Amazon’s progress toward an army of helpful robots is one step closer: a prize for the best warehouse-working “picker” machine has gone to a robot designed by a team from TU Delft Robotics Institute and Delft Robotics, both based in the Netherlands.
The competition was held in conjunction with Germany’s Robocup in Leipzig. Announced on Monday, the winners took home $25,000, while the university of Bonn’s NimbRo won $10,000 for second place and Japanese firm PFN was awarded $5,000 for third.
The contest, in Amazon’s words, “aimed to strengthen the ties between the industrial and academic robotic communities,” and ended with slightly fewer than half of the entrants scoring more than 20 out of 40 possible points, according to a report in TechRepublic. The technology is advancing quickly: all of those contestants would have surpassed the highest scorer in the previous Picking Challenge, held just three years ago.
To compete, on 29 January the robots tried to perform a number of tasks using “books, cubic boxes, clothing, soft objects, and irregularly shaped objects” and a box provided by Amazon. The robots had to take items out of the box and put them on the correct shelf, and remove shelved items and put them into the box.
Object recognition is difficult for robots. A UC Berkeley project from six years ago was hailed as a breakthrough when the robot developed by the university’s team was able to fold five never-before-seen towels over the course of an hour and 40 minutes. The team produced a video sped up 50 times for those in a hurry.
The project is close to what the US military division Darpa asked robots to do during its Robotics Grand Challenge in Pomona, California, in 2015. The idea of robots surpassing humans in service jobs or on the battlefield tends to give spectators pause, but for the moment, at least, the human factor isn’t going anywhere. The Delft team’s robot scored very high on the list of of tasks, but it still failed more than 16% of the time.
TechRepublic points out that the robots were able to retrieve 100 items an hour, where humans are able to retrieve 400 items per hour, and the human would be unlikely to drop 67 of those items.
Just as Darpa did last year, Amazon is downplaying the possibility that humans will lose jobs to robots. The company already uses robots made by Kiva – the robotics company it bought in 2012 for $775m – in 13 of its warehouses and points out that it has increased human head count since introducing android co-workers.
One US navy officer interviewed by the Guardian at the Darpa challenge theorized that Google’s purchase of Boston Dynamics, the firm that solidly outperformed its rivals during those contests, would be used to populate its server farms with robots that could maintain computers.
This article was written by Sam Thielman in New York, for theguardian.com on Tuesday 5th July 2016 20.15 Europe/Londonguardian.co.uk © Guardian News and Media Limited 2010