For online retail giant Amazon, speed and efficiency are the keys to expanding their retail empire. Just last week, the behemoth announced same-day delivery for millions of items in cities across the U.S., making the notion of a “store” all but obsolete. In their latest attempt at innovation, Amazon held a “picking challenge,” where robotics companies competed for a prize of $20,000 to see whose robot could “pick” items the fastest. The winner was Technical University Berlin’s Team RBO, blowing away the competition with 148 points, beating the MIT squad by 60.
The good news: Robots won’t be replacing human workers as Amazon pickers any time soon. The bad news? The day will come, eventually.
“This has been a fantastic team effort. Every single member of our team contributed with enthusiasm and ingenuity, enabling us to produce a compelling showcase for mobile manipulation as a winning approach to industrial manipulation,” Prof. Oliver Brock of TU-Berlin told media at the event.
Faster picking will play a huge role in Amazon’s expansion of their instant-delivery program. Presently, picking is done mostly by humans, who receive orders and then manually find items and place them into bins for processing and packaging. Workers describe the job as grueling, with strict hours, high accuracy and fulfillment requirements, and a lot of bodily wear and tear, owed to sometimes having to trek miles across Amazon’s huge warehouses. Automated shelves ease some of the burden, but ultimately a robot that never complained and never made a mistake would be far more efficient.
The competitors used a wide variety of robotic arms and gripping mechanisms, but TU’s winning entry won with its human-like arm and vacuum powered suction cup. The device only faltered on a few items, and had difficulty on items that were either very small or had a porous surface that didn’t work well with a suction cup.
Ultimately, Amazon is more interested in the algorithms the teams created that allowed the robots to identify items and plan the most efficient picking routes. It’s far cheaper to improve existing technology, after all, than to spend time and money developing better gripping mechanisms.
For right now, the robots work far too slowly to be a feasible replacement for human pickers, who can both move quickly and quickly plan their routes. However, as algorithms improve and Amazon pushes harder for instant order fulfillment, it’s not hard to imagine a day when your Amazon order is picked and packaged by robots before being delivered by drones, all within the span of an hour or so.