A Long Goodbye to Baxter, a Gentle Giant Among Robots

For a serious research robot, Baxter is a charmer. It’s sports-car red, with two big and deliberate arms. Its face is a flat screen that telegraphs “feelings” like embarrassment (rosy cheeks, upturned eyebrows). If you’re so inclined, you can sit in front of it and make it read your mind to fix its mistakes. Or you can point to objects for it to pick up. If it gets confused, it can actually ask you for clarification, a seemingly simple interaction that’s in fact a big deal for the budding field of human-robot communication.

It’s research like this that’s made Baxter a legend among engineers in these early days of advanced robotics. No longer are the machines confined to factories—they’re beginning to roll and walk among us. That means we need to figure out how the hell we’re supposed to interact with them, and vice versa. Baxter is a pivotal member of the vanguard making that happen, giving researchers a platform to crack massive problems like manipulation.

But last week, the machine’s maker, Rethink Robotics, suddenly announced it was planning to fold, sales of its robots having come up short. (Baxter was originally designed for industrial use, for instance picking and placing objects on a factory line, and was later adopted for research.) The announcement was a surprise, but perhaps not entirely surprising—the industry is booming with a gaggle of competitors vying for the market.

Baxter isn’t the only machine at a roboticist’s disposal, of course, so research on human-robot interaction isn’t about to grind to a halt. But Baxter, as well as its single-armed next generation, Sawyer, have been indispensable in robotic research. They may not have taken over industry, but they've certainly established themselves as pioneers.

“Everyone in robotics research today either has a Baxter or has a friend with a Baxter,” says Brown University roboticist Stefanie Tellex, who’s in the former group. “In terms of penetration, it's the closest thing we've had to a common platform for manipulation across different research labs.”

What made Baxter so special was the fact that it was an all-in-one solution—camera, arms, grippers, sensing. Released in 2012, it was a standalone being that roboticists could program and modify to their heart’s content. On top of that, it was cheap as hell, at least by robot standards—Tellex got hers for a cool $27,000, while other robotics platforms could set you back hundreds of thousands of dollars.

That, though, came with compromise. Generally speaking, there are two ways of going about engineering a robotic arm. The brute you see on an assembly line is extremely precise because it has to be, or you’re going to get car doors bolted to car roofs. Accordingly, its actuators (the electric motors that drive the arm) are very pricey.

The other approach is to pull back on extreme precision in favor of a sense of feeling. These are the collaborative robots, or cobots. If they’re working with humans and not car parts, you’re more concerned that they not crush skulls than be 100 percent accurate. So while Baxter isn’t a particularly precise robot, it has sensors that allow it to detect when it’s made contact with a person so it can stop. That makes it ideal for researchers who don’t want their robots flinging them across the lab.

“I wanted to put Baxter out there because it had two arms and it was safe to work around,” says Rodney Brooks, founder of Rethink. “You didn't have to lock away the grad students. You could leave the grad students alone with the robot at night, which is when they work, and no one was going to get hurt.”

Don’t just take Brooks’ word for it: “That was one of the features that let us throw undergrads at it without worrying too much,” says Tellex.

So yes, students could experiment with the robot’s manipulation powers without fear of … complications. But researchers also began working with Baxter as a study in human-robot interaction, or HRI. Which might have had something to do, at least in part, with the engineers forming bonds with the machine. It does, after all, have that flat screen on its face.

“Immediately everyone starting putting their own faces there,” says Brooks. “And a lot of researchers were doing human-robot interaction. That was not in my mind, it was really about manipulation research.”

Not that object manipulation and HRI can’t go hand in hand (sorry). Tellex’s work in particular uses Baxter to research how humans and machines can communicate in unexpected and largely unexplored ways. Her robot can, for instance, ask for clarification if it’s not sure what a human wants it to pick up.

And let’s not forget that although Rethink didn’t sell as many robots in the commercial space as it was hoping, Baxter was still a major departure from the hulking, unfeeling industrial robots that came before it. The thing was just simpler and safer. “Baxter proved that industrial robots can safely work side-by-side with humans in the workplace,” says Rick Faulk, CEO of Locus Robotics, which makes warehouse robots. “It also, and more importantly, showed that robotic programming doesn't need to be complicated to be effective, allowing for a wider range of applications of collaborative robotic technology.”

Baxter and its sibling Sawyer were treading on strange new territory. In order to stay competitive, American companies large and small are turning to cobots. And the competition for robots to work alongside humans is growing increasingly fierce. “Rethink Robotics had a vision that was both inspiring and practical: to place safe robots that were easy to task using intuitive interfaces in the hands of any factory worker,” says Daniela Rus, director of MIT's Computer Science and Artificial Intelligence Laboratory. “The robots Baxter and Sawyer galvanized the robotics community to push the boundaries on robot manipulation and helped small- and medium-sized companies to introduce custom automation.”

Which is not to say roboticists are going to walk into their labs tomorrow and find that their Baxters have disappeared. The machines will continue laboring as pioneering research platforms. Rethink Robotics is folding, but Baxter will live on, until inevitably something better comes along.

“We're ready for another platform,” says Tellex. “We're ready for something different. And it's OK. Part of how the field advances is by us trying our technology on new and different robots, and learning about the general problems, and how could we solve them for all robots.”

So long Baxter—eventually. You done good.