Ever since the rise of the internet as our main source of buying and selling, the speed in which we can get goods has been incredible. Every year, it feels simpler and faster to just do all of your shopping online, as it can often be delivered to you so fast and so easy. eCommerce, then, looks set to change the way in which we shop for at least a generation to come.
However, with the rise of machine learning and AI, we could be about to see another quantum leap: starting with ambidextrous robots (ambidextrous= using both hands equally well).
Massive warehouses are needed to help fulfill the incredible orders being taken online, which obviously means trying to get staff to help manage said orders. However, it looks like we’re moving towards trying to automate the whole experience instead, with the use of ambidextrous robots.
This became clear from a UC Berkeley paper which was produced in mid-January, in the Science Robots journal.
Engineers at UC Berkeley have been using decades of progress so far to come up with new algorithms that could help to compute robot pick points, allowing robots to work with a large variety of objects without having to undergo the training that they would need to normally.
The postdoctoral researcher on this project, Jeff Mahler, who is also the lead author, said: “Any single gripper cannot handle all objects,
“For example, a suction cup cannot create a seal on porous objects such as clothing, and parallel-jaw grippers may not be able to reach both sides of some tools and toys. ‘Ambidextrous’ robots offer greater diversity.”
At the moment, the robots used in the majority of fulfillment centers will be using the standard suction cup session mentioned above. This can limit the range and style of objects that it could actually pick up and work with, which would drastically limit their usage in a warehouse environment.
With ambidextrous robots, though, this would not be quite the case. This would make them compatible with more than one gripper type, allowing for much easier for the robot to decide which gripper is most likely to succeed when picking up any given item.
This would help to ensure that, by using a reward function, ensuring that fulfillment could be much faster and far less rigid in what it can and cannot pick up.
The researchers managed to train functions for using a parallel-jaw gripper and then a suction cup gripper to a two-armed robot. The result was that it could take on 25 objects never seen before with around 300 picks in an hour, with around 95% reliability: that’s incredible.
While this might not be commonplace yet just in every warehouse, it’s definitely something that we can expect to arrive in the next few years – with massive, positive results for the already in-demand eCommerce industry.