Manufacturing Minute: A Robot That Can Teach Itself How To Grip

Подробнее о видео

In this episode, a five-fingered robot overcomes the biggest challenge facing similar tech around the world: dexterous manipulation. Welcome to your Manufacturing Minute.

Research and development in the robotics field is expanding rapidly. One problem seems to be universal, though, and that is grip.
While some researchers have responded to this issue by programming movement – or “teaching” the robot – before it completes a task, a team at the University of Washington has created an alternative: a five-fingered robot that can teach itself.
This robot doesn’t require human assistance, but rather handles objects and learns from its own experience, showing improvement after each attempt. Programming rules into the robot to perfect a motion requires a lot of time and tweaking, whereas enabling the robot to learn on its own saves the engineers a lot of time.
That’s not to say that building this robot was easy, though; the team had to make sure the hand would possess enough strength, speed, flexibility and responsiveness to perform like a human hand does. The hand has 40 tendons, 24 joints, more than 130 sensors and cost about $300,000.
It may be a little pricey for factory use just yet, but the team says it opens up the door to further research and development toward improving robotic dexterity.

SO, WHAT DO YOU THINK?
Is this a good solution for the dexterity problem in robotic development? How could this technology benefit an industrial setting?
Email us or leave your comments below.
That’s all the time we have for today, but tune in every weekday for your next Manufacturing Minute.