MACHINE PERCEPTION AND ROBOTICS GROUP

Dept. of Robotics Science and Technology,
Chubu University

At the Amazon.com logistics warehouse in the United States, the Kiva Pod robot made by Kiva Systems (Amazon Robotics as of 2016) automatically conveys products from the storage shelves to the people responsible for picking. Manual labor is still needed to pick these products from the shelves, but it is expected that this task will eventually be automated by introducing picking robots. In an e-commerce business where there are many types of products stored randomly on shelves, the key to the introduction of automation is being able to perform stable pick-and-place operations by recognizing diverse objects on the shelves and gripping them correctly. Against this background, Amazon set up the Amazon Picking Challenge (APC) as a competitive event for robots in the automation of logistics. The first such event was APC 2015, which focused on the problem of picking diverse items. Contestants were required to build picking robots that could extract 25 different items from 12 frames (called “bins”) on a shelf [1]. At APC 2016 in the following year, the scope of the competition was made more realistic by having the contestants compete on two different tasks (“Pick” and “Stow”). This paper introduces the robot system of Team C^2M at Amazon Picking Challenge 2016, and its image recognition system.