University of Washington
Artificial Intelligence

This Five-Fingered Robot Hand Is Nimbler Than Your Own

Machine learning + an artificial hand = one dexterous robot.

Robot Hands

Robots have done and are able to do some of the most amazing tasks we never thought we could trust to a machine. Now, robots have a hand in healthcare, perform delivery tasks by themselves, even milk cows.

But one thing they have never been really good at is using hands. You get these amazing robots that look like humans, talk like humans, and have the pincers of a crab.

Until now.

A five-fingered robot developed at the University of Washington is capable of dexterous hand movements, and the ability to learn from its own experiences without human intervention.

In a paper to be presented at the IEEE International Conference on Robotics and Automation, the team revealed their creation of an accurate simulation model that allows a computer to analyze movements in real-time.

They married this program to their creation of a five-fingered robotic hand with the combination of speed, force
and compliance required for dexterous manipulation. Their latest demonstration showcased this merger of hardware and software, allowing the bot to do real-world tasks like spinning an elongated tube.

Getting a Grip

The making of the robot of course begins with hardware. The five-fingered robot—built at a cost of roughly $300,000—incorporates a Shadow Dexterous Hand skeleton and a custom pneumatic system. This allows it to move faster than a human hand.

This special set-up means that it is too expensive for routine commercial or industrial use, but the increased dexterity allows researchers to push core technologies and test innovative control strategies.

The other important aspect of the robot is the software package. This was developed by first creating algorithms that made the robot able to do complex hand tasks—in simulation. These algorithms were then plugged into the robot and made to do the tasks in real life.

As the robot performs all of the tasks, and fails at them, the system collects data from various sensors and motion capture cameras and employs machine learning algorithms to continually refine and develop more realistic models.

So goodbye pincer hands. Looking forward to the first handshake with a robot.

Keep up. Subscribe to our daily newsletter.

I understand and agree that registration on or use of this site constitutes agreement to its User Agreement and Privacy Policy
Next Article
////////////