Interviewer: Lucy Vernall (Project Director, Ideas Lab)
Guest: Dr Nick Hawes
Recorded: 02/08/2011
Broadcast: 08/08/2011
Intro VO: Welcome to the Ideas Lab Predictor Podcast from the University of Birmingham. In each edition we hear from an expert in a different field, who gives us insider information on key trends, upcoming events, and what they think the near future holds.
Robotic Voice: Today, we’re with Dr Nick Hawes, Lecturer in Intelligent Robotics at the University of Birmingham.
Lucy: Don’t worry, the file hasn’t been corrupted, that was actually the Aldebaran Nao robot who’s joining us for our podcast today, along with Dr Nick Hawes who is Lecturer in Intelligent Robotics. Welcome, Nick.
Nick: Hi.
Lucy: So Intelligent Robotics, what is that?
Nick: So it’s the intersection of robotics, so the kind of science and engineering field of building machines to kind of work with humans or work for humans, and artificial intelligence, the science of building smart things, intelligent machines and intelligent systems.
Lucy: So you spend your day tinkering with robots?
Nick: Yeah, well tinkering with software mainly, writing the kind of brains of the robots effectively to make them do the things you might want them to do in the future.
Lucy: It seems like we’ve been promised kind of robot butlers and intelligent robots to help us to do everything for decades but they never quite materialise do they?
Nick: No, the promises science fiction has made we can’t quite deliver on as a field. Building things that are as smart as humans has just proved to be a very very big problem because just the challenges involved in that in the sensing, in the action, in building the machines and building intelligence, there’s just so much in there that it’s not really possible to build these generally intelligent machines.
Lucy: It’s quite reassuring to know that we can’t be replaced just yet.
Nick: Well not just yet. I mean I think there might be cases where limited functions of humans could be replaced but they’re not going to rule over us yet.
Lucy: No. But robotics is growing as a field.
Nick: Yeah, I meant there’s been a significant growth in different strands of robotics, so you can think of industrial robots, so the kind of robots you’d find in car factories, there are millions, maybe something in the order of like 15 or 18 million robots in use and most of them are these kind of factory robots so the potential growth areas are service robots, so robots that can perform functions of humans, services for humans, and they are growing steadily, particularly the ones for our home. I think we’re looking at maybe one or two million new robots appearing in our homes in the next two years.
Lucy: That’s exciting but what exactly are we going to get? What are they going to be able to do for us?
Nick: Again, special purpose. So these general purpose intelligent robots, these general robot butlers, aren’t going to appear but what we’ll see are robots that can do one thing and one thing well. So already we’re getting vacuum cleaning robots, so in, I think 2009, there was like a million vacuum robots sold and something like 27,000 lawn mowing robots.
Lucy: Where have they been sold? Do you know anyone with a vacuum robot? I don’t know anyone.
Nick: Yeah, I know quite a lot of bachelors with vacuum cleaning robots.
Lucy: Are they bachelor computer scientists?
Nick: Yeah, most of them! There aren’t many other kinds.
Lucy: [laughing]
Nick: So yeah, special purpose robots, robots that can do one thing and one thing well. They can do hoovering, mopping, lawn mowing and I think we might see other special purpose robots outside of the home in areas where we wouldn’t necessarily think of robots being useful, so transport – our cars are turning into robots bit by bit. There's a lot of intelligent systems in a car that you would also kind of imagine as being part of robot functionality as well.
Lucy: As well as the robot we heard earlier on we are joined by another robot over in the corner of the room here. I’m told this is Dora.
Nick: Yeah, Dora the Explorer.
Lucy: Dora the Explorer. What can Dora do?
Nick: So Dora is a sort of prototype for one of these more generally intelligent robots. The idea of Dora would be to kind of explore what needs to be done to have a robot that could arrive and be unpacked from Amazon or Argos and then could start performing tasks for you in the home. So in particular the research we’re doing with Dora is funded as part of an EU project called CogX and we’re looking at ways that the robot can extend its own knowledge of the world. So autonomously explore its environment and be curious about the world it’s in.
Lucy: So that means you can put Dora in your home, provided it’s a bungalow or a flat! [Laughing]
Nick: Yeah, no stairs unfortunately!
Lucy: No stairs! And she can have a wander round and kind of start to get familiar with it.
Nick: Yeah, start to build up a map, start to look around, start to look for objects, try and learn what types of rooms they are so it’s really useful for Dora to know, if it’s going to perform tasks for you like make your breakfast or get you a cup of tea, it’s very useful for Dora to know where the kitchen is or where your living room is or where your bedroom is. So Dora can explore to find these areas and then look for objects that allow it to determine what type of room it is. So if it finds a kettle it might say ‘oh I’m in the kitchen’.
Lucy: That sounds great but she can’t at the moment make you a cup of tea can she?
Nick: Well she hasn’t got any arms.
Lucy: No arms!
Nick: Yes.
Lucy: Kind of crucial!
Nick: Yeah. I can’t defend the obvious! Even though there are no pictures – Dora has no arms. We do put arms, or a single arm, on the base but the problem of doing manipulation, it’s one of the toughest problems in robotics. So to actually manipulate something you need to be able to combine vision, so working out where the object is, with manipulation, so actually reaching out to grasp it. And the kind of errors you get there, the ability to estimate the position of an object just from vision is very very difficult. But there have been some interesting developments in sensing recently that are allowing us to kind of go beyond this a little bit. The Microsoft Kinect system, which is a sort of 3D sensing system for games, is actually turning into a huge new technology in the robotics field because it allows you to get 3D images effectively of the world using a very very cheap sensor. So at the moment Dora’s got one of these things strapped to her head and it’s allowing us to see things in Dora’s world that we couldn’t do before. So for example desks and tables weren’t visible to Dora before because Dora just uses a laser scanner at sort of knee height and things that aren’t bulky at knee height just get missed. So using things like vision and using things like the Kinect allow us to build up a richer map and get more information about the environment.
Lucy: And even recognising things is difficult because they need to be very solid, regular type objects for her to know what they are.
Nick: Yeah, well there’s two types of constraints, either solid and regular, so kind of clear geometric shapes, or highly textured or highly patterned. So very distinctive visually is also useful. There are sort of two different techniques you can use and it has to be one or the other.
Lucy: So things that would work well are things like cereal boxes.
Nick: Yeah, cereal boxes. There’s actually a Coco Pops box just behind her, not because I eat cereal in my office, but because these are great things for robots to be tested with.
Lucy: And the difficult things would be things like towels?
Nick: Towels because material is very irregular and also untextured. Also glass because light just passes through it which means that you can’t see it with Kinect because the infrared that Kinect uses doesn’t pick it up. You can’t really see it with vision because it’s very very hard to extract the edges of the objects using just kind of the image alone and that’s where humans again have a much greater advantage over robots because all of our common sense and our experience of the environment allows us to infer the presence of objects even though our kind of senses don’t give us conclusive evidence.
Lucy: So you’ll get your robot from Argos or Amazon, you’ll unpack it and it’ll go straight through the window! [laughing]
Nick: Yeah, if you’ve got a – it will be like a cat or a child crashing into a French window in your house.
Lucy: It does make you realise how brilliant humans are.
Nick: Yeah, I think that’s one thing this field teaches you, it’s that humans are incredible and in fact children even more so. When you see the way that children develop and learn about the world it makes you kind of despair sometimes for the robots we’re building.
Lucy: So the million dollar question, when are we going to see some of these things? OK, we’ve got the vacuum robots but aside from that, when are these things going to happen, really really?
Nick: I don’t, it’s hard to predict. The field itself is making great leaps forward. There's a company called Willow Garage who are for instance building and selling personal robots now, mainly for the research community, but these are robots that can fold towels and can do very very kind of scripted tasks but do quite advanced scripted tasks like making pancakes for example, but they can’t do the general things. This is a kind of trend that is leading us towards these robots in the home. It’s very awkward to try and put a number or a date on it. I think in five to ten years we’ll see that robots will be more and more commonplace and maybe not in our homes but maybe in places of work or in businesses.
Lucy: So what’s next for you in the more foreseeable future?
Nick: What interests me at the moment is looking at these hurdles we have to overcome to take the technology we’ve got in the lab at the moment and applying it to real world situations. So one of the interesting challenges for me is the sort of longevity of robots. At the moment Dora for instance can run for an hour or two but that’s about it. They’re not really built or designed from sort of an artificial intelligence perspective to run for long periods of time. So looking at issues of, for example, memory so episodic memory or long term memory as humans have and applying that to robots to try and get them, for example, to learn where you typically put your keys or where you typically find things like kettles in the house and building that experience over long periods of time to allow a robot to be a useful companion or assistant.
Lucy: And towards the end of the year we’ve got European Robotics Week.
Nick: Yeah, it’s an exciting week where us as scientists doing robotics can promote our work to sort of the rest of the world or the rest of Europe. It’s running from November 28th to the 4th December and there should be events all across Europe. We’re taking part in an event at the Science Museum called the Festival of Robots and Dora will be running there and there will be open events for schools and things.
Lucy: So that’s the Festival?
Nick: The Festival of Robots. I mean there's going to be I think twenty or thirty different robots, sort of cutting edge, state of the art things running there, also talks and demonstrations of the basic technologies.
Lucy: Great, so that’s 28th November to the 4th December at the British Science Museum and you’ll be there.
Nick: I’ll be there.
Lucy: [laughing] And so will Dora.
Nick: I hope so, if she makes the journey successfully.
Lucy: Dr Nick Hawes, thank you very much.
Nick: Thanks.
Outro VO: This podcast and others in the series are available on the Ideas Lab website: www.ideaslabuk.com. On the website, you can find out how to e-mail us with comments, questions or suggestions for future topics for the podcast. There's also information on the free support Ideas Lab has to offer to TV and radio producers, new media producers and journalists. The interviewer for the Ideas Lab Predictor Podcast was Lucy Vernall, and the producer was Andy Tootell.