Sucking Robot Arm Wins Amazon Picking Challenge

Team Delft’s Machine – Won Latest Warehouse Bot Competition

Team Delft’s machine, has won the latest warehouse bot competition of Amazon, a robotic arm which tends to combine a suction cup, a `two fingered’ gripper together with a 3D depth sensing camera, with its rivals at both the tasks. One comprised of choosing products from a container, picking them up and placing them on a shelf while the other did the action in the reverse. Amazon tends to use robots in moving goods with its buildings though depends on humans to stock the same on shelves.

Chief technologist at Amazon robotics, Tye Brady had commented that `their vision is humans and robots working alongside each other. It was inspiring to see 16 top teams with so many different approaches to the same problem and that they saw the advancements robotic technology had made since last year’. The Dutch winners had been awarded $50,000. They rivalled against two teams from Japan and five from the US countries who were known better for their robotics research, at the contest in Leipzig, Germany. There were various items representing some of the blockbusting products of the retailers that were used in the Amazon Picking Challenge.

Stow Task/Pick Task

A combination of various shapes inclusive of soft clothing, a boxed DVD, a bottle of water and a toothbrush were represented. Twelve various items were put in a red plastic box, in the `stow’ task and the robots were made to pick them out in an orderly manner, placing each one at a predetermined place on a shelf. While in the `pick’ task, a dozen specific products to be lifted, off the shelves, comprised of a combination of goods and packed in boxes.

 In some cases, the other items seemed to have been intentionally placed in the way of the target which seems to be tougher of the two tasks. In both the instances, the teams were given only a computer file defining the range of objects involved and instructions for which it should be moved, five minutes before to the start and once the task had started, the robots had to act independently. There were points deducted for damaging any item, dropping the items from a height of over 30 cm and leaving an object extending more than 0.5 cm beyond where it belonged on the shelf.

Variation in Time – Unable to have Single Picking Strategy

Delfts’ team, Kanter van Deurzen had clarified to BBC that due to the variation in items, one could not have a single picking strategy. Usually in industry, you would have a suction cup or a mechanical gripper intended for one kind of item or part. Here they had to handle dozens ranging from simple boxes to a T-shirt and a dumbbell with each needing a different approach. It had really been a big challenge to do all this with a single arm, to have to recognize how the items were focused on and to avoid collision with the other objects on the shelves.

 The Dutch team had come closer to a flawless score in the stow task but its only error was that when its arm’s suction cup had picked a small pack of glue, it also seemed to pick up a bottle-cleaning brush which then had dropped down on the floor. More errors were done in the pick task and initially secured points with the Japan’s Team PFN. However, Delft had been given the benefit since it had taken 30 seconds less time in making its first pick.