Most robots that fold laundry or greet party guests still look like rolling trash cans. The one now living in a University of Miami lab looks like a person.
The RoboCanes team, based in UM's College of Arts and Sciences, acquired the robot near the end of the spring semester. The Fourier Intelligence GR-2 stands nearly six feet tall and weighs 140 pounds. The team has spent months since then teaching it to complete household tasks.
A Field Advancing Fast
This summer, RoboCanes was one of only 24 teams worldwide to qualify for the RoboCup@Home league. It's an international competition where robots compete on tasks like folding laundry and welcoming guests to a party. It fielded the top-performing adult-sized robot in the league. It was also the only team from the Americas competing with a humanoid at all.
That mattered because this was the first year humanoid robots competed in RoboCup@Home. Previous years used smaller, wheeled robots.
"Now with the combination of new, fantastic hardware and the technology that AI is offering us, it's quite astonishing what we can do," said Ubbo Visser, the computer science professor who leads UM's robotics program.
Team leader Katarzyna Pasternak, a Ph.D. student studying human-robot interaction, says large language models did work that used to stall the whole team.
"There are problems that we had a hard time solving a couple years ago that are now very easily solvable by the large language models," she said. "Understanding complex sentences was a big task before. Now it's something that's very easy to understand and break down into actionable items."
Why the Hands Matter Most
RoboCanes is especially focused on the humanoid's fingers. They can open a bottle, unscrew a jar or hand an object to a person. That dexterity is the reason the team bought this specific robot.
"The beauty of having a large humanoid in the lab is that it opens up completely new functionality," Visser said.
The team's long-term goal is software for an autonomous household robot that could help care for an elderly person at home. Visser and Pasternak say that category could eventually fill labor shortages. Aging populations are growing faster than the workforce available to care for them.
The Gap Between Hardware and Reality
Getting there requires motion control, sensors, vision, language and navigation to work together seamlessly, something Visser says AI alone doesn't solve.
"What people underestimate is that to make a holistic humanoid that is autonomous and does something in the real world, a lot of integration is required," he said. "Sometimes, even with the help of AI, it's not that easy."
Human perception is its own obstacle. Pasternak still uses a smaller, three-foot robot for her own dissertation work. It involves teaching children about asthma medication in a clinic setting.
"Most kids are so excited to see the robot," she said. "I don't know how they're going to react to an adult-sized humanoid that looks a little bit like a Terminator."
What This Means for Miami
UM's robotics program just proved it can compete with the best-funded labs in the world on hardware most universities haven't touched yet. That result puts Miami on a very short list of places building humanoid robotics research at a competitive level.
The eldercare goal Visser and Pasternak describe is years out, not months. But the gap between "robot that wins a competition" and "robot that helps in someone's home" is exactly the gap this team is now working to close.