【英翻】中国《无极科技》机械手亮相惊艳全世界

摘要：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 bi

图

评论翻译

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。