Working in a warehouse is tough for robots. Unlike on a production line, where the same motion is constantly repeated, orders and tasks can change, requiring them to adapt.
On top of that, different objects require different – and often delicate – handling, and speed and precision are prerequisites.
US engineers think they are a step closer to making this a reality thanks to a combination of motion planning software and artificial intelligence that optimises robotic arm movement. And it’s the combination that’s important.
In earlier work, Ken Goldberg and Jeffrey Ichnowski at the University of California Berkeley created the motion planner, which can compute how a robot should pick up an object and how it should move to transfer the object from one location to another.
The motions tended to be jerky, however, and the calculations needed to tweak the software to smooth things out took around half a minute.
In new work, described in the journal Science Robotics, Goldberg, Ichnowski and colleagues say they were able to dramatically speed up this computing time by integrating a deep learning neural network.
A robot learns from examples, and can generalise to similar situations. And then it’s back to the motion planner.
“The neural network takes only a few milliseconds to compute an approximate motion. It’s very fast, but it’s inaccurate,” Ichnowski says. “However, if we then feed that approximation into the motion planner, the motion planner only needs a few iterations to compute the final motion.”
In combination, they cut average computation time from 29 seconds to less than one-tenth of a second.
Goldberg predicts that, with this and other advances in robotic technology, robots could be assisting in warehouse environments in the next few years. “This is an exciting new opportunity for robots to support human workers.”
The Royal Institution of Australia has an Education resource based on this article. You can access it here.