Innovation

Photos: A glimpse of Amazon's robotic future

By // July 11, 2016, 5:04 AM PST // nickjheath

Image 1 of 11

  • Robots struggle with the messy complexity of the real world, particularly with tasks like picking up objects from items jumbled together on a shelf.

    But in only its second year, a contest to build a warehouse picking robot has seen a threefold improvement in the winning bot's work rate.

    The robot arms taking part in this year’s Amazon Picking Challenge performed far better than bots in the preceding competition.

    Here you can see the winning bot made by Team Delft, a partnership between researchers from the TU Delft Robotics Institute and the company Delft Robotics in The Netherlands.

    Team Delft came first in the picking contest, where bots scored points for correctly selecting individual items from those bundled together on shelves. The items represented a cross section of products found in Amazon's warehouses, and ranged in shape and size from clothing to a toothbrush.

    The team trained its bot to recognise the 40 Amazon warehouse items by feeding a deep learning system a database of 3D scans of the items. Doing the grasping was a robot arm with seven degrees of freedom — a term that reflects its ability to move in any direction in 3D space, as well as yaw, pitch and roll. The arm used high-quality 3D cameras for vision and a combined gripper/suction cup, designed in-house.

    Image: Amazon

  • A team from Japanese machine learning specialist Preferred Networks (PFN), which secured second place in the picking contest with this pair of FANUC robot arms.

    Like the winning team, PFN researchers used deep learning algorithms to train the arms’ 3D cameras in how to spot and handle items, and built specialized tools for grabbing objects.

    The PFN bot scored 105 in the picking contest, the same as the Team Delft robot, but Delft was awarded first place for picking up its first item more quickly.

    Image: Amazon

  • The Team Delft bot reaches into one of the 12 bins on the shelf to pick up an item.

    The shelves were designed to be a simplified version of those found in an Amazon warehouse. This year’s contest was tougher, with more items mixed together and objects that were trickier to spot and to grasp.

    While the Team Delft bot was notably faster than last year’s champion, it still lags behind human pickers. A human can pick about 400 items from shelves per hour with minimal errors, whereas the Delft team's winning bot can select about 100 items per hour with a 16.7 percent failure rate.

    Image: Amazon

  • A team from the Massachusetts Institute of Technology (MIT) took part in the contest for the second year running.

    The team’s industrial ABB arm, with a gripper at the end, scored 67 points in the picking contest, earning MIT fourth place.

    MIT’s bot did better in the stow contest -- a new challenge where bots were tasked with taking an item from a tray and placing it on a shelf -- racking up 164 points to secure third place.

    Image: Amazon

  • Several teams in this year’s challenge used Baxter robots, including engineers from Duke University in North Carolina and from Oakland University in Michigan.

    Baxter is a six-foot tall, 300-pound industrial robot that can be trained to carry out tasks by taking its hand and guiding it through the motions it needs to repeat.

    Neither the Duke nor Oakland team placed above the halfway point in the picking contest.

    Image: Amazon

  • Engineers from Chubu and Chukyo universities in Japan teamed up with researchers from Mitsubishi to field this pair of robot arms.

    The C^2M placed eighth out of the sixteen teams taking part in the picking contest but fared less well in the stow contest, finishing fourth from bottom.

    Image: Amazon

  • A team from Rutgers University in New Jersey entered this dual-arm Motoman robot arm but failed to score in the contest.

    Only four teams didn’t score a single point in this year’s Amazon Picking Challenge, an improvement on last year’s competition when half of the teams crashed out.

    Image: Amazon

  • The robot arm from Team Delft that won the challenge.

    Team Delft also won the stow contest. One of the researchers from Team Delft suggested that robots based around its winning entry would be fast and accurate enough to be used by small businesses.

    Image: Amazon

  • The PFN team make adjustments to their FANUC robot arms.

    Amazon claims the idea of the contest isn’t to create bots to replace human workers in warehouses but to augment manual labor.

    Amazon uses Kiva robots to move shelves in 13 of its 123 warehouses worldwide and points out that warehouses that use robots increased their workforce after the bots were introduced.

    Image: Amazon

  • A close-up of the combined suction arm / gripper used on Team Delft’s winning arm. This combined approach to picking objects helped the bot handle the variety of differently-shaped and sized items found in an Amazon warehouse.

    The Amazon Picking Challenge is an annual competition and Team Delft won $50,000 for coming first in both the picking and stow contests.

    Image: Amazon

  • The Rutgers University team check everything is working on their dual-arm bot.

    Image: Amazon

Robots struggle with the messy complexity of the real world, particularly with tasks like picking up objects from items jumbled together on a shelf.

But in only its second year, a contest to build a warehouse picking robot has seen a threefold improvement in the winning bot's work rate.

The robot arms taking part in this year’s Amazon Picking Challenge performed far better than bots in the preceding competition.

Here you can see the winning bot made by Team Delft, a partnership between researchers from the TU Delft Robotics Institute and the company Delft Robotics in The Netherlands.

Team Delft came first in the picking contest, where bots scored points for correctly selecting individual items from those bundled together on shelves. The items represented a cross section of products found in Amazon's warehouses, and ranged in shape and size from clothing to a toothbrush.

The team trained its bot to recognise the 40 Amazon warehouse items by feeding a deep learning system a database of 3D scans of the items. Doing the grasping was a robot arm with seven degrees of freedom — a term that reflects its ability to move in any direction in 3D space, as well as yaw, pitch and roll. The arm used high-quality 3D cameras for vision and a combined gripper/suction cup, designed in-house.

Image: Amazon

Related Topics:

Innovation Amazon Digital Transformation Artificial Intelligence Internet of Things Smart Cities Hardware

About Nick Heath

Nick Heath is chief reporter for TechRepublic. He writes about the technology that IT decision makers need to know about, and the latest happenings in the European tech scene.