VijayKumar_2012關于四軸飛行器的演講稿(中英文對照).doc

早上好我今天想談談自主飛行沙灘球其實，是小型飛行器，像這一個我想和大家談談設計這些飛行器時的挑戰(zhàn)和使用這些飛行器能給我們帶來的很多用處這些飛行器源于無人駕駛的飛行器但是那些都體積很大通常上萬磅重毫無靈活型可言它們也不是真的自主飛行的事實上，很多這些飛行器都是受飛行團隊控制的包括好幾個飛行員感應雷達操作員和團隊協(xié)調員我們想設計的飛行器是這樣的這里有兩張照片是你能夠在超市里買到的那種小飛行器小型直升機，四個螺旋槳不超過一米長只不過幾磅重我們把它們稍微改造一下，加上感應器和處理器，它們就可以在室內飛用不著導航系統(tǒng)我現(xiàn)在拿著的這個飛行器是其中之一是兩個學生做出來的艾利克斯和丹尼爾這個僅僅比零點一磅稍微重一點只需要大約十五瓦的電源你能看到它的直徑大約只有八個英寸讓我給你們快速解釋一下這些飛行器是怎么工作的它有四個螺旋槳當四個螺旋槳轉速相同這個飛行器就浮在空中當所有螺旋槳的速度提升時這個飛行器就加速升高當然了，如果飛行器已經(jīng)是傾斜的向著地平線側過來就會向這個方向加速怎么能讓它側過來呢，有兩個途徑從這張照片你能看到四號螺旋槳旋轉加速同時二號螺旋槳轉速變慢這時飛行器就能向一邊倒反之亦然當三號螺旋槳加速一號減速時飛行器就向前倒最后如果任意兩端的螺旋槳的轉速大于另兩端的螺旋槳的轉速飛行器就能原地旋轉所以裝在飛行器上的處理器基本上能判斷需要執(zhí)行哪些動作然后把它們組合起來決定給螺旋槳下什么指令一秒鐘六百次簡單地說這些飛行器就是這么工作的這個設計的一個好處就是小巧這些飛行器很靈活這里的R是飛行器的長度其實是半徑當半徑變小時很多物理參數(shù)都會變最重要的一個參數(shù)是慣性， 也就是對于運動的阻力結果是慣性決定角速度它是半徑的五次方函數(shù)當半徑變得越來越小時慣性越來越快地減小另一個結果是角速度的加速度也就是這里的希臘字母alpha等于一除以半徑也就是半徑的倒數(shù)當半徑越小時飛行器能轉彎越快這個視頻清楚地顯示大家看右下角的飛行器正在做一個三百六十度翻轉只需要不到半秒連續(xù)翻轉，稍微時間長一點這里飛行器上用的處理器能夠從飛行器上的加速度計和陀螺儀得到反饋信息然后算出，就像我剛才講的一秒鐘六百個指令來穩(wěn)定控制這個飛行器在左邊你能看到丹尼爾把飛行器拋到空中你能看到飛行器的控制有多快不管你怎么扔飛行器都能恢復平衡飛回來為什么我們要設計這種飛行器呢？因為這樣的飛行器有很多用處你能把它們放進像這樣的大樓里作為報警器去尋找入侵者尋找生化泄漏或者煤氣泄漏你還能用它們建摩天大樓呢這里是飛行器在搬梁運柱架構一個立方體的建筑這里我想和大家介紹一下這些機器人能被用來運貨當然一個問題是這些小飛行器擔不了多少重量你可能需要很多飛行器來搬運重物我們新做了個實驗其實不那么新了在日本仙臺，地震后不久我們能把這些飛行器送進倒塌的樓房或者核反應堆大樓來探測放射性強度一個根本的問題是當這些飛行器需要自控飛行，它們自己得弄明白怎么從一個地點到另一個地點這就變得有點難度了因為這些飛行器的動力學是很復雜的事實上它們總在對付十二維的空間這里我們用了一點小技巧我們拿這個十二位的空間把它們轉換成平的四維空間這個四維空間包括了橫軸，縱軸和豎軸，還有旋轉軸這些飛行器只需要計劃一件事，我們管它叫最小化加加加速度軌道提醒大家一點點物理學這里我們有位置向量，導數(shù)，速度和加速度還有加加速度還有加加加速度這個飛行器把加加加速度最小化基本上它的工作是創(chuàng)造一個光滑優(yōu)雅的運動曲線這樣來繞開障礙物所以這個四維平面中，這個飛行器使用最小化加加加速度軌道， 然后轉換回到復雜的十二維空間飛行器必須這樣做來獲得控制和執(zhí)行動作讓我給大家看幾個例子這些最小化加加加速度軌道是什么樣的這是第一個視頻這個飛行器從一個地點飛到另一個地點中間經(jīng)停一下顯然這個飛行器能飛出一個曲線軌道還有這樣的打圈的軌道這里飛行器對抗兩倍的重力它們上方還有一個動感監(jiān)控攝像機，每秒一百幅畫面來告訴這些飛行器它們的位置也能告訴這些飛行器障礙物在哪里障礙物移動都不要緊當?shù)つ釥柊烟兹θ拥娇罩酗w行器就開始計算套圈的位置試圖預測怎么才能最有效地鉆過去作為一個科研人員我們總在試圖鉆出重重圈套，拿到更多經(jīng)費甚至訓練了我們的飛行器也來做這個（掌聲）另一個飛行器能做的事情是當我們預先編入一些軌跡或者它自己學著走過的，它能夠記住這里大家能看到飛行器能夠（在預設軌跡上）加上一個動作積聚動量改變它的定向，再回到預設軌跡上來它必須這樣做因為這個窗上的縫隙只比它的寬度大一點點所以就像是一個跳水運動員從跳板上起跳，聚集動量，做個旋轉，兩圈半然后優(yōu)雅地回到平衡這個飛行器是自主這樣做的它知道怎么把小段的軌跡組合起來來做這些高難度的技巧現(xiàn)在我想換個話題談談這些小型飛行器的不足之處，就是體積小我已經(jīng)提過我們需要使用很多飛行器來克服體積小的不便一個難點是怎么使得這些飛行器集體飛行？我們在大自然中尋找答案我想給大家看一個視頻是關于Aphaenogaster沙漠蟻的在史狄文普熱特教授的實驗室里，這些螞蟻一起搬運重物這是一個無花果事實上無論什么東西，只要蘸上無花果汁這些螞蟻都會把它們帶回巢去這些螞蟻沒有任何中央調控它們是靠感應鄰近的螞蟻它們也沒有明確的交流但是因為它們能夠感應鄰近的螞蟻也能感應抬著的重物整群的螞蟻有默契這樣的協(xié)調正是飛行器需要的當一個飛行器被其他飛行器環(huán)繞時讓我們注意 I 和 J 這兩個當它們成群飛行時我們希望這兩個飛行器能夠監(jiān)控它們之間的距離我們需要確定這個距離是在可接受的范圍里的飛行器要檢測這個變化在控制指令中計算進去也是每秒一百次這個控制指令每秒會被送到馬達六百次所以這個程序是分散化執(zhí)行的再有，如果你有很多很多飛行器要完成集體飛行任務，能足夠快地集中協(xié)調所有這些信息是幾乎不可能的加上這些飛行器只能依靠局部的信息來決定做什么動作也就是要靠感應鄰近的飛行器最后我們希望這些機器人不知道它們的鄰居是誰也就是匿名飛行下一個我想給大家展示的是這段視頻這二十個小型飛行器成群飛行它們在監(jiān)測鄰居的位置維持群隊群隊的形狀還能變它們可以在一個平面上飛也可以上中下地飛大家可以看到它們能從上中下的群隊變成平面的在飛越障礙物的時候它們能邊飛邊變換隊形我想強調，這些飛行器距離都很近比如這個群隊，八架飛行器相互距離不過幾英寸盡管在空氣動力學上這些螺旋槳相互干擾它們還是能夠維持平穩(wěn)飛行（掌聲）現(xiàn)在它們會成群飛了它們就可以合作抬重物這里展示的是我們能夠把飛行器的能力翻倍，翻三倍，四倍僅僅通過讓它們和鄰居合作，大家可以看到這樣做的一個不便之處就是當加大數(shù)量時比如使用很多飛行器來抬一個物體你其實是加大了慣性這樣它們就不夠靈活了，這是一個代價但是你可以增加載荷承載量另一個我想給大家展示的用處是這是在我們實驗室這是研究生昆汀林夕的工作他的算法程序告訴這些飛行器怎么使用桁架結構自動建造一個立方體他的算法程序告訴這些機器人該用哪一塊什么時候用，用在哪里從這個視頻我們可以看到這個視頻是十倍或者十四倍速度播放的大家可以看到飛行器在搭建很不一樣的構架并且，所有的運動都是自主的昆汀僅僅是給它們一個藍圖也就是他想建的設計所有這里展示的實驗所有這些演習都是靠著它們自己的動感檢測攝像機完成的那么，當它們離開實驗室來到真實世界的時候，又怎么樣呢？沒有衛(wèi)星導航會怎么樣？這個飛行器其實裝有一個攝像機和一個激光測距儀，一個激光掃描儀它可以使用這些探測裝置來描繪周圍的環(huán)境的地圖這個地圖包括很多細節(jié)玄關，窗戶人，家具還能弄清楚相對于這些東西它自己在哪里所以這里沒有整體的協(xié)調系統(tǒng)這個協(xié)調系統(tǒng)是靠飛行器自己來完成的它自己在哪里，前面有什么還能利用周圍環(huán)境為自己找到出路這里我想給大家再看一段視頻這個算法程序是法蘭克沈和南希麥克教授編的當這個飛行器第一次飛入一個建筑它是怎么邊飛邊畫地圖的這個飛行器弄明白了這些細節(jié)開始畫地圖弄明白了相對這些細節(jié)，自己在哪里，然后自我定位全以每秒一百次的速度發(fā)生這就給我們一個機會來控制這些算法像我之前講過的所以這個機器人其實是被法蘭克遙控的但是它自己也可以弄明白怎么飛假設我想放一個這樣的飛行器進一幢樓我并不知道里面是什么樣的我可以讓它飛進去創(chuàng)造一個地圖然后飛回來告訴我里面是什么樣的所以，這個飛行器不僅僅解決了怎么從一點到另一點的問題還能夠隨時知道最好的目標在哪里基本上，它知道該去搜索哪里因為那里的信息是最“未知”的這就是它怎么填充這個地圖這里我想展示給大家最后一個用途當然這個技術有很多很多用途我是個教授，我們很關心教育這樣的飛行器其實可以改變我們的小學和中學教育我們在南加州離洛杉磯很近所以我不得不放點娛樂元素進去我想給大家看一個音樂視頻我想向你們介紹艾利克斯和丹尼爾，他們是導演兼制作（掌聲）在我播放這個視頻前我想告訴大家這是他們在過去三天做出來的因為主持人克瑞斯給我打了個電話在這個視頻中表演的飛行器全是靠自控表演的你能看到九個機器人，演奏六種不同樂器當然了，這是為了今年的TED2012特別制作的請欣賞（音樂）（掌聲）Good morning.Im here today to talkabout autonomous, flying beach balls.No, agile aerial robots like this one.Id like to tell you a little bit about the challenges in building theseand some of the terrific opportunitiesfor applying this technology.So these robotsare related to unmanned aerial vehicles.However, the vehicles you see here are big.They weigh thousands of pounds,are not by any means agile.Theyre not even autonomous.In fact, many of these vehiclesare operated by flight crewsthat can include multiple pilots,operators of sensorsand mission coordinators.What were interested in is developing robots like this -and here are two other pictures -of robots that you can buy off the shelf.So these are helicopters with four rotorsand theyre roughly a meter or so in scaleand weigh several pounds.And so we retrofit these with sensors and processors,and these robots can fly indoorswithout GPS.The robot Im holding in my handis this one,and its been created by two students,Alex and Daniel.So this weighs a little morethan a tenth of a pound.It consumes about 15 watts of power.And as you can see,its about eight inches in diameter.So let me give you just a very quick tutorialon how these robots work.So it has four rotors.If you spin these rotors at the same speed,the robot hovers.If you increase the speed of each of these rotors,then the robot flies up, it accelerates up.Of course, if the robot were tilted,inclined to the horizontal,then it would accelerate in this direction.So to get it to tilt, theres one of two ways of doing it.So in this pictureyou see that rotor four is spinning fasterand rotor two is spinning slower.And when that happenstheres moment that causes this robot to roll.And the other way around,if you increase the speed of rotor threeand decrease the speed of rotor one,then the robot pitches forward.And then finally,if you spin opposite pairs of rotorsfaster than the other pair,then the robot yaws about the vertical axis.So an on-board processoressentially looks at what motions need to be executedand combines these motionsand figures out what commands to send to the motors600 times a second.Thats basically how this thing operates.So one of the advantages of this designis, when you scale things down,the robot naturally becomes agile.So here Ris the characteristic length of the robot.Its actually half the diameter.And there are lots of physical parameters that changeas you reduce R.The one thats the most importantis the inertia or the resistance to motion.So it turns out,the inertia, which governs angular motion,scales as a fifth power of R.So the smaller you make R,the more dramatically the inertia reduces.So as a result, the angular acceleration,denoted by Greek letter alpha here,goes as one over R.Its inversely proportional to R.The smaller you make it the more quickly you can turn.So this should be clear in these videos.At the bottom right you see a robotperforming a 360 degree flipin less than half a second.Multiple flips, a little more time.So here the processes on boardare getting feedback from accelerometersand gyros on boardand calculating, like I said before,commands at 600 times a secondto stabilize this robot.So on the left, you see Daniel throwing this robot up into the air.And it shows you how robust the control is.No matter how you throw it,the robot recovers and comes back to him.So why build robots like this?Well robots like this have many applications.You can send them inside buildings like thisas first responders to look for intruders,maybe look for biochemical leaks,gaseous leaks.You can also use themfor applications like construction.So here are robots carrying beams, columnsand assembling cube-like structures.Ill tell you a little bit more about this.The robots can be used for transporting cargo.So one of the problems with these small robotsis their payload carrying capacity.So you might want to have multiple robotscarry payloads.This is a picture of a recent experiment we did -actually not so recent anymore -in Sendai shortly after the earthquake.So robots like this could be sent into collapsed buildingsto assess the damage after natural disasters,or sent into reactor buildingsto map radiation levels.So one fundamental problemthat the robots have to solve if theyre to be autonomousis essentially figuring outhow to get from point A to point B.So this gets a little challengingbecause the dynamics of this robot are quite complicated.In fact, they live in a 12-dimensional space.So we use a little trick.We take this curved 12-dimensional spaceand transform itinto a flat four-dimensional space.And that four-dimensional spaceconsists of X, Y, Z and then the yaw angle.And so what the robot doesis it plans what we call a minimum snap trajectory.So to remind you of physics,you have position, derivative, velocity,then acceleration,and then comes jerkand then comes snap.So this robot minimizes snap.So what that effectively doesis produces a smooth and graceful motion.And it does that avoiding obstacles.So these minimum snap trajectories in this flat spaceare then transformed backinto this complicated 12-dimensional space,which the robot must dofor control and then execution.So let me show you some examplesof what these minimum snap trajectories look like.And in the first video,youll see the robot going from point A to point Bthrough an intermediate point.So the robot is obviously capableof executing any curve trajectory.So these are circular trajectorieswhere the robot pulls about two Gs.Here you have overhead motion capture cameras on the topthat tell the robot where it is 100 times a second.It also tells the robot where these obstacles are.And the obstacles can be moving.And here youll see Daniel throw this hoop into the air,while the robot is calculating the position of the hoopand trying to figure out how to best go through the hoop.So as an academic,were always trained to be able to jump through hoops to raise funding for our labs,and we get our robots to do that.(Applause)So another thing the robot can dois it remembers pieces of trajectorythat it learns or is pre-programmed.So here you see the robotcombining a motionthat builds up momentumand then changes its orientation and then recovers.So it has to do this because this gap in the windowis only slightly larger than the width of the robot.So just like a diver stands on a springboardand then jumps off it to gain momentum,and then does this pirouette, this two and a half somersault throughand then gracefully recovers,this robot is basically doing that.So it knows how to combine little bits and pieces of trajectoriesto do these fairly difficult tasks.So I want change gears.So one of the disadvantages of these small robots is its size.And I told you earlierthat we may want to employ lots and lots of robotsto overcome the limitations of size.So one difficultyis how do you coordinate lots of these robots?And so here we looked to nature.So I want to show you a clipof Aphaenogaster desert antsin Professor Stephen Pratts lab carrying an object.So this is actually a piece of fig.Actually you take any object coated with fig juiceand the ants will carry them back to the nest.So these ants dont have any central coordinator.They sense their neighbors.Theres no explicit communication.But because they sense the neighborsand because they sense the object,they have implicit coordination across the group.So this is the kind of coordinationwe want our robots to have.So when we have a robotwhich is surrounded by neighbors -and lets look at robot I and robot J -what we want the robots to dois to monitor the separation between themas they fly in formation.And then you want to make surethat this separation is within acceptable levels.So again the robots monitor this errorand calculate the control commands100 times a second,which then translates to the motor commands 600 times a second.So this also has to be donein a decentralized way.Again, if you have lots and lots of robots,its impossible to coordinate all this information centrallyfast enough in order for the robots to accomplish the task.Plus the robots have to base their actionsonly on local information,what they sense from their neighbors.And then finally,we insist that the robots be agnosticto who their neighbors are.So this is what we call anonymity.So what I want to show you nextis a videoof 20 of these little robotsflying in formation.Theyre monitoring their neighbors position.Theyre maintaining formation.The formations can change.They can be planar formations,they can be three-dimensional formations.As you can see here,they collapse from a three-dimensional formation into planar formation.And to fly through obstaclesthey can adapt the formations on the fly.So again, these robots come really close together.As you can see in this figure-eight flight,they come within inches of each other.And despite the aerodynamic interactionsof these propeller blades,theyre able to maintain stable flight.(Applause)So once you know how to fly in formation,you can actually pick up objects cooperatively.So this just showsthat we can double, triple, quadruplethe robot strengthby just getting them to team with neighbors, as you can see here.One of the disadvantages of doing thatis, as you scale things up -so if you have lots of robots carrying the same thing,youre essentially effectively increasing the inertia,and therefore you pay a price; theyre not as agile.But you do gain in terms of payload carrying capacity.Another application I want to show you -again, this is in our lab.This is work done by Quentin Lindsey whos a graduate student.So his algorithm essentially tells these robotshow to autonomously buildcubic structuresfrom truss-like elements.So his algorithm tells the robotwhat part to pick up,when and where to place it.So in this video you see -and its sped up 10, 14 times -you see three different structures being built by these robots.And again, everything is autonomous,and all Quentin has to dois to get them a blueprintof the design that he wants to build.So all these experiments youve seen thus far,all these demonstrations,have been done with the help of motion capture systems.So what happens when you leave your laband you go outside into the real world?And what if theres no GPS?So this robotis actually equipped with a cameraand a laser rangefinder, laser scanner.And it uses these sensorsto build a map of the environment.What that map consists of are features -like doorways, windows,people, furniture -and it then figures out where its position iswith respect to the features.So there is no global coordinate system.The coordinate system is defined based on the robot,where it is and what its looking at.And it navigates with respect to those features.So I want to show you a clipof algorithms developed by Frank Shenand Professor Nathan Michaelthat shows this robot entering a building for the very first timeand creating this map