UC Berkeley Robot Demonstrates Precision Grasping

Posted on June 3, 2017

Roboticists at UC Berkeley have built a robot designed to properly grip, pick up and move irregularly shaped objects. They claim their robot can pick up and move unfamiliar, real-world objects with a 99% success rate.

The robot was created by Berkeley professor Ken Goldberg, postdoctoral researcher Jeff Mahler and the Laboratory for Automation Science and Engineering (AUTOLAB) . The robot is called DexNet 2.0. It gained its dexterity through deep learning. There are 6.7 million data points in the neural network DexNet uses to learn grasps that will pick up and move objects.

There is a story about the nimble fingered robot in MIT Technology Review. Here is a video of DexNet 2.0 in action:



More from Science Space & Robots