Amazon Is Looking for the Perfect Warehouse Worker

Image
Geoff Robinson/Amazon.com

Amazon.com Inc.’s inaugural “Amazon Picking Challenge” inspired mechanical engineering and computer science students from around the world to design robots that can grab boxes of Oreo cookies and pencils from warehouse shelves and place them in bins, tasks ordinarily done by people.

The Seattle retailer hopes to make its challenge a regular event that encourages innovation in robotics and steers academic research toward e-commerce automation.

Participants, however, said Amazon will have to be more generous with prize money and travel vouchers in the future for that to happen.

The world’s largest e-tailer, which had 2014 revenue of $89 billion, budgeted a total of $26,000 for prizes and $60,000 for travel grants for more than 30 teams participating in the competition this week at the International Conference of Robotics and Automation in Seattle.



“If you really want strong teams, you need a bigger investment,” said Alberto Rodriguez, who led a team of graduate-level students from the Massachusetts Institute of Technology. “This probably needs 10 times the money.”

The feedback highlights the competition Amazon faces in using a contest to attract the resources and talent of academic researchers to solve its problems. The Department of Defense, for instance, will host the DARPA Robotics Challenge in Pomona, California, next month, featuring $3.5 million in prize money for teams that design the best robots that can assist humans in natural disasters.

Amazon has made big investments in automation to make its warehouses more efficient. It purchased warehouse robot-maker Kiva Systems Inc. for $775 million in 2012 and has 15,000 robots deployed in its facilities. Those machines move entire shelving units, but picking individual items is a difficult task for machines and remains better performed by people.

That was the focus of the Amazon challenge. Each team’s robot tried to pick up a shopping list of items of varying shapes and sizes — Crayola markers, a duck toy, tennis balls —  stored on shelves and place them in a bin.

It’s tricky for robots, which use sensors to identify objects that can be confused by plastic packaging. The machines also sometimes wrestle with how to figure out the best way to grab an object.

It’s a simple task for people. Last year, Amazon hired 80,000 temporary workers for its more than 50 warehouses during the holiday shopping season to do similar tasks.

The MIT team’s orange robot MCube nabbed seven of 12 items on the shopping list, using a pincher that resembled a large set of tongs with a suction cup on one end. One of the items that defied the robot was a tennis ball.

Amazon Chief Technology Officer Peter Wurman, who solicited feedback from participants, said he would make the case for additional funding. Only about half of the participants said they would return for future competitions with similar support from Amazon.

“I brought the executives who funded this by, and they were all very impressed,” Wurman said. “So we’ll work on them and see what we can do.”

A team from Technical University of Berlin won the contest with its vacuum-powered robot grabbing nearly all of the items on the list. Team leader Oliver Brock said he wasn’t sure the university would compete again.

It cost $6,000 to ship the team’s $250,000 robot, which was damaged on the way to Seattle, threatening the school’s ability to use the equipment in other research.

“This really got people’s attention and it’s a cool way to illustrate the state of the art,” he said of Amazon’s contest. “It’s just a big risk for us.”