Humans and robots: Stefanos Nikolaidis fosters anthro-machine collaboration in manufacturing 


Stefanos Nikolaidis, assistant professor of computer science at USC, studied what it means for humans to trust a robot as part of his work in algorithmic human robot interaction. (Photo courtesy of Stefanos Nikolaidis)

What do robots, manufacturing and stroke rehabilitation have in common? Stefanos Nikolaidis, an assistant professor in computer science in the Viterbi School of Engineering, has worked on these problems before, combining a different side of humanity with robotics.

In July, Agilent, a life sciences company bringing technologies to laboratories worldwide, presented Nikolaidis with the Agilent 2022 Early Career Professor Award for his work in human-robot collaboration in manufacturing. This award comes at a time where the ties between automation, manufacturing and the supply chain are more important than ever as a result of the pandemic, Agilent said in a press release.

Nikolaidis’ journey in robotics begins far earlier, though, as a university student in Greece, where he studied electrical and computer engineering at the National Technical University of Athens.

“One of my professors there was a robotics professor, but he was also in a wheelchair because he had [a] surgery that didn’t go very well,” Nikolaidis said. “Therefore, he had very limited mobility, but he was also doing research in robotics as a way to improve not only his own life … but also to improve other people’s lives that are in a similar situation. I found that very inspiring.”

From there, his journey has spanned multiple countries and continents, eventually bringing him to Los Angeles, where he currently studies how robots can be integrated into a human team — particularly in manufacturing.

“There has been a lot of work in robotics that has focused on how we can make robots more and more autonomous — for example, how we can have a drone fly by itself, how we can have a car drive by itself,” Nikolaidis said.

But Nikolaidis finds the reality of integrating robots into a human team to be very different from the sci-fi horrors of robots taking over the world. 

“The robot needs to be able to adapt to the preferences of the human teammates and have a shared mental model with the other teammates to be able to be a productive team member. This is where the core research question in human robot interaction is,” Nikolaidis said. “[We are researching] how we can develop algorithms that enable a robot to be productive and efficient, but at the same time to act in a way that is acceptable and in a way that is perceived as trustworthy by the human workers.”

This idea of trust has been an integral part of Nikolaidis’ work ever since his time at Carnegie Mellon University under his PhD advisor Siddhartha Srinivasa.

“[Nikolaidis] specifically looked at a couple of aspects of algorithmic human robot interaction. One of them specifically is around trust,” Srinivasa said. “What does it mean for the human to trust the robot? I think we all sort of qualitatively know what it means to trust someone, but how do you turn that into a mathematical model?”

Nikolaidis’ work is heavily focused on solving real world problems in manufacturing and other industries. When he first arrived in L.A., he visited stroke patients at the Rancho Los Amigos National Rehabilitation Center to see how he could use robotics to make an impact.

“I asked stroke patients there, ‘If you have a robot arm like this — I showed them some videos, some slides — what would you like the arm to do?’ People said things like assistive eating — something I worked on in the past — but what really made an impression is that a woman said, ‘I really have messy hair and that affects how I feel and my self image. I would really like a robot to help me comb my hair,’” Nikolaidis said. “We developed a hair combing system and we have been iterating on that over the years.”

Nikolaidis’ effort in speaking with those impacted stood out in particular to his faculty mentor, Maja Matari, professor of computer science, neuroscience and pediatrics.

“Engineers come up with many ideas about what stroke survivors may need. For example, engineers might think, ‘Well, they need a robot to reach things for them.’ But really, those are uninformed guesses. Unless you’re a member of the stroke survivor community, you really don’t know what it’s like and what is really needed, so you should not presume,” Matarić said. “It is really wonderful to see a young researcher like Nikolaidis not presuming, but taking the time to learn directly.”

Matarić recognizes the difficulty in these robotics problems, and said she resonates strongly with Nikolaidis’ willingness to tackle them.

“Too few researchers in robotics work on truly challenging real-world problems. Instead, researchers often find it easier to work on simulations or in toy evaluation domains, but Nikolaidis is willing to take on the hard problems, such as stroke rehabilitation, with real stroke patients,” Matarić said.

Since human-robot collaboration is still a relatively new field, there are many new outlets to be explored, especially the unpredictability of human nature. Srinivasa recognizes the difficulty of managing this mercurialness in fields such as autonomous vehicles.

“The problem is these pesky humans, because we drive and do things, like one [piece] is our intentions and our cost functions are hidden and latent and very personal. I think the second piece is that we are highly suboptimal agents, right? It’s not like we compute Nash equilibria and act accordingly,” Srinivasa said.

This challenge of finding a solution and tying together the volatility of human nature and the predictability of robots is what makes the work fulfilling for both researchers. Nikolaidis sees the potential for robots to help out at home, tying back to his original inspiration to pursue robotics.

“If we are to introduce robots at home, … by definition, a human environment where everything is chaotic and there are multiple users, then I think this is a very challenging application,” Nikolaidis said. “I think this is right now … one of the main goals in human robot interaction, which is how we can have a robot at home and be able to assist with everyday tasks and especially how it can help people with different types of disabilities and people with limited mobility.”

Although he is now at the University of Washington, Srinivasa said he cherishes the time that Nikolaidis spent with him.

“I find [Nikolaidis] to be fearless and committed to running user studies, to studying cognitive psychology papers, … to doing everything that is needed to make sure that humans and robots work effectively together. You have to be very selfless and accept the fact that to get a robot to do anything, you have to really do everything. I think [Nikolaidis] is one of the best that I’ve worked with when it comes to acknowledging that,” Srinivasa said.