Dex-Net 4.0 in Video

15 Mar, 2019

The AutoLab uploaded a video of BCNM faculty Ken Goldberg's Dex-Net 4.0 on Vimeo. In the video, you can see the robot explore "ambidextrous" grasping using two heterogeneous grippers.

From the video description:

We train policies for a parallel-jaw and a vacuum-based suction cup gripper on 5 million synthetic depth images, grasps, and rewards generated from heaps of three-dimensional objects. On a physical robot with two grippers, the Dex-Net 4.0 policy consistently clears bins of up to 25 novel objects with reliability greater than 95% at a rate of more than 300 mean picks per hour.

View the video directly on Vimeo here, or watch below:

Learning Ambidextrous Robot Grasping Policies - Dex-Net 4.0 from CITRIS Productions on Vimeo.