This site may earn affiliate commissions from the links on this page. Terms of employ.

Team Delft Amazon

For anyone wondering whether robots are gaining ground on people when it comes to performing repetitive tasks, you need wait no further than the Amazon Picking Challenge. Having long ago automatic the movement of goods in its warehouses through its acquisition of Kiva Robotics, Amazon is now looking at engineering science to reduce the number of people needed to pack boxes. Information technology recently hosted a DARPA-claiming-style contest, featuring over a dozen teams from the effectually the world, whose robots competed for the title of best autonomous box packer. This contest, its second one, took place in Leipzig, Germany, and featured a more than circuitous "Option" task than the first challenge Amazon held last year in Seattle, and a new "Stow" job for unloading.

Team Delft's robot holds up its first place award

Credit: Team Delft

Team Delft's robot, featuring both a two-fingered gripper and a suction device, achieved pinnacle scores and times in both the box picking-and-packing, and the reverse un-packing and restocking, to have home the $fifty,000 get-go identify prize. Amazon kept things interesting by using a dozen differently shaped objects in each job, and xl items overall in the competition. This meant the robot had to adapt its picking strategy to each specific particular. Teams were given a JSON file with an item list and work order 5 minutes before the challenge began.

Breaking new footing in flexibility

The flexibility to handle the broad-diverseness of objects found in an Amazon warehouse is perhaps the largest breakthrough needed to brand full general-purpose warehouse robots a reality. Currently about package handling robots are designed with a single, or small number of related, packet sizes and shapes in heed — or their cargo is prepackaged or paletted for easy automatic handling. Many of the robots, including the one from Team Delft, included a depth-sensing camera to help identify the objects and their exact size and location.

More classic warehouse robots like this one are designed to move a large number of very-similar objects, not a wide variety of different products

Credit: Team Delft

In add-on to timing the robots, Amazon deducted points for damaging an item, dropping it more than a foot, or misplacing it on the shelf. Objects ranged from a T-shirt to a dumbbell. Delft's point score was matched past Japan'due south Team PFN, but the Dutch team performed its job almost 30 seconds faster, giving information technology get-go place.

Squad leaders and event organizers provided some cheery words about a time to come where people and robots would work side-by-side in the warehouse — especially since current-engineering science robots are only expected to be able to handle well-nigh 50% of the variety of products — only the long-term trajectory is the replacement of workers performing repetitive tasks with machines.

A large part of what makes these new applications possible is deep learning. Using the same blazon of software tools that have enabled facial recognition and early-stage democratic vehicles, these prototype warehouse robots identified objects and pattern-matched their shapes and attributes with the appropriate picking and packing strategies.

[Top image credit: Amazon]

Now read: Artificial neural networks are changing the world. What are they? and Robots are coming for your job — but perchance that'south okay