- Title: A robotic hand with a human touch
- Date: 13th May 2016
- Summary: SEATTLE, WASHINGTON, UNITED STATES (MAY 12, 2016) (REUTERS) PERSON PLAYING PIANO
- Embargoed: 28th May 2016 17:53
- Keywords: robot robotic hand IEEE Conference computer learning AI
- Location: SEATTLE, WASHINGTON, UNITED STATES
- City: SEATTLE, WASHINGTON, UNITED STATES
- Country: USA
- Topics: Science
- Reuters ID: LVA0034HN6TZV
- Aspect Ratio: 16:9
- Story Text: Human hands are capable of incredible dexterity. They can learn the complicated moves needed to play the piano and have the speed required for keyboard skills such as typing. Recreating such dexterity in a robotic hand has proved to be a difficult challenge.
But now a robotic hand that mimics the human hand has been developed at the University of Washington in Seattle. Researchers in the Movement Control Lab at the University's Dept. of Computer Science & Engineering have been working on the project for several years.
Graduate student Vikash Kumar explained to Reuters what was required to design a successful robotic hand.
"So in order to manipulate any object what is required is three key ingredients. One of them is ability to act really fast because in case you are losing control of an object you want to quickly go and re-position yourself. So the ability to act fast is required in those situations when you're about to fail and you want to make a correction maneuver," he said.
"You have to look at the dexterity of the hand. That hand should be capable of doing a lot of different things and second, which is the most important piece of the puzzle, we need to have some kind of an intelligent unit that knows how to exploit that dexterity in order to solve those problems," he added.
The solution, according to the researchers, starts with computer software that drives a series of actuators, each controlling the cables that manipulate the hands' fingers and thumb. The hand that the University of Washington team has designed can operate at great speed, a key requirement if the hand is to have human-like reflexes.
What sets this robotic hand apart from earlier models is its ability to learn as it attempts to carry out a task, in much the same way humans learn 'on the job'. With each attempt at a task the hand improves its performance. While other robotic hands can usually only perform one task this hand could, potentially, perform unlimited tasks. Assistant Professor Sergey Levine wrote the algorithm that the hand uses to learn.
"If we can get robots that can learn autonomously, they can learn on their own, then we don't need to figure out ourselves how the robot should perform the skill. We just need to define the skill and have the robot figure it out on its own. I think that's tremendously powerful because then we can have robots that can figure out on their own how to do a wide variety of behaviors," he said.
The leader of the Movement Control Lab, Professor Emo Todorov, believes that colleagues from around the world attending the Robotics & Automation conference in Stockholm will sit up and take notice when they see how the hand deals with difficult tasks and teaches itself to handle them.
"Nobody had to hack it, nobody had to script it, the thing just learned all on its own. To us it's not surprising that it works because we knew it was going to work all along but people who haven't seen it, for them that might be a convincing proof of principle," he said.
Although there's a long way still to go, the commercial possibilities of a robotic hand that does not need programming but which can learn new tasks 'on the fly' are clear.
Sergey Levine outlined its potential. "We don't see a lot of consumer robots for instance, we don't see robots in the home and part of the reason for that is those settings are unstructured, they're not arranged in advance for the robot like a factory is. So I think in order to succeed in those unstructured environments learning will be a crucial component because we need robots to deal with uncertainty, to deal with variable environments, with a lot of complexity they simply don't see in a factory setting."
For now the hand continues to improve at the one task it's been set. The researchers next challenge is to design software that will enable it to cope with any challenge it is given. Only then can it truly be described as a 'robotic human hand'. - Copyright Holder: REUTERS
- Copyright Notice: (c) Copyright Thomson Reuters 2016. Open For Restrictions - http://about.reuters.com/fulllegal.asp
- Usage Terms/Restrictions: Audio restrictions: This clip's Audio includes copyrighted material. User is responsible for obtaining additional clearances before publishing the audio contained in this clip.