Warehouse robots getting better at sorting things out
Amazon competition challenges robots to face unknowns
MASASHI ISAWA and SATOSHI MOROTOMI, Nikkei staff writers
TOKYO -- The competitors at a global tournament for warehouse robots last month showed off some surprising advances in artificial intelligence and other state-of-the-art technologies.
Leading multinationals and research institutions brought their next-generation robots to the Amazon Robotics Challenge, hosted by the big internet company perhaps best known for its shopping site.
The tournament has been a cradle of future technologies.
Last month's was the third annual challenge but the first to take place in Japan. Sixteen teams from 10 countries and territories showed up in Nagoya, central Japan, to compete at the four-day event, which began on July 27.
The combatants had to compete in stowing and picking matches. In the stowing competition, each robot got to show off how efficiently it could pick up items and put them on a shelf. In the picking phase, the robots were to put specified items in target cardboard boxes. The best eight teams qualified for the final round.
This year, half of the 32 items were not revealed until the challenge began. New items are particularly difficult for warehouse robots to handle; the mechanical laborers must possess good recognition skills to get through their days. By not revealing some of the objects until the challenge began, the tournament organizers forced the robots to respond to unknown situations almost humanlike.
Let's vote on it
Tye Brady, Amazon Robotics' chief technologist, said that tournaments like the one in Nagoya last week help to push forward machine learning, object recognition and other advanced technologies. In turn, these innovations carry the warehouse robot industry into the future.
"I learned so much by participating in the competition," said Masaki Yamamoto, chief engineer at the robotics solution department of Panasonic. The Japanese company teamed up with the Nara Institute of Science and Technology; their robot finished sixth. It was their first ARC. "It changed our values and [the way we think about] nurturing talent," Yamamoto said.
Kazuyuki Ikeda, an executive officer of Askul, a Japanese office supply company that has been using robots at warehouses which handle online orders, said, "Unlike places like factories, warehouses are filled with merchandise that greatly vary in material, shape, size and weight."
The Panasonic-Nara Institute entry was imbued with so-called "democratic voting system" technology, which enhances a robot's ability to "recognize" even unknown items.
The system consists of several subsystems, such as AI, deep learning, high-performance cameras and weight sensors. Based on various input data, the system decides what the most plausible action would be by "voting" on it.
This way, even if one subsystem fails, the robot can carry on.
Huge technology gap
Yamamoto's team struggled in the final round because a spool of adhesive tape that was to be taken off a shelf got stuck on the robot's arm. The robot made numerous attempts and, with the help of its AI, was finally able to put the item in the target box.
"In terms of creating a reliable robot that can function continuously, we attained what we had aimed for," Yamamoto said.
Another Japanese team, made up of engineers from Mitsubishi Electric and other parties, also drew attention. It was the third time Mitsubishi Electric competed in the challenge. "I am interested in how robots can be used in the world of logistics," said Makito Seki of Mitsubishi Electric's Advanced Technology R&D Center. "The competition offers a chance to develop technology and learn how engineers from around the world are approaching [the same issues we are confronting]."
Mitsubishi Electric, Chubu University and Chukyo University developed a robot whose two arms function independently in picking up and placing products. The robot used to handle one object at a time, its two arms moving in a coordinated fashion.
The robot scored high in the preliminaries, handling unknown objects well by focusing on their weight. The robot is able to gauge the weight of items as it picks them up.
But there was a huge technology gap between the Japanese teams and the top finishers.
Good picking functions
Robotics experts in Japan were especially surprised by a team from Nanyang Technological University, a prestigious university in Singapore. The team's robot made almost no errors in recognizing items during the first day of the competition despite the venue's varied lighting, which made it difficult for the contestants to properly recognize products. The Nanyang team's image recognition technology drew sighs of admiration.
First place went to the Australian Centre for Robotic Vision, or ACRV, whose robot somewhat resembles an arcade-style UFO catcher. It picks up items from above rather than reaching out for them.
The robot also has unique AI, which allowed it to deftly handle the unknown objects.
The ACRV robot is able to "learn" items and how to handle them by taking in 200 to 300 pictures of each. Usually, deep learning only gets results after 10,000 or more pictures are processed. But the robot is preinstalled with data on a range of products and was able to compare the 200 to 300 pictures with the image data in its memory.
Only last year did Amazon Japan install robots at its logistics centers. But they are primitive machines, like conveyor shelves. And moving shelves are not going to receive products from suppliers and get them packaged for customers nearly as efficiently as robots with good picking functions.
Amazon is still looking for the right picking robot, and these annual competitions allow it to spot cutting-edge technologies.