USA-INTUITIVE ROBOTS Sous chef of the future? Robot learns to slice vegetables using a knife

Record ID: 401940

  • Title: USA-INTUITIVE ROBOTS Sous chef of the future? Robot learns to slice vegetables using a knife
  • Date: 6th October 2014
  • Summary: KODIAK, INTUITIVE PR2 ROBOT (SOUNDBITE) (English) IAN LENZ, PH.D. STUDENT OF COMPUTER SCIENCE AT CORNELL UNIVERSITY, SAYING: "I think it's still a while out until the personal robotics setting. I could see things like this making it into more controlled settings like maybe you have a food prep robot that you use in a kitchen or something like that because it's a more contr
  • Embargoed: 21st October 2014 13:00
  • Keywords:
  • Topics: General
  • Reuters ID: LVAA44JIJZZ67U68MDVXC3PP72V3
  • Story Text: Having your own personal robot to help with kitchen work may be closer to reality than you think.
    Kodiak, a PR2 robot, is slicing a cucumber using a sharp knife. While it looks easy enough, the robot is actually figuring out 'intuitively' how to hold the knife and the cucumber in the right place.
    Researchers at Cornell University's Robot Learning Lab have been developing new algorithms and software that will enable robots like Kodiak to figure out how to work in unstructured human environments and adapt to individual's preferences.
    Ph.D. student Ian Lenz has been working on Kodiak, hoping he'll one day help him cook dinner. He says that rather than having a robot that someone programmed, you could have a robot that does exactly what you want.
    "It would have some intuitive understanding of how you work. And it would start out maybe with an idea of how in general people work but then it figures out you're this kind of person so I'm going to work this way for you," said Lenz.
    For instance, he explained, "I like to cook but when I get home right now I'm tired and I don't want to go home and cook myself a meal on a weekday. But if the robot has done all the prep work for me, he's cut all the food and all that for me, then great, alright, I'll come home, I'll throw it in the frying pan, I'll cook it. It's things like that, it's just those little things that maybe you don't think about that would really make your life a lot better to actually have the robot doing it for you rather than having to do yourself."
    Lenz and other researchers are using a technique called 'machine learning', that mimics the same process of learning used by humans, which is based on experience and memory.
    They program the robot to downloaded images from the Internet about the human environment, including appliance manuals and language samples. Then the robots use a machine-learning algorithm to process the information they see in their new environment. Based on its stored 'memory' the robot is able to infer how to manipulate similar devices.
    "The idea is to let the learning algorithm do those things more and more for us so rather than I told it exactly where to move it, it just figured out this is a good place to put it on the cutting board because now I know I can get at it with a knife and then I know I'm going to start cutting so I know I can cut it this way," Lenz explained.
    The robot has some special features, including hi-tech cameras that perceive depth, and sensors on their fingertips that indicate to the robot how much pressure is being felt on its fingertips.
    Lenz says scientists are still quite a way from bringing these 'thinking' bots to the masses, but says this isn't very far away.
    "I could see things like this making it into more controlled settings like maybe you have a food prep robot that you use in a kitchen or something like that because it's a more controlled environment you can just give him his corner of the kitchen, you have people that are trained to work with him, so it's a lot safer to have him around then the general home setting."
    Kodiak isn't quite ready. While he's cutting slowly and has figured out the right placement and slicing motion, he still has a few kinks to work out before he can rival Lenz's kitchen proficiency.


  • Copyright Holder: REUTERS
  • Copyright Notice: (c) Copyright Thomson Reuters 2014. Open For Restrictions - http://about.reuters.com/fulllegal.asp
  • Usage Terms/Restrictions: None
Register