Historically, robotic innovation has had a fairly high barrier to entry. If you wanted to develop the next big thing in any one area, you had to build a complete system on which your idea could be implemented: a physical device, of course, but also the control systems, interface, and inspection tools required to get the robot up and running as a test platform.
As a Ph.D. student, Gerkey was thwarted by this problem in the course of his research. “We needed to build software infrastructure before we could do anything,” he recalled.
Gerkey’s team built its own infrastructure, called it Player, and released its code on Sourceforge — the best repository for open-source software at the time.
It kicked off an obsession with open-source tools that distracted Gerkey from his work, much to the chagrin of his advisor.
“She’d say, ‘Stop doing that! You need to do science and get a Ph.D.!'” he said. “I was always more motivated by the tool building.”
So it was no surprise, then, that he was drawn to an opportunity presented to him by the Willow Garage team in 2009: “They said, ‘It’s kind of like Player, but bigger and better and can handle all these different use cases.’ I said, ‘Let’s go do that full time.'”
An end to reinventing the wheel
At Willow Garage, the team set out to develop a basic robotic operating system (OS) that would support the PR2 research platform. “At the outset, the initial goal was to stop people from reinventing the wheel,” Gerkey said.
About the OSRF
- Founder & principal: Brian Gerkey, CEO
- Products: ROS, Gazebo
- Available: Yes. Free.
- What it is: An open-source robotic operating system
- Market niche: Originally, focused on research teams in robotics; today, increasingly focused on commercial applications.
- Funding: Government contracts and grants. Total amount undisclosed.
- Industry Partnerships: Spin-off of Willow Garage. Numerous collaborations and relationships across the sector.
A long history of academic and commercial research had solved many of the basic software problems in robotics, such as basic mapping or localization; the solutions were published as algorithms in journals or littered around the market, if you knew where to look.
“But these were rarely packed in a form where you could pick them up and use them,” said Gerkey.
The team began to identify these best-of-breed solutions and build them into a basic platform for researchers. “We wanted to raise the level from which everyone starts when they’re going to make a robot do something,” Gerkey said.
The team initially focused on applications for graduate researchers. The result was a complete robot OS, known simply by the acronym ROS, that was focused on maximum flexibility. The team attempted to build a system with no constraints or assumptions about what the end-user would do with it once it was installed on robots in the lab.
The platform gained traction in research communities around the world, through the constant circulation of interns and researchers who were collaborating with (or interning at) Willow Garage or using the PR2.
But researchers weren’t the only audience attracted by this flexibility; robotics startups were quickly getting hip to ROS, as well.
“We always had in mind that the problems we were solving would be applicable on other platforms; we wanted to make sure the software was portable across robots,” said Gerkey. “At first, we did not have the commercial use cases in mind. It was something of a refreshing surprise when we started hearing from companies who were picking up ROS and using it.”
Out of the lab, into the market
There have been robotics startups that did really interesting work with ROS. Clearpath, Rethink Robotics, Unbounded Robotics, Neurala, Blue River Technologies — are all among the growing number of companies that are using ROS to build (or support) their next-generation technologies.
It’s a virtuous cycle: The more ROS users in the market, the more advantageous it is to build using ROS.
“It’s starting to let people really work with our systems,” Ryan Gariepy, CTO of Clearpath said. “They’re able to integrate and work with our platforms very quickly.”
Neurala’s CEO, Max Versace, agreed that the spread of ROS has made it really easy for companies like his to integrate with many partners at the same time. “In the consumer space, many robots will be extremely open and standardized, making it really easy to hook into them,” he says.
In 2012, the ROS team spun out of Willow Garage as a separate nonprofit, the OSRF, charged with maintaining and promoting ROS and other open-source tools for the robotics community. That mission has grown substantially as open source tools have spread out of the lab and into the market.
Today, OSRF balances its need to stay on top of the latest research and continue to develop new capabilities and packages for ROS, with its need to maintain existing modules, build relationships, and support its increasingly diverse community.
Its current funding relies heavily on government contracts and some grants from industry sources. In the future, Gerkey says he hopes the organization will benefit from the growth of the startups it has helped get off the ground.
“Today, it’s a lot of startup companies working with ROS, and they don’t have a lot of cash on hand to make donations. It will get better as the industry matures,” Gerkey told me, optimistically. “Many big tech companies support Apache and Linux, because they need that stuff to keep working, to keeping doing what they do.” He hopes to see the same trajectory for OSRF.
To make sure that happens, OSRF has started to pay more attention to the needs of its commercial user base.
“At first, we started thinking about the grad student user, but now we’re definitely thinking about what would the product manager in a medium-sized company want to see if they’re going to roll out a product and supporting it based on ROS,” said Gerkey. “That goes into the QA process, the level of testing, and how the underlying communications system is built.”
ROS 2.0: Faster, smarter, better communicators
Over the past year or so, the OSRF team has started working on what is casually known as ROS 2.0. It’s a rewrite of the core system, largely focused on improving the distributed computing system that allows different parts of the system to exchange messages with one another.
“We’re rewriting how that works, in part to get the features in there that were missing, but also to take advantage of the opportunity to build something that we know our commercial users will believe in from the beginning,” said Gerkey.
The full timeline and product road map for ROS 2.0 won’t be fleshed out until later this year, but among the planned improvements is native, high-reliability real-time control.
Currently, the communication system relies on external real-time communication systems. But with the growing use of robotic arms and legs in the commercial space, OSRF thinks its time to improve this capability.
“It was all robots on wheels, with cameras or lasers, when I was in grad school,” Gerkey said. “Hundreds or even tens of hertz was a fine; but if you want your robot to balance on legs or move its arms fast and not wiggle once they get to where you set them, you need a tight control loop, in the thousands of cycles per second range.”
The team is also confronting its own human communications challenges: making it clear to potential ROS users what the capabilities, value, and limitations of the system are. Better QA tools and a campaign to highlight the systems’ robust security features are at the top of the list.
Despite the ubiquity of open-source software, many potential users are still perplexed by the claim that open source software is secure.
“The fact that you can look at the code doesn’t have anything to do with whether you can build a secure system based on that code,” Gerkey said.
It’s a message OSRF is hoping to drive home by highlighting its work with NASA, which is using ROS to control some of the robotic equipment installed at the international space station. ROS talks to the space station via a fully authenticated, encrypted, VPN.
The future of the OSRF
The future is bright future for ROS. Many of the leading startups in the space are turning to ROS. As they build revenue and become market leaders in the huge range of industries that are ripe for transformation through robotics — manufacturing, transportation, agriculture, home health care — it will be in their best interest to give back to the community. But in the meantime, it might be in the best interest of large companies that aren’t using ROS to step up, as well.
Creating a lingua franca in which potential partners can communicate with old-guard companies will speed up innovation and allow market leaders to maintain their relevance and customer relationships, while capturing the benefits of new technologies for themselves.
ROS provides that language, and the emerging startup ecosystem is taking advantage of it. If they don’t want to be left out in the cold, big companies would do well to sit up and pay attention.
For small companies, Gerkey said there are plenty of ways to contribute before the big bucks are rolling in.
“The thing that we would benefit the most from is more people taking over to maintain the packages that are out there already,” Gerkey said. “You should identify the packages in the ROS ecosystem that you care about and consider contributing some in-kind labor contribution.”
For any individual package, OSRF invites ROS users and engineers to step up and help answer questions, look into bugs, and apply patches.
“It doesn’t take that much time for each piece,” he said. But altogether, “it can take up a lot of our time, which gets in the way of new development.”