�Թ���

When Brown University professors Sydney Skybetter and Stefanie Tellex first saw Spot®, both instantly had visions of how the yellow quadruped robot from �Թ��� could revolutionize their respective fields.

For Tellex, a professor of computer science, Spot represented a stable, nimble platform to support her research on natural language processing and autonomous navigation. “The first time I saw a video of Spot moving in a house, I was spellbound,” she recalls. “I loved the idea of a robot that you could take anywhere, and it could perform tasks in human settings.”

And for Skybetter, a dance professor and faculty director of the Brown Arts Institute, the sight of Spot dancing in the viral videos put out by �Թ��� was nothing short of revolutionary. “I was agog at the rhythmic complexity, the balancing and dexterity, the wit of the choreography,” he says. “This was, to my knowledge, the first rigorously choreographed, thoughtfully performed robotic performance that reached millions and millions of people.”

Both Tellex and Skybetter separately contacted �Թ���, looking for ways to use Spot in their own work. In 2021, an Office of Naval Research grant awarded to Tellex brought two Spot robots to Brown, with two more coming later thanks to additional funding. Spot’s arrival on campus sparked what some might consider an unlikely partnership between the Computer Science and Theatre Arts and Performance Studies departments at Brown. On one day, the robots will be following complex language commands from Tellex and her team, and on another they will be following the steps to Dua Lipa’s “Don’t Start Now” in Skybetter’s classroom.

Spot became commercially available in 2020, and many early use cases have centered on repeatable tasks in sectors like energy, construction, and public safety.

“The more capabilities a robot has, the faster we can go with the software stack, putting AI on top of that hardware. Spot was the first robot I’d seen that had that physical capability and control, with hardware that can actually perform in a human environment… very quickly it became the platform of choice for everything we do in the lab.”

Stefanie Tellex, Professor of Computer Science

Spot Comes to Campus

When the robots first arrived in Providence, Tellex expected to encounter the same complications with Spot as she did with previous robotics platforms—even wondering how much time the robots would spend suspended from a gantry, awaiting repairs. Instead, from the first time she and her team took Spot around campus, Tellex recognized that the hardware was markedly more reliable than anything she had encountered in the past.

“I remember the first time we took it out on a tour of Brown’s campus,” Tellex says. “I never, ever, in a million years thought we would have a robot that could go out like that. With our wheeled robots, we would be too scared that they would fall over. But with Spot, it was easy. It was no big deal.”

For Skybetter, seeing Spot arrive on campus was “close to a religious experience.”

Skybetter had already been planning a course that married robotics and dance using a  different platform. Once Spot became available, he immediately reconfigured the curriculum to revolve around Spot. The class was open to both dance and computer science students, and it quickly filled up. Even people who weren’t enrolled in the class would linger at the periphery, eager to interact with Spot; students who didn’t get into the class asked to be put on its mailing list.

“The robots are incredibly popular,” Skybetter says. “Within minutes of going online in our course offering system, the class was deluged with students.”

Max Merlin, a computer science graduate student at Brown, became the class’s course co-designer and student advisor. “The interface was so intuitive,” Merlin says. “The students didn’t need to know any sophisticated programming. I helped the teams a handful of times, but they were mostly self-sufficient.”

A Robust Research Platform

Tellex’s research focuses on getting robots to understand and respond to natural language commands—especially for navigation and wayfinding purposes. In the past, much of Tellex’s work was confined to indoor (and even simulated) environments, given the hardware constraints of previous robotics platforms. But Spot’s reliability and mobility opened up new possibilities for testing in a variety of real-world settings.

“The more capabilities a robot has, the faster we can go with the software stack, putting AI on top of that hardware,” she says. “Spot was the first robot I’d seen that had that physical capability and control, with hardware that can actually perform in a human environment. It can go up and down stairs, it can go through doors, and it can automatically avoid obstacles. The functionality on Spot is so good and so powerful that very quickly it became the platform of choice for everything we do in the lab.”

Tellex and her team focused on getting Spot to follow complex navigation commands. (Instead of “Go to this address,” the instructions might be “Go to the coffee shop, but stop at the bank first, and avoid the pharmacy.”) The team developed a system that can process information from Spot’s camera and translate English commands into robotic behaviors, and in some tests, the robot was successful up to 80 percent of the time. In one example, Spot was able to search for a “lost cat” (actually a stuffed animal) and find it by autonomously looking under a couch.

Tellex envisions a future in which robots are able to understand and follow navigational commands with the same success rate as humans. “You should be able to tell the robot to do anything that it’s physically capable of doing, and the robot should be able to do it,” she says.

Complex Questions, Exciting Opportunities

In Spot, Skybetter sees not only the future of dance, but also echoes of the art form’s past. He notes that “The Nutcracker” is about a “mechanical apparatus that comes alive” and “intervenes productively” in a fantastical social environment.

“‘The Nutcracker’ is the most popular ballet of all time, and it is fundamentally about a dancing robot,” Skybetter says. “The Western dance tradition has been dreaming about robots for centuries.”

For Skybetter, Spot’s emergence not only presents new artistic opportunities, but also poses complex, difficult questions about the nature of dance. For instance: Can a machine really be said to be “dancing”? Does it matter that Spot can’t hear or respond to music, but instead performs pre-programmed, prescribed moves? And what does it mean for the field that a company like �Թ��� has blazed the path for artists who want to incorporate robots into dance?

“It’s super complicated,” Skybetter acknowledges. “Corporations and emerging art forms have existed in dialogue with each other since forever. But what is new, and what is challenging, is when companies such as �Թ��� own the means of production. Ultimately, what it means to be an artist right now is to exist in relation to these tensions.”

The choreorobotics course that Skybetter developed alongside computer scientists at Brown attracted students from both technical and artistic backgrounds. Teams created everything from “languid moving solos,” to complex performances featuring both humans and robots, to one routine where Spot danced in time to a speech by former president Barack Obama. Despite some initial concerns about safety and technical complexity, students quickly mastered Spot’s intuitive interface, creating performances that impressed even �Թ���’ own choreographers.

“This course ended up proving that the robots are not limited to one particular dance vocabulary,” Skybetter says. “They are as dexterous aesthetically as they are technically, and that means that we have to do all kinds of work—in the classroom and in the lab—to understand what these robots are capable of.”

Now, as faculty director of the Brown Arts Institute, Skybetter is hatching plans to bring Spot to a larger stage—the school’s 500-seat performing arts center. “We’ve got a theater here that I want to fill with dancing robots,” he says.