The next big challenge for robots: Insights from the new director of CMU’s Robotics Institute


Matthew Johnson-Roberson, new director of Carnegie Mellon University’s Robotics Institute, spent his teenage years in Seattle, and did his undergraduate studies at CMU. (GeekWire Photo / Todd Bishop)

PITTSBURGH — A few months into his new job as the director of Carnegie Mellon University’s storied Robotics Institute, Matthew Johnson-Roberson still finds the whole experience a little surreal.

A robotics engineer and researcher with experience in land-based and underwater autonomous vehicles, he received his doctorate in robotics in Australia at the University of Sydney, and co-founded a startup, Refraction AI, developing AVs that use bike lanes for last-mile deliveries. He was previously at the University of Michigan, co-directing the UM Ford Center for Autonomous Vehicles, and founding and leading the Deep Robot Optical Perception (DROP) Lab.

But his new role brings him back to where he started. Johnson-Roberson was an undergraduate at CMU’s School of Computer Science, and worked under industry pioneer Red Whittaker on an AV for the 2004 DARPA Grand Challenge.

“That was the turning point in my career that made me want to go into robotics,” Johnson-Roberson said, describing his new role as a “full-circle experience.”

As a point of local interest in the Pacific Northwest, Johnson-Roberson spent part of his childhood in Seattle, attending Garfield High School when his mom was a grad student at the University of Washington.

CMU’s Robotics Institute has been the source of numerous breakthroughs in robotics and automation since its founding in the late 1970s, developing robotic innovations in areas including nuclear cleanup, autonomous cars and space exploration.

We sat down with Johnson-Roberson in his office at the Robotics Institute this week to talk about his new role; the experience of working on robotics in two historically industrial regions; the opportunities and obstacles ahead in robotics; and the sometimes challenging relationship between research institutions and the titans of the tech industry.

Listen to the conversation here, and continue reading for edited highlights.

GeekWire is in Pittsburgh reporting on the future of robotics, AI and automation in conjunction with the Cascadia Connect Robotics, Automation & AI conference, organized by Seattle-based Cascadia Capital, which is underwriting our independent coverage.

Listen above, or subscribe to GeekWire in Apple Podcasts, Google Podcasts, Spotify or wherever you listen.

The next big challenge in robotics: A couple years from now, you’re going to walk into a bunch of different businesses, and you’re going to see robots next to people. That’s the next big challenge: to have robots operating in and around people safely and effectively.

You’re going to begin to see that in the streets with AVs, and I think we’re going to continue to see advances there. But you’re going to see that in the home, you’re going to see that in restaurants, you’re going to see that in kitchens. Those are the places where I think the stuff that is being worked on today is really going to come to fruition.

Robots and the evolving labor market: The pandemic has exposed many of the fault lines that were there, but had not really been paid attention to, around the supply chain, around the danger and costs of some of the jobs that people were doing. It’s been an incredibly accelerating force, making people think about what we can automate. It has also pointed out that we have a lot of work to do.

But so much of the conversation that society is having now relates to things that have been talked about in robotics for a long time. So I’m really excited about the future because of that.

Refraction AI delivery vehicles vs. Amazon Scout sidewalk robots: You’re going to need a bunch of different types of robots to actually serve this problem. Sidewalk robots represent one part of that. They can handle that last quarter of a mile, last tenth of a mile. [For longer distances], we are firm believers in the idea that these robots should travel in this liminal zone between the sidewalk and the street. There’s these really fundamental limitations to traveling fast on sidewalks.

Research vs. commercialization, and working with tech giants: We need to be working on things that are not driven by the business models of the companies that exist today, particularly for developing new technology like robotics. The business models for robotics have not been worked out yet. And they are not going to look the same as the information capitalism of the FAANG companies of the world.

Without an investment in the long-term future of robotics, we may get a couple of short-term wins, but I don’t think we’re going to see the benefit to humanity that we will if we really stay invested. That’s 100 years, not next quarter.

Let me not come off as negative on trying to translate your research into the commercial sphere. I’m deeply encouraging of it here at CMU, and of the faculty that are doing it here. It’s just critical to make sure you do that with eyes open.

Contrasting the industrial evolution of Pittsburgh and Detroit: With the exit of steel, you see the rise of healthcare as this massive component of the economy here in Western Pennsylvania. And that provided an economic base for the universities to be able to really grow and thrive.

In Southeast Michigan, the auto industry has gone through ups and downs, but it didn’t have this immediate and final departure, which then forced a lot of reconfiguration of the economy (in the Pittsburgh area).

They’re just on different trajectories. They both have a lot of the same challenges, which are trying to figure out how you bring that economic benefit to all the residents of both cities.





Source link