中国《无极科技》机械手亮相惊艳全世界 众网友:我有一个大胆的想法！

lego_batman
It's a nice piece of engineering.
This is going to be a guess, but having a look I expect very high speed motors inside the phalanges and large reductions with a bit of serial compliance to prevent shock loading of drivetrains. High gear ratio is the only way to get good strength in a small form factor with electric motors, input and output encoders to me is further evidence of this. Input encoders are necessary for FOC control of the brushless dc motors, though the didn't say why type of input encoders so I wouldn't confidently say that, but it would allow good static force control for the finger tips.
Oh yeah... And custom absolutely everything.
Fundamentally I expect this to be expensive and not particularly robust, hands and dexterity are perhaps the hardest thing to engineer of all aspects of a humanoid robot.
 
“这绝对是工程学的杰作。
我只能猜，但仔细一看，我认为指节内部可能装有超高转速电机，配合大减速比传动机构，再加上一点串联柔顺机构，以避免传动系统承受冲击载荷。要在小尺寸内用电机实现高负载能力，大传动比是唯一的办法。而输入编码器和输出编码器的存在，进一步印证了这一点。输入编码器是无刷直流电机实现磁场定向控制的必要部件，不过他们没说具体用的是什么类型的输入编码器，所以这一点我不能完全确定，但它确实能让指尖实现良好的静态力控制。
哦对了…… 还有一点，所有部件肯定都是定制的。
从根本上说，我觉得这玩意肯定会很贵，而且耐用性可能不太行。在类人机器人的所有部件中，手部和灵活操作能力或许是最难设计的。”
 
 
lorepieri
Can you elaborate on not being robust? How many open-close cycles before breaking?
 
你所说的‘耐用性不行’是啥意思？是指它能进行多少次抓握才会坏？
 
 
Ronny_Jotten
From their brochure:
Its reliability leads the industry, having undergone million-cycle lifespan validation internally and undergoes rigorous 300,000-cycle aging tests before shipment. ... Empty-load gripping lifespan: > 300,000 cycles - Maintains stable performance during prolonged, frequent use
 
根据他们的产品手册：产品可靠性行业领先，已完成百万次循环寿命内测验证，出厂前还通过严格的30万次循环老化测试…… 空载抓取寿命：＞30 万次循环 —— 可在长时间、高频使用中保持稳定性能。
 
 
lego_batman
Yeah it's just speculation on my behalf having designed these kinds of things. Gears, especially tiny little ones wear out, and cyclic loading in reversed directions under potentially highly varying loads is like the worst condition for them. Add that to the fact that there is 20 DoF and you've got a recipe for finicky real world reliability.
 
我之前说的只是基于经验的推测， 我设计过这类产品。齿轮，尤其是微型齿轮，这东西是很容易磨损的；而在负载变化极大的情况下，齿轮还会承受反向循环载荷，这对齿轮来说是最恶劣的工况。再加上这款机器人手有 20 个自由度，综合来看，它在实际使用中的可靠性可能会比较不稳定。
 
 
Clean_Firefighter922
Looks like it might be Solenoids?
 
它会不会用的是电磁铁？
 
 
qTHqq
It seems the motors ARE the bones of the finger
Feels like that to me too. Like a tiny serial robot arm
I also have to presume the batteries and motor controllers are either in palm or outside of the hand?
Probably
I have to presume the downgearing is built into the custom motor
Yeah, I expect so
Also I was told that large diameter pancake shaped BLDC motors have high torque and narrow motors like this are low torque high speed. So the downgearing would have to be a large gear ratio I thought
There are pros and cons to every type of motor. Inrunner/conventional motors are a lot easier to cool because the stator is outside where you can heatsink it.
It is true that maximizing gap radius is good but you also have to watch the inertia of the motor.
And high gear ratios are not good for sensitive, torque controlled manipulation. First off they're not as backdriveable just because of inefficiencies of the high gear ratio.
Also, even if they're efficient and backdriveable at low accelerations of the end-effector, high ratios can be bad because the effective inertia of the manipulator chain when pushed on by external forces to rotate the motor rotor scales like the SQUARE of the gear ratio.
A small diameter rotor helps with this too because its real rotational inertia is smaller than a big outrunner rotor as the rotating moving part.
When something slams into the finger on this thing like a baseball it's catching you don't want that to slam hard against the N-squared amplified rotary inertia of the motor rotor. You want it to squish against the magnetic field.
So I'd look to use a modest gear ratio... Enough torque to get the job done but low enough to be nicely backdriveable avoid excessive reflected inertia.
The past whatever 10 years or so have had a flurry of papers and products leveraging this kind of system-optimized "quasi-direct-drive" (QDD) manipulation. I think it's a weird name for gear ratios of O(10) and I think we should just call it bidirectional-dynamics-optimized-actuation or something.
BDOA? IDK maybe I need to workshop that.
But there's been a lot of work and a good emerging design philosophy and sorta-recipes for actuators for manipulators that actually need to move at high speed and make constant contact with a relatively unstructured environment.
The best examples do not use plain language or rules of thumb for design shortcuts. They go crazy with the physics and math and measurements of each and every subsystem so you can push everything a little more to the limit.
Like how Ben Katz went through and considered the magnetic saturation of the motors he was using for backdriveable robot dog legs.
I don't know anything about this hand to know exactly what's going on or how real it is.
You can build small systems that look really great but are destroying themselves too.
Still, China is getting really good at precision machining and I could see the designer following a QDD pathway and really mathematically optimizing things to deliver something pretty new.
Take a look at Duatic robot arms for a similar design philosophy applied to a large arm. Also fairly crazy looking compared to more conventional arms.
One of the tricks to all this is you're inevitably giving up kinematic precision in favor of good dynamic performance but that's exactly how humans work. Sloppy and slack and flexy AF but you wrap that in a good visual servoing loop with a sophisticated physics-aware world model and force feedback and we can do anything, including many quite precise things we could never do under kinematic control.
It's been hard to apply to robots but we're getting better at it.
 
看起来确实是用电机作为手指骨骼控制结构
我也有同感，就像微型串联机械臂
我推测电池和电机控制器应该位于手掌内部或手部外侧？大概率如此
减速机构应该是集成在定制电机内部，反正我是这么认为的
另外我了解到大直径扁平无刷直流电机扭矩较高，而这种细长型电机属于低扭矩高速类型。因此我认为减速机构需要采用大传动比设计，每种电机都有各自的优缺点。
内转子或传统电机更易冷却，因为定子在外围容易安装散热片。
增大间隙半径有利于提升性能’这话没错，但同时也要考虑电机的转动惯量。
高传动比不利于灵敏的扭矩控制操作：首先高传动比本身的效率损失会降低反向驱动能力；
其次即使它们在末端执行器低加速度运行时效率良好且可反向驱动，高传动比仍可能带来问题——当外部力作用导致电机转子旋转时，机械臂传动链的等效转动惯量会随传动比的平方增长。
小直径转子在这方面具有优势，因为作为旋转部件，其实际转动惯量小于大型外转子电机。
要是有个棒球之类的东西撞到这款机器人手的手指上，你肯定不希望冲击力通过‘转动惯量 × 传动比平方’的放大效应传递到电机转子上，而是希望冲击力被磁场化解掉。
因此我会选择适中传动比——既保证足够任务扭矩，又确保良好的反向驱动特性，避免过大的反射惯量。
过去十年间涌现了大量基于这种系统优化"准直接驱动"操作的论文和产品。我认为用这个名称描述10倍量级传动比有些奇怪，我更倾向于叫它‘双向动力学优化驱动’之类的名称。
BDOA？不知道，或许还得再琢磨一下命名。
不过目前这一领域已经有了大量研究，形成了一套成熟的设计理念和‘配方’，可用于设计能高速运动、且需与非结构化环境持续接触的机械臂驱动器。
最优秀的设计不会依赖通俗的经验法则或捷径，而是会深入研究每个子系统的物理原理、数学模型和实测数据，从而将各部件的性能推向极限。比如本・卡茨（Ben Katz）在设计可反向驱动的机器狗腿部时，就专门研究了所用电机的磁饱和特性。
我对这款机器人手的具体原理和实际性能一无所知。毕竟有些小型系统看起来性能惊艳，实则内部损耗很大，很容易损坏。
但中国在精密制造方面已经非常厉害，我相信设计者完全有可能采用准直接驱动（QDD）方案，并通过数学优化，做出真正创新的产品。
大家可以看看 Duatic 的机械臂 —— 它的设计理念和这款机器人手类似，只是应用在大型机械臂上。和传统机械臂相比，它的外观也相当独特。
这类设计的关键技巧之一，是在运动精度上做出妥协，以换取更好的动态性能 —— 而人类的手部运作方式正是如此。人类的手虽然极度松散、迟钝且柔性过大，但通过精准的视觉伺服控制、复杂的物理感知世界模型和力反馈，我们几乎能完成任何动作，包括很多单靠运动控制无法实现的高精度操作。
要让机器人实现这一点并不容易，但我们正在不断进步。”
 
 
Antique-Gur-2132
New era of robot hands
 
机器人手的新时代来了。
 
 
humanoiddoc
Chinese innovation again.. Who will use shadow robotics hand now?
 
中国又一次展现了创新实力…… 现在谁还会用影子机器人公司（Shadow Robotics）的机械手呢？
 
 
oiratey
I'm skeptical, many of the details appear too humanlike—it's just like those humanoid demonstration videos made by F&T I watched this year
 
我有些不敢信，因为好多细节看起来太像人了 —— 就像我今年看过的那些 F&T 公司制作的人形机器人演示视频一样。
 
-----------------
 
@abnormaltaro
i have never seen a robot hand this articulate while being this compact. this is some revolutionary stuff!
 
我从没见过这么小巧的机器人手，居然有着这么高的灵活度。这绝对是革命性的技术！
 
 
@BabyBear-tn2wh
This hand is better than mine?
 
这只手难道比我的还好用？
 
@tesBye
 @BabyBear-tn2wh  This hand can work 24 hours a day without a break. How about yours？
 
这只手能 24 小时不间断工作，你的能吗？”
 
@freddybell8328
Optimus V3 hand is better
 
特斯拉Optimus V3的手不比这好多了。
 
@miroslavhavlas1443
 @freddybell8328  How do you know? Have you seen it?
 
你怎么知道？你见过？”
 
 
@JinKee
Forget the hand, i want those servos!
 
别管这只手了，我想要那些伺服电机！
 
 
@ChristoBoshoff
Dexterous, indestructible, and doesn't feel pain. I can't wait to be choked to death by such an amazingly engineered piece of technology.
 
灵活、耐造，还不知疼痛。真等不及被这么个神奇高科技给掐死了。
 
 
@timbehrens9678
Now add the full-area tactile sensorics and we are talking.
 
现在要是加上全域触觉传感系统，那就真的厉害了。
 
 
@tsrmmercy836
no need, as long as it can grasp a cylindrical tube and shift vertically rapidly I’m buying it and fitting it onto whatever.
 
没必要，它只要能握住管状物体并能上下快速撸动，我肯定会买，然后装到某个东西上。
 
@jooeyyjn
The ultimate test will be for a robot to successfully pull off a magic trick using slight of hand.
 
对机器人手的终极测试，将是看它能否用手法技巧成功地变魔术。
 
 
@family66478
I think the most difficult things for robot's hand was to slove the knot of rope like human
 
我觉得机器人手最难做到的事，就是像人类一样解绳结了
 
 
@--Wingman--
Remember Terminator it all started with a hand
 
还记得《终结者》吗？一切都是从一只手开始的
 
@xiaozhong5198
Wuji Tech is a Chinese company that has developed a highly advanced, human-sized robotic hand known for its exceptional dexterity
 
无极科技是一家中国公司，研发出了一款高度先进的真人大小机器人手，以其极高的灵活度广受认可。”
 
 
@RupertBreheny
Nice progress. Humanoid robots are only as good as their hands, so I'm happy to see this,
 
进展不错。类人机器人的水平取决于其手部性能，所以看到这样的成果我很开心。
 
@artsyomni
I’m impressed. I’ll be even more impressed if you can hand it a bunch of keys on a key ring and it can isolate a single key.
 
太惊艳了。要是能把一串挂在钥匙圈上的钥匙递给它，而它能单独挑出其中一把，那会让我更震惊。
 
 
@krit9125
This needs to be a prosthetic!!
 
这完全可以做成义手啊！！
 
 
@snitox
Imagine getting a haircut remotely at your home.
 
想象一下能在家（给他人）远程理发的场景。
 
 
@teocharger
Waiting for the robot revolution from Japan since early 2000's... Guess what!?
It comes from China ❤
 
从 21 世纪初就等着日本来场机器人革命…… 你猜猜怎么着！？这场革命居然是中国带来的！
 
 
@foxtraner
Japn has been left in the dust
 
日本已被远远甩在后面了。
 
@nicosantu1402
They close their eyes too much!!
 
他们（日本）太固步自封了！
 
@村田让
After being trounced by the United States in the 1980s, Japanese companies became deeply conservative, failing to effectively apply various new technologies. Corporate executives were often stale and stubborn, with no desire for innovation. This led to a gradual decline in Japanese IT technology starting in the 1990s, with the latest generation of AI technology relegated to the Stone Age.
 
日本在80年代被美国打垮之后，日本企业变得极度保守，未能有效应用各类新技术。日企高管思想僵化、固执己见，毫无创新意愿。这导致日本 IT 技术从90 年代开始逐渐衰落，最新一代的人工智能技术更是被规划到了石器时代水平。
 
@lexxxx2
The Japanese are lagging behind in big data, software, and artificial intelligence. Robots that rely solely on hardware are doomed to fail.
 
日本在大数据、软件和人工智能领域都处于落后地位。仅靠硬件的机器人注定失败。
 
@hugoelec
Almost ready for  iPhone assembly
 
这都快能组装苹果手机了！
 
@Atmos_Glitch
Though it can move in impossible ways a human couldn't, for the most part it's highly convincing! If you were to tell me that it was an actual human hand in that glove I'd believe you!
 
虽然它的动作方式超越了人类生理极限，但整体观感却极具说服力！如果有人说那里面套着是真人的手，我肯定相信！
 
 
@noviceartisan
It may as well be in this video, it's 1-2-1 controlled by a human wearing a sensor glove that just copies the motions exactly, which is very impressive, just it is effectivley the saem as a human hand in that glove lol
 
“这段视频里的情况可能就是如此 —— 它是由戴着传感手套的人类‘一对一’操控的，只是精准复制动作而已。这确实很厉害，但本质上和‘手套里套着真人手’没区别，哈哈。
 
@thsGuy
@noviceartisan  I find it more impressive that it can be operated like that. Give it to an AI and… wait maybe don’t
但我觉得‘能这样操控’反而更惊艳。要是把它交给 AI…… 等等，还是算了吧。
 
 
@shanrana7669
Why are women so excited about this tech lMao?
 
为什么女性对这项技术这么兴奋啊，哈哈？
 
@TheThink5op
Cant wait to see a reddit post asking how to unstuck a soft cylinder from these jammed hand
 
等不及想看 Reddit 上出现帖子问‘如何把卡在机器人手里的软圆柱形物体弄出来的主题帖了
 
 
@idf3da
The future of robotics is in economic efficiency and component specialisation; what we are seeing right here is company who's superb hand modules will be used as mountable components on other robots.
Remember: robots are not a fantasy, they are an economic miracle; if they become cheap enough to produce they will be all around us; and China seems to be on the forefront of that
 
机器人的未来在于经济效益和部件专业化；我们现在看到的，正是一家‘凭借出色手部模块成为其他机器人可搭载部件供应商’的公司。记住：机器人不是幻想，而是经济奇迹；这东西只要生产成本足够低，它们就会遍布我们身边，而中国似乎正走在这一领域的前沿。
 
 
@davidbosak7503
Yeah, the component specialization makes so much sense.  One company trying to build the entire thing is too hard and expensive.  And results in some parts being good and some bad.  The specialization will result in the entire thing being good.  Very promising approach!
 
是的，部件模块化太有道理了。一家公司想打造完整的机器人，不仅面临难度大、成本高，还会出现那些部件好、那些差的问题。而专业化分工能让整个机器人的所有部件都达到最佳水平，这方法太有前景了！
 
@村田让
Unitree has already taken the lead in the general R&D platform market. Their affordable G1 R&D robot has dominated the global market in just one year. Numerous robotics companies, AI companies, research institutes, and university labs have purchased this robot and formed a R&D community around it, paving the way for rapid iteration of their robots. It's likely that other companies will use Unitree as the standard for supporting technology in their future robotics components.
 
宇树已经在通用研发平台市场占据领先地位。他们家价格亲民的 G1 研发型机器人，仅用一年就主导了全球市场。无数机器人公司、AI 公司、研究机构和高校实验室都买了他们机器人，并围绕它形成了研发社群，为机器人的快速迭代铺平了道路。未来其他公司在研发机器人部件时，很可能会以宇树的技术作为行业标准。
 
@Wilzig
HOLY MOLY, This is seriously impressive.
 
我的天，这也太惊艳了吧。
 
@atom2ueki
China is really No.1 now
 
中国现在已成为名副其实的NO.1了
 
 
@Just_Another_Random_Human
truly amazing! i just wonder how loud the pneumatics are
 
太惊人了！我就是好奇它的气动装置噪音有多大
 
@loganagle746
Holy crap this looks amazing! Please bring us more Demo-Videos like this, Wuji! People need to see what this thing can do!
Imagine the possibilities for this thing.
Imagine these kinds of mechanical hands on Robots. Imagine these kinds of mechanical hands on amputees...
 
我的天，这看起来也太牛了！无极，求多放些这样的演示视频！大家都想看看这玩意儿到底能做什么！ 
想想它的潜力吧。想想机器人装上这种机械手会怎样，想想残疾人装上这种机械手会怎样……
 
 
@Pickle_Juice_Studios
I’ll be impressed if it can do Connor’s coin tricks from Detroit become human
 
这要是能做出像《底特律：变人》里康纳那样变硬币的魔术，我才真的服。
 
@smetljesm2276
Excellent!Can it gently handle soft shafts?Asking for a friend.
 
太厉害了！这玩意能轻柔地撸某种圆柱形物体吗？帮朋友问的。
 
 
@wowzande
Every guy on earth is your friend.
 
全世界的男生都能成为你‘朋友’的。 
 
 
@Maormizrahi4192
Maybe instead of placing the obxt exactly where and how you want it, hand it to wuji and let wuji place it itself, then see how wuji moves it.
 
与其把物体按你想要的位置和角度放好，不如直接给无极机器人手，让它自己摆放，看它会怎样做动作，这样更能测试性能。
 
 
@노방래-f8b
대단하다~거의 사람손인데 더 강하다!
 
太厉害了简直跟真人的手一样，甚至力道更大！
 
 
@alessandrasolinas5349
Teleoperation - Looks great for Telework like in surgery or in every other job with ionizing radiation
 
远程操控——这对远程工作太有用了，比如外科手术，或是在有电离辐射的环境工作。
 
 
@theDarthLord01
I want to see wuji x boston , make it happen guys
 
我想看无极和波士顿动力合作，大伙赶紧撮合啊！
 
 
@jayjaysaves6589
First the robots came for your jobs, now they are coming for your hand jobs
 
先是机器人抢了你的工作，现在连‘手上的活儿’都要被抢了。
 
@Ab_Sense
We got five fingers AI before GTA6
 
《侠盗猎车手6》还没出，我们倒先有了‘五指AI手’。
 
 
@kogasoldier9379
“Do not imitate”. Bro i doubt I’ll ever even be in the same room as one of these.
 
‘请勿模仿’。兄弟，我怀疑我这辈子都未必能和这玩意儿待在同一间屋子里。
 
 
@AAAAAA-tj1nq
China superior tech
 
中国的技术太先进了。
 
 
@luizpereira7165
This robotic hand is fantastic! Adding tactile sensors would make it even better and significantly simplify machine learning.
 
这机械手太绝了！要是能加上触觉传感器会更厉害，还能大大简化机器学习的难度。
 
 
@苌正仁
我有一个大胆的想法
（原文如此）
 
 
@LightCrasher
Wow, I know some people that could really use a third hand!
 
哇，我知道有些人真的很需要第三只手！
 
 
@orn8471
This sh!t more articulate than my own hand
 
感觉这玩意比我自己的手还灵活呢！
 
 
@TeHPHoBoS999
If you told me this was just a guy in a green suit with his hand on the table and they were faking the hand that way, I'd believe it. This is genuinely incredible.
 
是你告诉我‘这是有人穿绿幕服，手放在桌子上，靠后期仿作机械手效果’，我真的会信。这也简直太不可思议了。
 
 
@MasterThief117
Imagine someone with two of these cosplaying as Fragile from Death Stranding.
 
想象一下有人装两只这玩意儿，COS《死亡搁浅》里的芙拉吉尔。
 
@giuseppeverdi8284
Can you make a version with warming feature and gentle like Arianna Grande?
 
你们能不能做一个带加热功能的、像爱莉安娜·格兰德的手那样的筋道轻柔的版本吗？
 
 
@mechwarrior-oz7pn
The future portrayed in I, Robot, Ex MaChina and A.I.  is just around the corner.
 
《我，机器人》《机械姬》《人工智能》里描绘的未来，已经近在眼前了。