Vivian Chu Working to Provide Robots Basic Building Blocks for Cognition
Georgia Institute of Technology robotics student Vivian Chu shares a familiar path to computer science with plenty of other students:
As a child, she loved engineering and computer science, taking things apart and putting them back together. Both of her parents were software engineers, so her road to STEM was paved long before she had the means to travel it.
She attended the University of California, Berkeley, for her undergraduate degree, where she earned a Bachelor’s degree in Electrical Engineering and Computer Science. She focused on Embedded Software, and, by and large, she enjoyed the experience.
But something was always missing when she took her computer science or electrical engineering classes. In them, she might design an algorithm or program a circuit, but she wasn’t seeing visual representation of her work in the way she wanted.
“You could put these things together, but there wasn’t a lot that you could actually see happen,” she said. “Then I took this one class where we got to program a Roomba to climb ramps or do other actions with an accelerometer. That was the first time things kind of clicked.”
She began to see and appreciate how a robot could understand how to interact with the world and also how it processed the information it gathered.
She got another taste a year later as a senior while working on an autonomous helicopter project. The helicopter didn’t do much – just hovering a few feet off the ground – but she realized during her work that she could sit in the lab for 12 hours without realizing it and come back excited to work the next day.
“That drove it home,” she said.
Now a Ph.D. student in robotics at Georgia Tech, Chu is interested in how to advance robotics to a point where robots could be deployed in care facilities or the home. Specifically, she is taking an approach of teaching robots the basic building blocks of cognition.
There are certain things humans learn as children that help them develop an understanding of the material world around them. A cup is a cup because it is fully containable, able to hold something like water inside; a spoon is a spoon because it can scoop other materials and hold them within its concave structure.
“If you could teach these robots these basic components, these basic building blocks, then when they go into your home, they could better reason how to perform other tasks,” she said.
Like making pasta. If a robot knows it needs something containable to hold something, heat to cook, and a spoon to stir, it could carry out that and other similar jobs.
Her inspiration came when she was working on her Master’s degree in robotics at the University of Pennsylvania. She attended a guest lecture by Georgia Tech alum Alex Stoytchev, who is now an assistant professor at Iowa State University. In the talk, Stoytchev discussed developmental psychology in children, how they explore basic actions and movements.
“A lot of my research is similar in that I want to teach these building blocks by having robots play with objects the way children play,” Chu said. “Adults give a child a nudge in the right direction here or there. Rather than having a robot do it blindly, we can have someone in the room and give it a bump here or there.
“It presents something that is much faster than a robot doing it on its own.”
The ideal goal is for a robot to truly understand its different sensory inputs. People use touch, sight, and sound, for example, to accomplish a task like turning on a lamp. Currently, robots are either very visual, which is the majority of the research, or incorporate touch.
“There’s very little being done to sort of merge these senses,” she said. “Audio is almost unheard of.”
Chu would like to achieve a scenario where the robot could understand that to turn on a lamp there is a touch component (learning the correct force with which to pull the rope), a visual component (to see where to pull, as well as whether the light turns on or not), and an auditory component (to hear the click as it pulls the rope).
“Those are all things I’m trying to research for my thesis,” she said.
The applications for this research are wide-ranging, but the enormous potential for the aging population is one of the aspects that interests Chu the most.
“As people get older, how do I make sure they could retire and have a dignified lifestyle toward the end of their life?” she asked.
Although she is still pursuing answers to these questions and is yet to defend her thesis, she was already recognized in the robotics community by Robohub’s 2016 list 25 Women in Robotics You Need to Know About.
The inclusion on the list took Chu by surprise, but she said it was rewarding because it acknowledges the importance of the work she is doing.
“As a Ph.D. student, the biggest fear is that you are going to write your thesis and no one is going to know about it,” she said. “That it’s just a document that gets tossed aside and doesn’t have an impact. It’s nice to know that, on a high level, there’s acknowledgement of what I’m working on.”
She plans to complete her degree within the next year, and still has plenty of goals she’d like to achieve going forward. While she is undecided whether she’ll pursue a career in academia or one working with a startup – a lifelong goal of hers – she knows she ultimately wants to impact society as a whole in any way she can.
“I often joke with my wife about the ways in which we can try to save the world,” she said. “But all jokes aside, for me, technology for just technology’s sake isn’t enough. The goal really is: How can the things I’m working on help improve the lives of those around us?”