Robots that fly ... and cooperate | Vijay Kumar

譯者: kane tan 審譯者: Joan Liu

Good morning.

早安。

I'm here today to talk about autonomous flying beach balls.

今天我想要來談一談

會自動飛行的海灘球。

(Laughter)

不是啦，是靈巧的飛行機器人，就像這一個。

No, agile aerial robots like this one.

I'd like to tell you a little bit about the challenges in building these,

我想告訴大家製作這種東西的挑戰性

以及一些很棒的可能性

and some of the terrific opportunities for applying this technology.

來運用這種技術。

So these robots are related to unmanned aerial vehicles.

這些機器人

算是一種無人的飛行器。

不過，如你所見，它們的尺寸都比較大。

However, the vehicles you see here are big.

它們都有幾千磅重，

They weigh thousands of pounds, are not by any means agile.

一點都不靈巧。

They're not even autonomous.

它們甚至並不是自動操作的。

In fact, many of these vehicles are operated by flight crews

事實上，大部分這些飛行器

是由飛行小組所操作，

可能有好幾個駕駛員

that can include multiple pilots,

operators of sensors,

同時在操控著感應器

and mission coordinators.

以及任務協調器。

What we're interested in is developing robots like this --

我們想要開發的機器人是像這個樣子 --

左邊這裡另外兩張照片--

and here are two other pictures --

這些你都可以買到現成的。

of robots that you can buy off the shelf.

這些是一種具有四個螺旋槳的直昇機，

So these are helicopters with four rotors,

它們大約是一公尺大小，

and they're roughly a meter or so in scale,

也有好幾磅重。

and weigh several pounds.

於是我們將它們進行感應器與處理器的改良，

And so we retrofit these with sensors and processors,

讓這些機器人能夠在室內

and these robots can fly indoors.

不靠GPS飛行。

Without GPS.

我手中所拿的這個機器人

The robot I'm holding in my hand

就是這種飛行器，

is this one,

這是由兩位學生所製作的，

and it's been created by two students,

Alex 以及 Daniel。

Alex and Daniel.

它的重量大概是

So this weighs a little more than a tenth of a pound.

十分之一磅左右。

It consumes about 15 watts of power.

它消耗的能量大概是15瓦。

如你所見，

And as you can see, it's about eight inches in diameter.

它的直徑大概是8英吋大。

讓我替大家簡單介紹一下

So let me give you just a very quick tutorial

on how these robots work.

這些機器人的原理。

這裡有四個螺旋槳。

So it has four rotors.

If you spin these rotors at the same speed,

當這四個螺旋槳速度相同時，

the robot hovers.

機器人就會懸浮在空中。

If you increase the speed of each of these rotors,

如果這些螺旋槳速度增加，

機器人就會飛起來，往上加速。

then the robot flies up, it accelerates up.

Of course, if the robot were tilted,

當然，如果機器人傾斜了，

相對於水平線來說，

inclined to the horizontal,

then it would accelerate in this direction.

它就會往這個方向前進。

So to get it to tilt,

想讓它傾斜的話，這裡有兩種方法可以辦到。

there's one of two ways of doing it.

在這圖片中，

So in this picture, you see that rotor four is spinning faster

你可以看見4號螺旋槳轉速變快一點，

and rotor two is spinning slower.

而2號螺旋槳轉速變慢一點。

And when that happens,

當這種情況發生時，

there's a moment that causes this robot to roll.

就會讓機器人進行翻轉。

另一種狀況是，

And the other way around,

if you increase the speed of rotor three and decrease the speed of rotor one,

當3號螺旋槳的速度上升，

1號螺旋槳的速度下降時，

機器人就會往前傾斜。

then the robot pitches forward.

And then finally,

而最後一種可能，

if you spin opposite pairs of rotors

當對角線的兩組螺旋槳

faster than the other pair,

轉得比另外一組快時，

then the robot yaws about the vertical axis.

機器人就會在垂直方向偏移。

有一個內置處理器

So an on-board processor

essentially looks at what motions need to be executed

一直在監控著該進行什麼動作，

並且將這些動作進行組合，

and combines these motions,

然後以每秒600次的速度

and figures out what commands to send to the motors --

決定出該對這些螺旋槳下達什麼指令。

600 times a second.

That's basically how this thing operates.

這就是它操作的基本概念。

這種設計的其中一項優點是，

So one of the advantages of this design

當你將它的尺寸縮小時，

is when you scale things down,

機器人自然就會變得很靈巧。

the robot naturally becomes agile.

這邊的 R

So here, R is the characteristic length of the robot.

代表著機器人特性的長度。

事實上是直徑的一半。

It's actually half the diameter.

And there are lots of physical parameters that change as you reduce R.

而當你將 R 縮減時，

許多物理係數就會跟著變動。

The one that's most important is the inertia,

其中最重要的

就是慣性或稱為阻止變動的抵抗力。

or the resistance to motion.

結果，

So it turns out the inertia, which governs angular motion,

控制了角運動的慣性，

大小約是 R 的 5 次方。

scales as a fifth power of R.

所以當 R 變小時，

So the smaller you make R,

the more dramatically the inertia reduces.

慣性會急遽的下降。

So as a result, the angular acceleration,

結果，角加速度，

denoted by the Greek letter alpha here,

這裡用希臘字母的 α 表示，

goes as 1 over R.

變成了 1 / R 。

It's inversely proportional to R.

它和 R 成反比。

The smaller you make it, the more quickly you can turn.

當尺寸越小時，它就越容易旋轉。

用這個影片說明會清楚一點。

So this should be clear in these videos.

在右下角，你可以看見一個機器人

On the bottom right, you see a robot performing a 360-degree flip

正在進行 360 度翻轉

in less than half a second.

在不到 1/2 秒的時間內。

Multiple flips, a little more time.

多次的翻轉，只要稍微長一點點的時間。

在這種狀況下，內置的處理器

So here the processes on board

接收了加速器

are getting feedback from accelerometers and gyros on board,

以及陀螺儀回傳的資訊，

and calculating, like I said before,

然後進行計算，如先前所說，

commands at 600 times a second,

用每秒600次的速度發出指令，

to stabilize this robot.

讓機器人保持平衡。

So on the left, you see Daniel throwing this robot up into the air,

在左下角，Daniel 正將機器人拋向空中。

and it shows you how robust the control is.

這會讓你知道它的操控能力有多強大。

不論你怎麼丟，

No matter how you throw it,

the robot recovers and comes back to him.

機器人都能恢復平衡然後回到他的手中。

So why build robots like this?

為什麼要將機器人設計成這樣呢？

嗯，這種機器人有很多種運用方式。

Well, robots like this have many applications.

你可以將它派遣到這種建築物裡，

You can send them inside buildings like this,

as first responders to look for intruders,

擔任先遣部隊去找出侵入者，

或是去找尋生化物質洩漏，

maybe look for biochemical leaks,

或是瓦斯洩漏等。

gaseous leaks.

你也可以將它們運用在

You can also use them for applications like construction.

例如建築上面。

So here are robots carrying beams, columns

這裡的機器人正運送著橫梁、柱子，

並且組合成立方體形狀的建築物。

and assembling cube-like structures.

I'll tell you a little bit more about this.

我再告訴大家詳細一點。

The robots can be used for transporting cargo.

這些機器人可以用來運送貨櫃。

So one of the problems with these small robots

但這些小機器人的困難在於

is their payload-carrying capacity.

它們對於重物的負載能力有限。

So you might want to have multiple robots carry payloads.

所以如果你可能會希望能有多一點機器人

一起來搬運這個重物。

This is a picture of a recent experiment we did --

這是我們近期實驗的照片 --

事實上已經不算是近期了 --

actually not so recent anymore --

in Sendai, shortly after the earthquake.

在地震過後的仙台市(日本)。

So robots like this could be sent into collapsed buildings,

這種機器人可以被派遣進入傾倒的建築物裡面

去評估天災造成的損害，

to assess the damage after natural disasters,

或是派遣到反應爐裡

or sent into reactor buildings,

to map radiation levels.

去勘查輻射等級。

So one fundamental problem that the robots have to solve

如果這些機器人想有自主能力的話，

它們必須先解決這個問題，

if they are to be autonomous,

is essentially figuring out how to get from point A to point B.

就是必須能夠判斷

怎麼從 A 點到達 B 點。

So this gets a little challenging,

這有一點難度，

because the dynamics of this robot are quite complicated.

因為這個機器人的動力學是相當複雜的。

In fact, they live in a 12-dimensional space.

事實上，它們活在 12 維空間裡。

所以我們運用了一些技巧。

So we use a little trick.

We take this curved 12-dimensional space,

我們將這個 12 維空間的曲線

轉換成為

and transform it into a flat, four-dimensional space.

一個平面的四維空間。

在這個四維空間之中，

And that four-dimensional space consists of X, Y, Z,

包含了 X, Y, Z 還有偏移的角度。

and then the yaw angle.

And so what the robot does,

所以這個機器人所做的是，

is it plans what we call a minimum-snap trajectory.

去找出我們所說的最小震盪軌跡。

複習一下物理參數，

So to remind you of physics:

You have position, derivative, velocity;

我們有位置，接著衍生出速度，

then acceleration;

以及加速度，

and then comes jerk,

還有加加速度，

and then comes snap.

然後是震盪。

So this robot minimizes snap.

所以機器人將震盪進行最小化。

So what that effectively does,

這實際上的結果就是

is produce a smooth and graceful motion.

產生出柔順且優美的動作。

And it does that avoiding obstacles.

它還可以用來避開障礙物。

而這些最小震盪軌跡在這個平面空間中

So these minimum-snap trajectories in this flat space are then transformed

又會被轉換回

back into this complicated 12-dimensional space,

這個複雜的 12 維空間，

才能夠讓機器人去進行

which the robot must do for control and then execution.

控制以及執行任務。

So let me show you some examples

讓我給大家看一些例子

of what these minimum-snap trajectories look like.

說明這些最小震盪軌跡是什麼樣子。

在第一段影片中，

And in the first video,

you'll see the robot going from point A to point B,

你可以看見機器人經過中繼點

由 A 點到達 B 點。

through an intermediate point.

(Whirring noise)

所以機器人確實可以

So the robot is obviously capable of executing any curve trajectory.

去執行任何曲線軌跡。

這些是環狀軌跡，

So these are circular trajectories,

where the robot pulls about two G's.

機器人牽引著大約 2 G 的重力。

Here you have overhead motion capture cameras on the top

在上面有個置頂動態影像攝影機，

that tell the robot where it is 100 times a second.

它會以每秒100次的速度告訴機器人自己在哪裡。

It also tells the robot where these obstacles are.

它也會告訴機器人這些障礙物的位置。

這些也可以是移動中的障礙物。

And the obstacles can be moving.

And here, you'll see Daniel throw this hoop into the air,

你將會看見 Daniel 將這個鐵環丟向空中，

while the robot is calculating the position of the hoop,

機器人會計算鐵環的位置，

然後試著去找出穿過鐵環的最佳方式。

and trying to figure out how to best go through the hoop.

身為一個學術人員，

So as an academic,

we're always trained to be able to jump through hoops

我們總是被訓練得能夠赴湯蹈火才能籌措研究經費，

to raise funding for our labs,

所以我們也要我們的機器人做到。

and we get our robots to do that.

(Applause)

(掌聲)

這機器人還能做另一件事，

So another thing the robot can do

就是去記住軌跡的片段，

is it remembers pieces of trajectory

that it learns or is pre-programmed.

不論是它自行發現的或是事先輸入的。

所以你可以看見機器人會去

So here, you see the robot combining a motion that builds up momentum,

組合一項動作

讓它產生動量，

and then changes its orientation and then recovers.

接著改變自己的行進方向在回復過來。

它必須這麼做，因為這個窗戶的缺口大小

So it has to do this because this gap in the window

is only slightly larger than the width of the robot.

只比機器人的寬度稍微大一點。

就像是跳水選手站在跳板上，

So just like a diver stands on a springboard

and then jumps off it to gain momentum,

接著會跳起來用以產生動量，

然後快速旋轉，翻轉兩周半進行穿越，

and then does this pirouette, this two and a half somersault through

最後優雅的回復，

and then gracefully recovers,

this robot is basically doing that.

這就是機器人所做的事。

So it knows how to combine little bits and pieces of trajectories

它懂得如何去結合這些零碎的軌跡

to do these fairly difficult tasks.

以達成這些相當困難的任務。

我想換個話題。

So I want change gears.

So one of the disadvantages of these small robots is its size.

這些小機器人的缺點之一就是尺寸。

如同先前所提，

And I told you earlier

that we may want to employ lots and lots of robots

我們想使用大量的機器人

來解決尺寸上的限制。

to overcome the limitations of size.

但有個困難點是

So one difficulty is:

How do you coordinate lots of these robots?

你要如何去協調這些機器人呢？

And so here, we looked to nature.

這部份我們觀察了自然界。

So I want to show you a clip of Aphaenogaster desert ants,

我想讓大家看一段影片，

關於沙漠盤腹蟻

在 Stephen Pratt 教授的實驗室裡搬運東西。

in Professor Stephen Pratt's lab, carrying an object.

事實上這是一小塊無花果。

So this is actually a piece of fig.

事實上你可以把任何東西沾附一層無花果汁

Actually you take any object coated with fig juice,

螞蟻們就會將它搬回巢穴裡。

and the ants will carry it back to the nest.

So these ants don't have any central coordinator.

這些螞蟻並沒有中樞協調者。

它們能感覺到旁邊的鄰居們。

They sense their neighbors.

不用進行明確的溝通。

There's no explicit communication.

但因為它們能感覺到鄰居，

But because they sense the neighbors

因為它們能感覺到東西，

and because they sense the object,

they have implicit coordination across the group.

它們在團體間有著隱性協調能力。

這種協調能力

So this is the kind of coordination we want our robots to have.

就是我們希望機器人能有的。

So when we have a robot which is surrounded by neighbors --

當我們的一個機器人

被周圍的機器人包圍時 --

看看機器人 I 和機器人 J --

and let's look at robot I and robot J --

我們希望機器人做的事情是

what we want the robots to do,

當它們以特定隊形飛行時

is to monitor the separation between them,

as they fly in formation.

去偵測它們之間的距離。

And then you want to make sure

你期望能夠確保

that this separation is within acceptable levels.

這個距離是在可接受的範圍內。

於是機器人們偵測著這個誤差值

So again, the robots monitor this error

and calculate the control commands 100 times a second,

然後以每秒100次的速度

去估算控制指令，

which then translates into motor commands,

接著以每秒600次的速度對螺旋槳進行動作指令。

600 times a second.

這必須是在

So this also has to be done in a decentralized way.

沒有中央控制的方式下進行。

Again, if you have lots and lots of robots,

當你有許許多多機器人的時候，

想要以中央協調訊息的方式

it's impossible to coordinate all this information centrally

fast enough in order for the robots to accomplish the task.

快速的讓所有機器人完成任務是不可能的。

Plus, the robots have to base their actions only on local information --

再加上機器人們必須依靠

它們自身去偵測到鄰近機器人

以獲得訊息來進行動作。

what they sense from their neighbors.

最後，

And then finally,

we insist that the robots be agnostic to who their neighbors are.

我們堅持機器人必須無法預知

鄰近機器人會是誰。

So this is what we call anonymity.

也就是匿名的方式。

接下來我將要給大家看

So what I want to show you next is a video of 20 of these little robots,

一段影片

關於20個這些小機器人

flying in formation.

以特定隊形進行飛行。

它們正在偵測鄰近機器人的位置。

They're monitoring their neighbors' positions.

它們正在保持著這個隊形。

They're maintaining formation.

The formations can change.

這些隊形可以改變。

They can be planar formations,

可以是平面的隊形，

they can be three-dimensional formations.

也可以是三維空間的隊形。

如你所見的，

As you can see here,

they collapse from a three-dimensional formation into planar formation.

它們從三維空間的隊形變換成平面的隊形。

在穿越障礙物時，

And to fly through obstacles,

they can adapt the formations on the fly.

它們可以在飛行中調整隊形。

這些機器人移動時真的靠得很近。

So again, these robots come really close together.

As you can see in this figure-eight flight,

在這個 8 字飛行隊形中，

they come within inches of each other.

它們的距離只有幾吋而已。

儘管在這些螺旋槳葉片之間

And despite the aerodynamic interactions with these propeller blades,

有著空氣動力的交互影響，

they're able to maintain stable flight.

它們仍然能維持穩定的飛行。

(Applause)

(掌聲)

一旦你知道要怎麼進行特定飛行隊形，

So once you know how to fly in formation,

你就能準確的協力拿起物體。

you can actually pick up objects cooperatively.

而這是要告訴大家

So this just shows that we can double, triple, quadruple

藉由將機器人組合成小組後，

我們可以將機器人們的力量

the robots' strength,

by just getting them to team with neighbors, as you can see here.

放大兩倍、三倍、四倍，就像是你將看到的這樣。

但這樣做有一個缺點，

One of the disadvantages of doing that is, as you scale things up --

當你將尺寸放大以後 --

so if you have lots of robots carrying the same thing,

如果你有很多這些機器人載運同一個東西，

你一定會有效地增加慣性，

you're essentially increasing the inertia,

and therefore you pay a price; they're not as agile.

於是你將會付出代價，它們會失去靈巧性。

But you do gain in terms of payload-carrying capacity.

但你可以相對獲得載運負重能力。

另一項我想給大家看的運用 --

Another application I want to show you -- again, this is in our lab.

這也是在我們的實驗室裡進行的。

This is work done by Quentin Lindsey, who's a graduate student.

這是由 Quentin Lindsey 完成的，他是一位研究生。

他的演算法告訴這些機器人們

So his algorithm essentially tells these robots

如何能夠自主性的

how to autonomously build cubic structures

將綑狀的材料

建造成立體建築。

from truss-like elements.

他的演算法告訴機器人

So his algorithm tells the robot what part to pick up,

該拿起哪一個部份，

以及什麼時候該把它放在哪裡。

when, and where to place it.

你可以在這短片中看到 --

So in this video you see --

and it's sped up 10, 14 times --

這是以 10 倍、14 倍速播放 --

你可以看見這些機器人們建造了三種不同建築。

you see three different structures being built by these robots.

再次提醒，一切都是自主性進行的，

And again, everything is autonomous,

而 Quentin 所做的是

and all Quentin has to do

is to give them a blueprint of the design that he wants to build.

給這些機器人一張藍圖

記載著他想要的建築設計。

So all these experiments you've seen thus far,

你所看見的這些實驗，

all these demonstrations,

這些展示，

have been done with the help of motion-capture systems.

都使用了動作擷取系統。

如果離開了實驗室，

So what happens when you leave your lab,

走進真實世界會變成怎麼樣呢？

and you go outside into the real world?

And what if there's no GPS?

如果沒有 GPS 會怎樣呢？

So this robot is actually equipped with a camera,

這個機器人

裝置了一具攝影機，

一具雷射H搜尋器，雷射掃描器。

and a laser rangefinder, laser scanner.

它使用這些感應器

And it uses these sensors to build a map of the environment.

來製作一張周圍的地圖。

這地圖然有著一些環境特徵 --

What that map consists of are features --

例如大門、窗戶、

like doorways, windows, people, furniture --

人、家具 --

接著它會辨識出相對於這些環境特徵

and it then figures out where its position is,

它所處的位置。

with respect to the features.

So there is no global coordinate system.

這裡並沒有整體座標系統。

座標系統是機器人自身定義出來的，

The coordinate system is defined based on the robot,

where it is and what it's looking at.

藉由它所在的位置以及它所看到的東西。

And it navigates with respect to those features.

接著它對這些環境特徵進行探索。

我想給大家看一段影片，

So I want to show you a clip

of algorithms developed by Frank Shen and Professor Nathan Michael,

關於 Frank Shen 以及 Nathan Michael 教授

所開發出來的演算法，

that shows this robot entering a building for the very first time,

這個機器人第一次進入一個建築物，

and creating this map on the fly.

然後在飛行中製作了這個地圖。

So the robot then figures out what the features are,

於是機器人知道環境特徵是什麼東西。

it builds the map,

它製作出地圖。

it figures out where it is with respect to the features,

它知道自己相對於環境特徵的位置，

and then estimates its position 100 times a second,

然後每秒100次的速度

估算出自己的位置，

allowing us to use the control algorithms that I described to you earlier.

讓我們可以利用

剛剛說過的控制演算法。

So this robot is actually being commanded remotely by Frank,

事實上這個機器人正被

Frank 以遠端遙控的方式下指令。

但這個機器人也能自行判斷

but the robot can also figure out where to go on its own.

它應該往哪裡走。

假設我把它送進一個建築物，

So suppose I were to send this into a building,

而我完全不知道這個建築物的樣子，

and I had no idea what this building looked like.

我可以命令機器人進入，

I can ask this robot to go in,

製作出一張地圖，

create a map,

然後回來告訴我建築物的樣子。

and then come back and tell me what the building looks like.

So here, the robot is not only solving the problem

所以機器人並不只是解決

如何從地圖上的A點到B點這個問題，

of how to go from point A to point B in this map,

but it's figuring out what the best point B is at every time.

它甚至知道

每一次的最佳B點是哪個位置。

於是它知道該往哪裡去

So essentially it knows where to go

to look for places that have the least information,

以找出還沒有訊息的位置。

這就是它如何把地圖裝滿的方法。

and that's how it populates this map.

So I want to leave you with one last application.

最後，

我想再給大家看一樣應用。

And there are many applications of this technology.

這個技術有許多運用方式。

I'm a professor, and we're passionate about education.

我是一個教授，我們對教育充滿熱情。

這種機器人可以改變

Robots like this can really change the way we do K-12 education.

我們進行12年國教的方式。

我們身在南加州，

But we're in Southern California,

很靠近洛杉磯，

close to Los Angeles,

所以我想用關於娛樂的例子

so I have to conclude with something focused on entertainment.

來作為最後的結尾。

我想用一段音樂影片來作為結尾。

I want to conclude with a music video.

我要為大家介紹

I want to introduce the creators, Alex and Daniel, who created this video.

這個影片的作者，Alex 和 Daniel。

(掌聲)

(Applause)

So before I play this video,

在我播放影片之前，

I want to tell you that they created it in the last three days,

我想告訴大家他們在接到 Chris 電話後的三天內

after getting a call from Chris.

就將這段影片製作完了。

And the robots that play in the video are completely autonomous.

影片中演奏的機器人

都是完全自主性的進行。

You will see nine robots play six different instruments.

你可以看見 9 個機器人們演奏著 6 種不同的樂器。

當然，這是為了 TED 2012 特別製作的。

And of course, it's made exclusively for TED 2012.

讓我們一起來欣賞。

Let's watch.

(Sound of air escaping from valve)

(Music)

(Whirring sound)

(Music)

(音樂聲)

(掌聲)

(Applause) (Cheers)