Amazon handles millions of orders per day, shipping products around the world from its network of warehouses and fulfillment centers. And while a large network of third-party sellers partner with Amazon to list their products for sale on Amazon’s marketplace, others participate in the Fulfillment by Amazon program, through which they store their inventory at Amazon’s fulfillment centers, and Amazon ships orders when their products are sold.
To handle the huge demand for products, which often need to be delivered on the same day or within two days to Amazon’s millions of Prime members, the company has turned to building automated systems to streamline the process of picking and packing orders. These automation systems enable the human workers in Amazon’s warehouses to work more quickly and more efficiently.
Will Knight reports for MIT’s Technology Review that Amazon is organizing a competition to spur the continued development of robots that leverage computer vision and machine learning to do the work required by Amazon’s sprawling fulfillment centers. Robots that participate in the competition, scheduled for May, will earn points by locating products within a stack of shelves, retrieving them safely, and packing them into cardboard boxes. If they crush a cookie or drop a toy, they’ll have points deducted. The team whose robot accumulates the most points will win $25,000. Amazon’s competition is intended to catalyze the development of better technology and algorithms, and eventually to accelerate the rate at which Amazon automates the process of fulfilling the millions of orders it receives daily.
Knight notes that Amazon has already automated some of the product-picking and packing that takes place in its fulfillment centers. In a few locations, robots made by Kiva Systems, a company Amazon acquired in 2012 for $678 million, send shelves laden with products to human workers, who then package them. These robots reduce the distance that workers need to walk to find products.
However, no robot can yet pick and pack products with the speed and reliability of a human worker, and the industrial robots that are already widespread in other industries are limited to precise, repetitive work in highly controlled environments. People have a much easier time identifying objects, manipulating them, and grasping them with exactly the right amount of force than robots do. These tasks are difficult for a robot to complete if the object is unfamiliar, awkwardly shaped, or sitting on a dark shelf with an assortment of other objects. Amazon’s contest will require robots to perform without any guidance from their creators.
Pete Wurman, the chief technology officer at Kiva, says that 30 teams from academic departments around the world will participate in the challenge, to be held at the International Conference on Robotics and Automation in Seattle. In each round of the competition, robots will pick and pack one of 25 different items from a stack of shelves like those in Amazon’s warehouses. Wurman told Technology Review, “We tried to pick out a variety of different products that were representative of our catalogue and that pose different kinds of grasping challenges. Like plastic wrap; difficult-to-grab little dog toys; things you don’t want to crush, like the Oreos.” Some teams are developing their own robots, while others are adapting existing commercial systems with their own grippers and software.
Knight notes that Amazon’s contest could offer an opportunity to judge the progress that’s been made in the past few years as cheaper, safer, and more adaptable robots have emerged. New types of manipulators and “robot hands” enable machines to pick up awkward objects, copying the flexibility and sense of touch in human hands. Progress in machine learning enables robots to learn how to manipulate unfamiliar objects, identifying what kind of grip will be appropriate for new objects.
CNET ‘s Donna Tam reported late last year that Amazon had already integrated 15,000 Kiva robots with its warehouse technology across the 10 warehouses in the company’s network, which has more than 50 facilities in the United States. Despite worries that robots will take over large numbers of human jobs, Amazon told Tam that it didn’t eliminate any jobs with the introduction of Kiva. While Amazon wouldn’t say how many jobs it had added since incorporating Kiva into its warehouse systems, it said it had hired 61,110 employees since 2011, the year before it acquired Kiva.