《HodLipson_2007[霍德.利普森][“自我感知”机器人].pdf》由会员分享，可在线阅读，更多相关《HodLipson_2007[霍德.利普森][“自我感知”机器人].pdf（3页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、www.XiYuS锡育软件So,where are the robots?啊，机器人们都在哪儿呢？00:15Weve been told for 40 years already that theyre comingsoon.四十年前人们就说它们很快就要来了00:17Very soon theyll be doing everything for us:theyll becooking,cleaning,buying things,shopping,building.Butthey arent here.很快，它们就能为我们做一切事情 它们会做饭，打扫，买东西，购物，甚至是建房子。但直到今天，
2、它们也没能进入我们的生活。00:20Meanwhile,we have illegal immigrants doing all the work,but we dont have any robots.这段时间里，非法移民承担着这些工作，但我们什么机器人都没有。00:28So what can we do about that?What can we say?我们又能做些或说些什么呢？00:34So I want to give a little bit of a different perspective of howwe can perhaps look at these things
3、in a little bit of adifferent way.我想给大家带来点不同的启发 看看今天我们能不能换一个角度看待这些事情00:38And this is an x-ray picture of a real beetle,and a Swisswatch,back from 88.You look at that-what was true thenis certainly true today.这是一张X光片 上面有一只活甲虫，和一只88年的瑞士手表，你看 无论是当时还是现在00:48We can still make the pieces,we can make the r
4、ight pieces,we can make the circuitry of the right computational power,but we cant actually put them together to make somethingthat will actually work and be as adaptive as these systems.我们都可以做出这些零件，一模一样的零件，我们能做出有同样计算能力的电路，但我们并不能把它们放在一起，再做出个什么东西 能和这些系统（甲虫）有一样的适应力。00:57immigrants:n.移民(immigrant的复数)be
5、etle:vi.急忙来回;突出/vt.用槌打/n.甲虫;大槌 circuitry:n.电路;电路系统;电路学;一环路 computational:adj.计算的 adaptive:adj.适应的,适合的So lets try to look at it from a different perspective.那么就让我们试着换个角度再看看这个问题。01:11Lets summon the best designer,the mother of all designers:lets see what evolution can do for us.让我们召集最好的设计师，所有设计师的鼻祖，看看
6、进化论能给我们做些什么。01:13So we threw in-we created a primordial soup with lots ofpieces of robots:with bars,with motors,with neurons.所以我们就找来了-我们创造了一锅“原汤”，里面有组装机器人需要的各种零件：条状的，带马达的，还有些带神经元的。01:20Put them all together,and put all this under kind of naturalselection,under mutation,and rewarded things for how we
7、llthey can move forward.把它们都放在一起，然后把所有这些放到一种自然选择，可能产生突变的环境中，并奖励那些发生“进化”的零件组。01:28A very simple task,and its interesting to see what kind ofthings came out of that.这是一个非常简单的任务，并且得出的结果非常有趣。01:36So if you look,you can see a lot of different machines comeout of this.They all move around,they all crawl i
8、n differentways,and you can see on the right,that we actually made acouple of these things,看一下，你能看到各式各样的机器 在这种模式下制造出来，它们到处移动，以不同的方式爬行，在右边你可以看到 我们真的做出了几个这样的玩意儿，01:42and they work in reality.These are not very fantastic robots,but they evolved to do exactly what we reward them for:formoving forward.So
9、 that was all done in simulation,but wecan also do that on a real machine.它们在现实中真的能工作。这些还算不上非常先进的机器人，但它们的确按照我们奖励的方向进化了：那就是向前进化。这些都是模拟的，但我们在真的机器上也做成功了，01:53summon:vt.召唤;召集;鼓起;振作 primordial:adj.原始的;根本的;原生的 motors:n.汽车;发动机(motor的复数);汽车公司证券 mutation:n.遗突变;变化;元音变化 rewarded:劳经奖赏 move around:v.走来走去;绕着来回转 s
10、imulation:n.仿真;模拟;模仿;假装Heres a physical robot that we actually have a population ofbrains,competing,or evolving,on the machine.这是一个真实的机器人，我们动用了大量的人力脑力 让机器们相互竞争，共同进化02:05Its like a rodeo show:they all get a ride on the machine,andthey get rewarded for how fast or how far they can make themachine move
11、 forward.这有点像一场套马表演：他们都骑在机器上 根据他们能够驾驶机器前进的速度和距离 得到奖励02:15And you can see these robots are not ready to take over theworld yet,but they gradually learn how to move forward,andthey do this autonomously.你可以看到这些机器人都还没完全准备好 占领这个世界，但是 它们渐渐地学会如何前进 并且是自发地学习。02:23So in these two examples,we had basically mac
12、hines thatlearned how to walk in simulation,and also machines thatlearned how to walk in reality.所以从这两个例子中，我们已经基本上 得到了能在虚拟中学习走路 和在现实中学习走路的机器人。02:33TED演讲者：Hod Lipson|霍德.利普森演讲标题：Robots that are self-aware|“自我感知”机器人内容概要：Hod Lipson works at the intersection of engineering and biology,studying robotsand
13、the way they behave and evolve.His work has exciting implications for design andmanufacturing-and serves as a window to understand our own behavior and evolution.霍德.利普森演示了他的几个神奇的小机器人。它们不仅有学习能力，理解自己，甚至还能进行自我复制！But I want to show you a different approach,and this is thisrobot,over here,which has four
14、legs,it has eight motors,fouron the knees and four on the hip.现在我还要再给你们展示另一项进展，就是这里的这个机器人，在这儿，它有四条腿 它身上安了八个马达，膝盖上四个，腿上四个02:42It has also two tilt sensors that tell the machine which wayits tilting.它还配备了两个倾斜度传感器，可以告诉自己 正在向哪个方向倾斜。02:52competing:adj.相互矛盾的;抵触的/v.竞赛(compete的ing形式)evolving:adj.进化的;展开的/v.进
15、化;展开(evolve的ing形式)rodeo:n.竞技表演;竞技者/vi.竞技 autonomously:adv.自治地;独立自主地 tilt:vi.倾斜;翘起;以言词或文字抨击/vt.使倾斜;使翘起/n.倾斜 sensors:n.自传感器,感应器;感测器(sensor的复数)tilting:n.倾卸台/adj.航气象倾斜;倾卸/v.使倾斜(tilt的ing形式)But this machine doesnt know what it looks like.但这个机器并不知道自己长啥样02:58You look at it and you see it has four legs,the m
16、achinedoesnt know if its a snake,if its a tree,it doesnt have anyidea what it looks like,but its going to try to find that out.你可以看到它长了四条腿，但它自己并不知道自己是一条蛇还是一棵树 它完全给蒙在鼓里，不晓得自己的相貌 但它马上就要试着找到自己03:00Initially,it does some random motion,and then it tries tofigure out what it might look like-and youre seei
17、ng a lot ofthings passing through its minds,首先，它会做一些随机的动作 然后试着弄清楚那些动作都看起来是什么样子的 你可以看到它的脑海里闪过许许多多的东西，03:09a lot of self-models that try to explain the relationshipbetween actuation and sensing-and then tries to do asecond action that creates the most disagreement amongpredictions of these alternative
18、 models,like a scientist in alab.Then it does that and tries to explain that,and prune outits self-models.大量的自我尝试的动作模型，试着理清 行动和感官之间的关系然后它将再做第二个动作 在所有可能的动作模型中 最诡异的一个动作，就像科学家在实验室里的试验。接着，它重复那个动作，并且试着解释那个动作，然后梳理出自己的动作模型。03:16This is the last cycle,and you can see its pretty much figuredout what its self
19、 looks like,and once it has a self-model,it canuse that to derive a pattern of locomotion.这是最后一个环节，你可以看到它已经基本上 清楚自己的样子了。一旦它理清自己的动作模型，就可以从模型中得出一种运动模式。03:35Initially:adv.最初，首先；开头 actuation:n.冲动,驱使;刺激;行动 sensing:n.传感;感觉;测知/adj.敏感的/v.感觉,了解(sense的现在分词)predictions:n.预测,预言(prediction复数形式)prune:vi.删除;减少/vt.
20、修剪;删除;剪去/n.深紫红色;傻瓜;李子干 derive:vt.源于;得自/vi.起源 locomotion:n.运动;移动;旅行So what youre seeing here are a couple of machines-apattern of locomotion.好了，你现在看到的是几个机器 嗯，一种运动模式，03:46We were hoping that its going to have a kind of evil,spiderywalk,but instead,it created this pretty lame way of movingforward.我们期待它
21、能产生一种邪恶的，蜘蛛式的运动，结果它却自创出这种相当脑残的前进方式。03:50But when you look at that,you have to remember that thismachine did not do any physical trials on how to moveforward,nor did it have a model of itself.但当你看着它前进的时候，你必须记得 这个机器并没有接受任何物理指令，控制着它们向前进，它也没有任何已有的自我模型。03:58It kind of figured out what it looks like,and h
22、ow to moveforward,and then actually tried that out.它相当于是自己推理出了自己的样子，以及应该如何向前进 并且进行了亲身的尝试。04:09(Applause)So,well move forward to a different idea.鼓掌?那么现在，我们再来看看另一个想法。04:16So that was what happened when we had a couple of-thats what happened when you had a couple of-OK,OK,OK-那是我们将几个.把几个.放在一块儿就会.好啦好啦好啦
23、04:25(Laughter)（笑）04:34-they dont like each other.So theres a different robot.它们不大喜欢对方，所以啦 这是另外一个机器人。04:36Thats what happened when the robots actually are rewardedfor doing something.刚才的那些都是在机器人做对了动作，获得奖励的情况下发生的。04:41What happens if you dont reward them for anything,youjust throw them in?那么如果我们不给它们奖
24、励，只是把它们扔到一块，又会怎么样呢？04:45spidery:adj.细长足的;蜘蛛一般的;蜘蛛网一般的So we have these cubes,like the diagram showed here.所以我们拿来了这些立方体，就像这里的这些图，04:48The cube can swivel,or flip on its side,and we just throw1,000 of these cubes into a soup-this is in simulation-anddont reward them for anything,we just let them flip.W
25、epump energy into this and see what happens in a couple ofmutations.它们能旋转，或者翻筋斗 我们把1000个这样的立方体放入“原汤”这是模拟效果我们没给它们任何奖励，我们就让它们自己活动。我们给它们注入了些能量，看看经过几次突变，会发生点什么。04:51So initially,nothing happens,theyre just flipping aroundthere.刚开始的时候，什么也没发生，它们光在那儿跳来跳去。05:06But after a very short while,you can see these
26、blue things onthe right there begin to take over.但过了一小会儿，你可以看到这些蓝色的小东西，它们在右边逐渐地开始占取主动。05:09They begin to self-replicate.So in absence of any reward,the它们开始自我复制。由此可见即使没有任何奖励They begin to self-replicate.So in absence of any reward,theintrinsic reward is self-replication.它们也会用自我复制的方式来奖励自己。05:15And weve
27、 actually built a couple of these,and this is part ofa larger robot made out of these cubes,its an acceleratedview,where you can see the robot actually carrying out someof its replication process.事实上我们已经造了好几个这类的玩意儿，这是一部分以这些立方体为单位造出来的大机器人，这是快进的效果，可以让你看到机器人 进行自我复制的过程。05:22So youre feeding it with more
28、 material-cubes in this case-and more energy,and it can make another robot.如果你给它多喂点儿就是这些立方体 再多给它点能量，它就能自己造出另一个机器人。05:32cubes:n.数立方体,数据库(cube的复数形式)/v.使成立方形(cube的第三人称单数形式)swivel:n.转环;转椅座架/vt.使旋转/vi.旋转 mutations:n.遗突变;变化;转变(mutation的复数形式)flipping:adj.非常的;糟透的/adv.非常;该死地/v.轻弹;急动;蹦蹦跳跳(flip的ing形式)intrinsic
29、:adj.本质的，固有的 accelerated:adj.加速的;加快的/v.加速;促进(accelerate的变形)replication:n.复制;回答;反响So of course,this is a very crude machine,but were workingon a micro-scale version of these,and hopefully the cubeswill be like a powder that you pour in.当然，这还是一个非常粗糙，不成熟的机器，但我们正在研究微缩版的这类机器人，希望这些立方体能小到像倒出的面粉一般。05:39OK,s
30、o what can we learn?These robots are of course not veryuseful in themselves,but they might teach us somethingabout how we can build better robots,and perhaps howhumans,animals,create self-models and learn.好的，那么我们都了解到了什么？这些机器人当然自己并不是有很大用处，但它们能教会我们一些东西，让我们知道如何造出更好的机器人，甚至是人类，动物创造自我模型和学习机制的原理。05:47And o
31、ne of the things that I think is important is that we haveto get away from this idea of designing the machinesmanually,but actually let them evolve and learn,likechildren,and perhaps thats the way well get there.Thankyou.我觉得这其中最重要的，就是我们必须摒弃之前的观念，手动地设计这些机器 而是让它们自己进化，学习，像孩子一样，这大概才是我们成功的必经之路。谢谢！06:03(Applause)（鼓掌?）06:14pour in:大量地涌入,倾倒;蜂拥而来 manually:adv.手动地;用手*Warning:本文是由生成导出，请用于个人学习，不要用于商业用途。否则，导致的一切法律后果，均由您个人承担，锡育软件概不负责。*