Researcher Jeff Mahler (L) and Professor Ken Goldberg (R) organize objects for AUTOLAB’s robot to grasp.
13
have dramatically reduced robot training time by
using simulations instead of painstaking labeling
and image-learning.
When Dex-Net first began, the source of data
used to train the robot’s algorithm was tediously
hand-labeled images or examples collected from
a physical system. Researchers collected millions
of data points in a process that required a year or
longer. That’s no longer the case.
“The idea behind Dex-Net is to automate the
collection of training data by using simulation,”
says Mahler. “We use analytic models based on
physics and geometry to automatically determine
whether or not a robotic grasp would successfully
pick an object up. We also use a technique called
domain randomization to randomize parameters
of the simulator, such as object mass, friction,
and camera parameters, which aids in transferring
learning from simulation to reality.”
“The result is that we can collect millions of
useful data points in less than a day,” he adds. In
commercial settings, where time is money, that’s a
powerful innovation. The team makes much of the
training data and tools available in an open source
library for training other robots.
Last year, the Dex-Net robot won the Amazon
Picking Challenge—Amazon’s annual event that
benchmarks picking progress—with an astonishing
200 to 300 picks per hour, a tremendous increase
from the standard 70 to 95 picks per hour.
While in the midst of launching their own company,
Mahler and Goldberg are still in the AUTO-
LAB refining what they’ve learned and leveraging
advances in deep learning. “We are developing new
methods [of teaching] robots to perform tasks,
such as surgical needle insertion, rope-tying, and
assembly,” Mahler says.
Such teaching advances may elevate robots
from repetitive tasks on assembly lines to more
intricate tasks, like suturing in an operating room.
And that puts entirely new use cases for these
dexterous robots within grasp. ■