
Original Article
Asa Whitney Professor of Mechanical Engineering Mark Yim has completed the second phase of a human-robot interaction research project at Penn’s ModLab.
The project primarily aims to develop Quori, a robot designed for cost-effective non-contact interaction between humans and robots. The project will be presented at 20th Annual IEEE/ACM International Conference on Human-Robot Interaction in Melbourne, Australia in early March.
The Quori robot is equipped with an expressive head, arms, a spine adapted for both forward and backward posture, and an omnidirectional base.
Yim and Simon Kim, associate professor of architecture were a part of an initial research team along with researchers from the University of Southern California, who built 10 copies of an early version of Quori to ship to schools including Johns Hopkins University, Carnegie Mellon University, and the University of San Diego.
Penn researchers were responsible for hardware, while USC researchers took the lead on software development. The project began in 2015 and was funded by the National Science Foundation.
Kim is responsible for contributing to robot design on the Quori project, including decisions about rotation, the placement of joints, and physical appearance. An architect and industrial designer by training, Kim said he had experience in assembling and designing structures but had not done so specifically for a human-robot interaction before.
“The idea of its expression of gender was important. The shape was important. Whether it was sharp or soft has big ramifications on how HRI is conducted,” Kim said. “For the HRI community, there’s quite a lot of statutes about how robots should look, especially for human-robot interaction. I had to quickly get caught up with that.”
The first Quori robot was previously exhibited at the Philadelphia Museum of Art as part of the “Design for Better Features” collection.
The second version of Quori — which will be presented at the conference — came about through collaboration with Doug Dooley, a Pixar animator who had worked on “Wall-E” and “Monsters Inc.” The project is funded by another NSF grant that aimed at producing 50 new models of the Quori robot.
“When it goes there, hopefully it’ll make a big splash. I mean, these are, this is the target audience. It’s the HRI conference. You’re talking about the one. We’ll have our first prototype. And it’s like, virtually done. It should be done, shipped out there, and then, hopefully we’ll make a big splash. We’ll have to just see,” Yim said.
Kim added that “there’s a huge amount of growth potential for robots and how they are going to be excited into care robots for companionship.”
The team held the last concept meeting for the face and head of the robot in early February, and was working on producing the actual parts for fabrication.
“Fingers crossed that it gets fabricated and sent to the conference,” Yim said. “I think it’ll look great.”
Andrew Specian, a 2022 School of Engineering and Applied Science Ph.D. graduate who served as the project lead for the first version of Quori — and was a former student of Yim — said he was drawn to the Quori project because it aimed to decrease costs associated with hardware, which is oftentimes the most expensive part of HRI research.
“It brings us a lot of emotion. When I was building these things, 10 seemed like a small number. Man, did I learn a lot of lessons that I don’t think Mark could have taught me verbally, walking through that challenge of building 10 human sized robots,” Specian said. “It’s great to see those efforts continue.”
As the project lead, Specian also worked on the math and programming for the robot’s motion. He worked in the lab until the first 10 robots were assembled and shipped out.
“[Yim] probably could have sat down there and wrote up the code and got it right in the first round, but going through the season of fine-tuning and understanding each line of code in each part of the algorithm, you can’t learn that in class,” Specian said.
“When you have that implemented in the real world, and you see, okay, this isn’t working because the controller is slow, the batteries are not plugged in. Really taking in those lessons and building and testing. You have to have that kinesthetic memory of robotics and not just the intellectual to be able to properly troubleshoot,” Specian added.
Specian described working on Quori as a learning experience, adding that he came into his Ph.D. with a much stricter mindset about readiness and testing, but is now more willing to try out the things he has tested to gather feedback for further iterations.
“Mark had another saying that I use a lot, and that’s to fail fast.” Specian said. “It probably took me all of my years to become as comfortable as I was. I’m still trying to practice that at work every day.”
Featured People
Director, GRASP Lab; Faculty Director, Design Studio (Venture Labs); Asa Whitney Professor, MEAM