Thanks to the Robot Operating System, there is a community platform for robotics programming, but there has not been one for scaling of robot fleets — until now. The Robot Operations Group is a new collective of people, companies, and organizations across industries intended to help accelerate the adoption of robotics.
“Our motivation was that there are a lot of places for people to go to learn how to build a robot using the Robot Operating System [ROS] for all sorts of things, but there was no community for building robots at scale. It was holding back the industry,” said Florian Pestoni, co-founder and CEO of InOrbit Inc., which is developing fleet management software. “We started the Robot Operations Group, or ROG, about a year ago.”
“Although I was hired three years ago to lead operations for fleets of robots and grew that fleet from dozens to 11,000 mobile robots, Florian and I are not representing InOrbit or Brain,” said Joe Wieciek, manager of software operations at Brain Corp. “We both have experience in high-scale systems and are looking to scale the industry.”
“We’re looking to change how you do DevOps,” Pestoni told The Robot Report. “It’s not just tools; it’s also best practices and mental models. We wanted to bring some of the ideas from software development to robotics and to have an inclusive community.”
Robot Operations Group starts with meetup, manifesto
The Robot Operations Group started with Meetups, said Pestoni. “We were meeting virtually before it was fashionable, since we wanted people from the East Coast, Europe, and Asia,” he said. “We’ve got more than 35 members joining from different robotics companies, academia, and equipment manufacturers.”
ROG (pronounced “rogue”) had its first in-person meeting at RoboBusiness 2019, where it began working on its Robot Operations Manifesto, which describes the organization’s goal of “developing best practices for robot operations at scale.”
“The manifesto was a way of describing challenges, and we spent our first four to six meetings focusing on it,” said Wieciek. “We looked at four pillars of robot operations. Interoperability was debated pretty heavily — is it a core element or a feature? We described the feedback loop: scaling, learning, making changes to fleets, seeing to make sure what they have learned.”
“It’s open-source, in GitHub, so anyone can comment and make submissions,” said Wieciek. “We’re at an inflection point, moving from 50 to 5,000 robots. How can we now get the industry to millions of robots out in the world?”
“The punchline is ‘effective operations at scale’ — we have to accept failure and not plan for perfection but design for resilience,” added Pestoni. “We’re not inventing everything from scratch but are bringing in existing frameworks and tools. Recognizing issues and closed-loop corrective actions are some measures that predate autonomous mobile robots [AMRs].”
Industry drives ROG
Unlike other developer organizations, the impetus for the Robot Operations Group is coming from industry rather than academia, said Pestoni and Wieciek.
“The LISA conference for USENIX was instrumental in propelling my career,” Wieciek recalled. “Robots are no longer in the lab, but are in the real world. I couldn’t find a community, so I wanted to provide something like this. No matter what applications or fleet-monitoring software there is, we want to focus on best practices for scaling nondenominationally, without focusing on individual technology.”
“It’s not a matter of building tools and expecting use cases to follow,” he said. “You should build tool against a pillar rather than the other way around. Interventions [to troubleshoot AMRs] will happen for a long time.”
“As the industry matures, we will see an ecosystem where we’re going to see a lot of players in that circle, while some will be in just one slice,” Pestoni said. “That growth of options is healthy. Then we’ll see some specialization, like what we’ve seen with the cloud and AWS [Amazon Web Services]. We want to see the same thing with robots — for millions, you can’t have three engineers per robot.”
“ROG is not a prescriptive standards body. It’s really about creating a forum and exchanging ideas,” said Pestoni. “There are not too many people like Joe, who have scaled to 11,000 robots. There’s a lot of great experience out there that can be shared without pulling out proprietary knowledge. Our goal is to facilitate, not dictate.”
“At the end of the day, most developers want to work on robots that benefit society, so the more people who can run fleets of robots, the better,” he added. “We do have people from academia who have joined, but they have tried to make things practical as opposed to theoretical.”
The Robot Operations Group is actively looking for individuals as new members.
“We’re talking to people with shared interests, not to represent their companies in an official capacity,” said Pestoni. “It’s not that different from ROS. We’ve created our own Discourse group, where people can propose topics for our next meetings, which are about every month — every six weeks since COVID-19.”
“We spent a lot of time getting to the manifesto, but now that it’s published, we have interesting meeting topics ahead,” Wieciek said. “We’re discussing what people are actually working on.”
The September ROG meeting featured a discussion around defining and setting user roles. It included people from companies producing robots for cleaning, taking retail inventory, restaurant automation, and cleaning and disinfection, among other applications. Attendees also discussed software for Internet of things (IoT) integration, remote support, and machine learning, as well as how a robot operations center (ROC) is different from a help desk.
What are some common themes that have come up? “Integration and working in different legacy environments have received a lot of focus,” replied Pestoni. “We ask people when they join what they’re doing in the field, what challenges they’re facing, and what they’d like the working group to focus on. Any area could be blown up into 10 topics, such as connectivity and cybersecurity.”
“It’s the elephant and blind philosopher problem — we must find the problems, test, mitigate, prevent, and fix when unavoidable,” he said. “We’re also open to representation from big AMR users.”
“Users are still ramping up, and the market is still very fragmented,” Pestoni said. “The lack of a framework or models has held back scaling. You can send a robot to a contract manufacturer, but then you have to run them. We’d love to get more perspectives from the community.”