-
(单词翻译:双击或拖选)
Robyn Williams: And this is The Science Show on RN where we just heard Amanda Bauer talk about all those robots running the giant instruments. Here's one of them.
Amanda Bauer: This is the robot. The robot reaches down and clamps onto one of these little metal pieces and moves the optical fibre to a place on the metal plate and drops it, and its little magnet allows it to stick very precisely1. Each one of these fibres has a little light that comes out of it, so the robot can detect that light and know exactly how accurately2 it's pointing and pick it up and adjust it and then move it a little bit.
Robyn Williams: I suppose if you did it by hand it would take about four years.
Amanda Bauer: Well, have you heard of the Sloan Digital Sky Survey? This is a big survey that is based in New Mexico, and they were developing their technology about the same time this was, and they decided3 to have big metal plates that they drilled holes in, and they have humans plugging these optical fibres into those plates, and that's what they've been doing for 20 years. So you never quite improve your efficiency that way, whereas we decided to go with the robotic technology. And robots are a bit finicky. They have little attitudes, but we have robot whisperers, as you see around here, who know how to talk to it.
Robyn Williams: A chain is being moved…
Amanda Bauer: Technical things are happening! So with the robotic technology we can still improve the software and improve aspects of it over the years. So it's a bit better now than it was when we first developed it, but it's a funky4 little robot.
Robyn Williams: Amanda Bauer with the robot running the telescope. But many other robots don't seem quite so funky. Over in Birmingham for instance is one that is trying to pick up objects you'll find in your kitchen. You and I pick them up and put them down without thinking, but this robot…well…
So you are about to give me an object to put on the table in front of the robot. Any particular one? The funnel5, the pan?
Jeremy Wyatt: Pick which you like.
Robyn Williams: I rather like the funnel, it's a red funnel. So I put it over there…
Jeremy Wyatt: We're going to put it on here first, and now you put the final on top.
Robyn Williams: There we are…
Jeremy Wyatt: Oh, see, you've broken it already…
Robyn Williams: I've broken the…aw! Let me just to describe, while you're trying to work out what needs to be fixed6, it's huge, each arm has got two arms and they are something like twice the size of a normal human arm, lots of joints7, and a kind of head up the top with seeing eyes perhaps. Anyway, tell me who you are first and where we are.
Jeremy Wyatt: So I'm Jeremy Wyatt, we are at the University of Birmingham in the robot lab.
Robyn Williams: Right, and you've got a whole number of robots. Which is this one?
Jeremy Wyatt: This is Boris.
Robyn Williams: Boris…but he doesn't have a blond mop.
Jeremy Wyatt: He doesn't have a blond mop, no, I'm not going to draw comparisons with the Mayor of London. They look like quite different characters to me.
Robyn Williams: Right, okay. And what you are trying to do here is pick up an object, and this is an object that is unexpected. In other words, it's not the kind of routine picking up and movement that we can see in places where robots build motorcars, this is de novo, this is something that it's got to learn and home in on afresh. So you are teaching it to learn in many ways, are you?
Jeremy Wyatt: Yes, exactly. So what we're doing is we showed it one object and we showed it exactly how to grasp that one object, and then what we are doing here, as you say, de novo, we are showing it a new object which it hasn't seen before, and you cunningly put the funnel upside down, and so that's going to break it in this case, but if we turn the funnel the other side up it will figure out how to grasp it, because this particular grasp, it knows how to grasp a rim8.
Robyn Williams: Can I dare do it without breaking it again? Is that okay?
Jeremy Wyatt: That's looking pretty good.
Robyn Williams: And it will decide at some point…
Jeremy Wyatt: Up on the screen over there you can see its little hands appearing, and those hands are where it thinks it might be able to grasp it.
Robyn Williams: So we are looking at its brain, are we, in other words?
Jeremy Wyatt: You are looking at a visualisation of what it's thinking about.
Robyn Williams: Okay. The tension is killing…beginning to…no, no, it's just still thinking…
Jeremy Wyatt: You'll know when it wants to do something because by looking at this screen over here it will tell you it wants to go and grasp it.
Robyn Williams: Is it a real challenge, teaching it to do something new every time?
Jeremy Wyatt: Yes, I think is the one-word answer to that! The interesting thing is building it so that it's robust9 even to a relatively10 small range of variation. So it's a real challenge to build a robot that is robust even to variations that to a human are trivial. So, for example, if you pick up a glass, it's all shiny. That foxes current robots. Metal, shiny, that's also really hard. Things that are too brittle11, things that are too soft. So to bring robots out into the real world, into our world, to get them to work with us, you've got to cope with all of this variation…
Robyn Williams: It's doing it now, it's doing it now! It has picked up the funnel, it's grasping it, a big hand, lifting it up, and the task is to bring it across to a tray full of polystyrene where presumably it will be placed securely. Moving towards the tray, and up and down it comes…drops. And that's good, is it?
Jeremy Wyatt: That's good, I'm happy with that.
Robyn Williams: What is the task that it might do when you've got full computerisation of the actions and so on? What can it do in the household, for example?
Jeremy Wyatt: So the task of this project, the objective is to get the robot to load a dishwasher tray. So you going to have a bunch of objects on the table, scattered12 as you might have them on a regular kitchen surface, then the robot is going to look through the set of objects, find the one it wants to pick up, figure out where to put it into the dishwasher, and load it. And that is not because we think that dishwashing-loading robots are a critical social need…
Robyn Williams: You could have a Polish au pair to do that for you more easily.
Jeremy Wyatt: I couldn't possibly comment. But the thing is that it's a typical task that humans engage in that requires all of your manipulative faculties13 that evolution spent hundreds of millions of years developing. And so by being able to put that into a robot we hope to make the robots more flexible in the future. So immediately it's going to do the dishwasher loading task.
Robyn Williams: Now, a naive14 question; we already have driverless cars which can adapt to presumably the unexpected in a busy highway. How does that compare with what I've got in front of me in terms of robotics?
Jeremy Wyatt: In terms of the machine learning…so machine learning has really been the revolution in all kinds of computing15, so you can see a lot of data mining, you can see self-driving cars, robots like this, they are all doing machine learning in one form or another, so in that way they are the same. What's a little bit different here is that you are coming into contact with the world to change it, whereas in the car you are moving along the road and you are trying not to change the world, you're trying not to touch things other than the road surface with the wheels.
And so what we are doing here is we are really trying to alter the world, and that's what humans do because we are tool-using animals, that's how we made our world. And so it is thinking about almost a kind of a third generation of robots, because the first generation was industrial robots and they manipulate the world when it is very precisely controlled. And now what we are getting is drones and we are getting self-driving cars and mobile robots that can move around in our world and share it with us safely, even though that world is very uncertain and novel to them. But the moment you want to contact the world and change it, manipulate things in your fingers, use tools, make stuff, assemble things, work with people, then you've got to be able to contact it and that then makes it much, much harder again if you are going to do it under uncertainty16.
Robyn Williams: Well, that's Boris. What sort of other robots do you have in this laboratory?
Jeremy Wyatt: So if we walk over here we've got Baxter. So Baxter has come from America. So Boris we built here, but Baxter is an American robot. And whereas Boris cost about $350,000, the Americans have made Baxter for about $20,000. So this is a really good example of trying to get some of the tech that actually in future we are developing in Boris could go into a robot like this because this is meant to reshore small-scale manufacturing. So that was the idea of Rod Brooks17, its inventor.
And it's a big red robot with two arms, and these arms are very floppy18 and flexible. So what you can do is you can teach the robot by moving the arms around and showing it how to do tasks. And it can pick things up. It also has little cameras underneath19 its wrists. So that one is, again, not so much for assembly but for picking and placing things in production lines.
Robyn Williams: Again he is just about the same size as a fairly hefty human. And Rodney Brooks is in fact an Australian at MIT, is he not?
Jeremy Wyatt: He is, yes. And over here we have an educational robot which is based on a French platform called a Nao, made by a company called Aldebaran, and that robot, what it's doing is some children when they are learning they get a lot from interacting with a robot. So some children prefer to interact with a robot to a human, and so what happens the robot is used to give feedback to the children on what they are doing.
And then outside we've got our mobile robot, Bob. And Bob is a robot who we've been…well, he's been on work placement as a security guard at G4S, the security company. And he's developing technology for running over very long periods of time. So a very long period of time for a research robot is anything more than a day, because trying to make these systems reliable, particularly when they are encountering humans, because humans always do something that you don't expect. Bob has got to cope with that. And so Bob tries to build up a picture over time by learning how the humans behave, so then he can adapt to the way that the humans work.
Robyn Williams: How far off are you from doing something that's almost like a human in its adaptability20 and flexibility21?
Jeremy Wyatt: I think we are possibly still a really long way. But the message here is that even a little bit of adaptability and flexibility goes a really long way. So we are going to have robots coming with just a little bit of intelligence and they are going to solve little tasks for us. So, for example, James Dyson has just announced that he is going to release his new robot vacuum cleaner. That's going to be the first robot vacuum cleaner in the world that can see and can recognise where it is in the room. So it's just a little bit of intelligence but it's really, really useful for the task of doing the robot vacuum cleaning and going back and cleaning up the bits of the house that you know people tend to leave really dirty.
Robyn Williams: Okay, you're in Birmingham, you're in the computer department here. How would you compare Britain with robotics to, say, Germany, you mentioned America, in the league table?
Jeremy Wyatt: In terms of research the UK actually does really well. We have lots of really interesting ideas coming out of the labs. The Germans are really, really good at industrial robotics and they are the powerhouse in the world. I mean, other than the Japanese, Germany makes all the world's industrial robots, so they are brilliant at that.
In America they do a lot of work in military robots and they also have some wonderful labs. Australia is actually the hotbed for field robotics. So down in Australia, particularly for the mining industry, there's a huge amount of work going on in field robotics there, and they are amazing at that.
Robyn Williams: Thank you.
Jeremy Wyatt: Thank you.
点击
收听单词发音
收听单词发音 |
1
precisely
|
|
| adv.恰好，正好，精确地，细致地 | |
参考例句： |
|
|
|
|
2
accurately
|
|
| adv.准确地，精确地 | |
参考例句： |
|
|
|
|
3
decided
|
|
| adj.决定了的，坚决的；明显的，明确的 | |
参考例句： |
|
|
|
|
4
funky
|
|
| adj.畏缩的，怯懦的，霉臭的；adj.新式的，时髦的 | |
参考例句： |
|
|
|
|
5
funnel
|
|
| n.漏斗；烟囱；v.汇集 | |
参考例句： |
|
|
|
|
6
fixed
|
|
| adj.固定的，不变的，准备好的；(计算机)固定的 | |
参考例句： |
|
|
|
|
7
joints
|
|
| 接头( joint的名词复数 )； 关节； 公共场所（尤指价格低廉的饮食和娱乐场所） （非正式）； 一块烤肉 （英式英语） | |
参考例句： |
|
|
|
|
8
rim
|
|
| n.(圆物的)边，轮缘；边界 | |
参考例句： |
|
|
|
|
9
robust
|
|
| adj.强壮的，强健的，粗野的，需要体力的，浓的 | |
参考例句： |
|
|
|
|
10
relatively
|
|
| adv.比较...地，相对地 | |
参考例句： |
|
|
|
|
11
brittle
|
|
| adj.易碎的；脆弱的；冷淡的；(声音)尖利的 | |
参考例句： |
|
|
|
|
12
scattered
|
|
| adj.分散的，稀疏的；散步的；疏疏落落的 | |
参考例句： |
|
|
|
|
13
faculties
|
|
| n.能力( faculty的名词复数 )；全体教职员；技巧；院 | |
参考例句： |
|
|
|
|
14
naive
|
|
| adj.幼稚的，轻信的；天真的 | |
参考例句： |
|
|
|
|
15
computing
|
|
| n.计算 | |
参考例句： |
|
|
|
|
16
uncertainty
|
|
| n.易变，靠不住，不确知，不确定的事物 | |
参考例句： |
|
|
|
|
17
brooks
|
|
| n.小溪( brook的名词复数 ) | |
参考例句： |
|
|
|
|
18
floppy
|
|
| adj.松软的，衰弱的 | |
参考例句： |
|
|
|
|
19
underneath
|
|
| adj.在...下面，在...底下；adv.在下面 | |
参考例句： |
|
|
|
|
20
adaptability
|
|
| n.适应性 | |
参考例句： |
|
|
|
|
21
flexibility
|
|
| n.柔韧性，弹性，（光的）折射性，灵活性 | |
参考例句： |
|
|
|
|
22
brute
|
|
| n.野兽，兽性 | |
参考例句： |
|
|
|
|
23
cosmos
|
|
| n.宇宙；秩序，和谐 | |
参考例句： |
|
|
|
购买
在百度中搜索