The Australian Centre for Robotic Vision (ACRV) has claimed first place in the 2017 Amazon Robotics Challenge with a low-cost Cartesian robot for unstructured environments. Built from the ground up for the competition, "CartMan" outshone its costlier rivals by picking and stowing more items in its allotted test time.
On the surface, picking an object and packing it in a box sounds relatively simple, but it involves the synergy of object recognition, pose recognition, grasp planning, compliant manipulation, motion planning, task execution and error detection and recovery. In other words, robots need to find the correct items, how to reach and grab the item, how to place it on the ground and, if things go wrong, fix the problem on the fly.
Why this specific skillset? Well, online retailers like Amazon are reliant on human-staffed fulfilment centers in which workers take items from the shelves and manually box them up for delivery. The online giant uses robots to shuffle products around these centers, but it hasn't managed to automate the picking and packaging process.
Having identified the correct item using a bespoke vision system and machine learning, CartMan automates the pick and stick process by moving on three axes (vertical, horizontal and forward/back) like a gantry crane. It has a rotating gripper on the end, which uses suction cups or a two-finger grip to grab the correct item.
The ACRV team headquartered at the Queensland University of Technology (QUT) isn't new to the Amazon Robotics Challenge. It used a modified off-the-shelf robot last year, and managed to finish sixth. This year, the team poured more than 15,000 hours into developing CartMan from the ground up.
"We opted to build our own robot from scratch – a three-axis Cartesian robot that acts much like a gantry crane you see at ports," says Dr Juxi Leitner from QUT. "With six degrees of articulation and both a claw and suction gripper, CartMan gives us more flexibility to complete the tasks than most robots can offer. CartMan is robust and tackles the task in an innovative way and is also cost effective. We learnt from our experience last year when we used an off-the-shelf robot. I think we had the lowest cost robot at the event!"
The Amazon Robotics Challenge was held in Nagoya, Japan as part of the RoboCup and attracted a field of 16 teams from 10 countries. The Australian Centre for Robotic Vision team takes home US$80,000 for winning the event.