson {pagination} MIT Wrist-Robot Adds Extra Fingers One hand is better than two R. Colin Johnson PORTLAND, Ore. — Ever try to get the lid off a jar with one hand, or open an envelope, or 1,000 other two-handed tasks? Well now you can, that is if you are wearing the new wrist-robot from the Massachusetts Institute of Technology (MIT). Still in the prototype stage, the wearable robot nevertheless offers a new paradigm in assistive robotics called "synergism." "It is well known that the motion of the human hand is controlled by synergy -- which is the idea that groups of muscles are activated together by a single control signal," said MIT doctoral candidate Faye Wu in a video about her presentation this week at Robotics: Science and Systems 2014 (July 22 through July 24, Berkeley, Calif.) "We want to extend the synergy-based control to wearable robots." MIT doctoral candidate Faye Wu shows that by following the motions of the fingers the two robotic fingers adjacent to thumb and pinky can help users perform tasks with one hand that would ordinarily take two. (Source: MIT) To do so, the wrist worn robot has a glove attached that measures the angle and orientation of each finger, then passes a control signal to the extra fingers to react synergistically. For instance, if a large object is being picked up, the normal fingers only need grasp its top while the longer robotic fingers grab its bottom portion. Because the user does not have to give explicit control signals to the new digits, the learning curve for working with extra fingers for assistance if not very steep. "We use a knife and fork, or we drive our car, and if you use these tools for a long time, they feel like and extension of your body. So that is exactly what we are trying to do with robotics. If you could have extra fingers or a extra arm and communicate with them very well, then you've got to feel too that they are an extension of your body," said Harry Asada, the Ford professor of engineering in MIT's Department of Mechanical Engineering. Asada's group is creating synergistic algorithms that allow extra robotic appendages to follow-along helping out with tasks that would be too difficult for one-hand, and someday an extra hand for tasks difficult with two-hands, or even an extra arm for tasks too difficult for two-arms. No explicit commands will be given to the robot, it just follows along helping out synergistically the way all your fingers cooperate to grasp something. Stirring a drink with one hand is easy if the robotic fingers hold the cup. (Source: MIT) Officially, the extra robotic fingers are called "supernumerary" because they consist of actuators linked together to exert forces of equal strength to actual human fingers. Through studying the way that humans grasp objects, the researchers found that synergistically there are only two basic algorithms used -- bringing the fingers together and twisting them inwards. Grasping most objects is just a combination of these two general patterns. The researchers hope that the bio-mechanical synergy of the other limbs will also follow a few basic patterns, but right now their next step is to add force into the mix. For instance, how do the fingers know to exert more force on small objects that are heavy or slippery? Secondly, different people use slightly different approaches to grasping objects, making the compiling of a library of those techniques a goal -- so that the robot can learn to recognize and assist more efficiently by picking the correct algorithm from the library for the person they are helping. The group is also working to downsize the mechanism so that it could fit inside a bracelet, with the extra fingers popping out when necessary, but then folding back up when no longer needed.