Germany’s Team RBO Takes the Amazon Picking Challenge Top Prize
Photo: AUVSI.
A team from Technische Universität Berlin named RBO took home the top prize and $20,000 in Amazon’s Picking Challenge at this year’s International Conference on Robotics and Automation.
A team from MIT took second place and $5,000, and Team Grizzly from Michigan’s Oakland University took third and $1,000.
Twenty-five teams competed in the challenge, which is intended to develop technology and software that will allow robots to quickly and accurately select items from bins. Amazon already has automated warehouses where robots move items for shipping, but the company says some “unstructured” environments remain, and that’s where they need help.
RBO team member Sebastian Höfer credits the win to the team’s testing — and a bit of luck. The team began work on the project two and a half months ago and a few weeks ago stopped adding new features and focused on testing to make sure its system was “bulletproof,” he says.
“Some teams did have fancier components, but then they got stuck,” he says.
RBO used a Barrett Technology-built WAM arm sitting atop a university-built platform that has been around since the 1990s, Höfer says.
There was one problem: When their shipment arrived at the ICRA show from Berlin, “our arm had been broken,” he says.
Fortunately, Barrett is one of the companies that is exhibiting at ICRA 2015, and company representatives were able to repair the arm.
In the competition, the robots were faced with an inventory shelf loaded with random items, including stuffed animals, Cheez-Its, Elmer’s Glue and copies of “The Adventures of Huckleberry Finn.” They had to reach in and bring out some of the items, getting at least one from each bin — some bins contained a single item, and some had more than one. Each team had 20 minutes to complete the task.
The teams could either build their own systems or use existing ones, so there was a mix of home-grown robots — Willow Garage PR2s, Yaskawa Motoman systems, Fanuc arms and others.
Some teams used human-style grabbing hands while others relied on two-finger robot hands or, as RBO did, air-suction systems.
“At Amazon, we’re excited to inspire the next generation of technologists and engineers to think of new, innovative ways to incorporate robotics and automation within the fulfillment industry,” the company said in a statement. “We have more than 15,000 Amazon robots operating in our fulfillment centers today, enhancing jobs for employees and speeding up order delivery times for customers. We’re excited to continue creating great, full-time jobs where employees interact with high-end technology as part of their day-to-day roles.”
Emulating Nature
Many researchers at the conference recounted their efforts to replicate systems found in nature. One Wednesday keynote speaker, Harvard University’s Radhika Nagpal, has had more experience with that than most and says it’s not an easy thing to do.
Her team has built a system of 1,024 tiny robots, dubbed kilobots. They are simple devices that use buzzing motors, such as you’d find in a cell phone, to move around on a level surface.
The goal is to try to understand and replicate the behaviors of insects such as ants and termites, which have no centralized authority, yet are able to carry out complicated actions. If you knock off the top of a termite mound, for instance, they will build it back without anyone telling them to do it.
“What makes these fascinating is that many are not how we think about human organization,” she says. “There’s no leader, no professor of the research group and no chair in charge of the department.
By interacting only with the robots around them, the kilobots are able to autonomously form complex shapes, although Nagpal said there have been plenty of challenges in getting to that point.
The robots push each other around, they vary in speed and their motors burn out — all things that are hard to predict.
“We had slow drivers and we had reckless robots, and that was an interesting thing that we just had not modeled in our simulation,” she says.
Her team is still working on these behaviors, but she noted that the kilobots operate only on a two-dimensional plane, where ants and termites move in a much more complicated environment.
“We can build systems but we don't know how to make them autonomous in three dimensions,” she says.
Embracing Innovation
Helen Greiner, a cofounder of iRobot and CEO and founder of the drone maker CyPhy Works, spoke Thursday morning about the importance of innovation for startup companies.
Most innovation comes from solving a customer need, she said. In its early military robot days, iRobot struggled with the question of how to move through rugged and complicated environments. Some observers said they would need a legged robot. Instead, they built a tracking system with one degree of freedom, and the PackBot was born, she said.
Likewise, the company struggled with the question of how to clean up against walls with its Roomba vacuum. The solution was digital infrared, not more complicated rangefinders.
CyPhy Works tackled a different problem: how to meet the military’s need for persistent surveillance with small unmanned aircraft that typically can only fly for a few minutes at a time. The solution was to feed power through a filament, allowing them to fly for days at a time.
“You must really innovate to survive. … It is about the money. Money fuels more investment. Money fuels your internal research and development budget. If you can get a jump on that, you can feed it back in, create that virtual cycle. … I hate to say it in this community, but it is about the money,” she said.
