Socializing Robots

Heather Knight’s Research is Helping Robots Function in the Real World


  • newswise-fullscreen Socializing Robots

    Credit: OTTO Motors™

    Oregon State University roboticist Heather Knight is working to improve artificial social intelligence of robots such as Otto, an autonomous, mobile-transport robot developed by Clearpath Robotics.

Newswise — In “Star Wars,” R2-D2 is the perfect example of a likable and effective robot. Though he looks and sounds nothing like a human — with no face or hands, and communicating with only whistles and beeps — he clearly has a connection to his human co-workers.

“R2-D2 does a good job of illustrating that he’s paying attention. That’s important for people, especially in a collaborative scenario where you really want to understand what the other person needs,” said Heather Knight, an assistant professor of computer science at Oregon State University’s College of Engineering.

Although moviemakers have long seen the purpose in creating charismatic robots that fit in with human culture, roboticists have traditionally overlooked this area. But that is rapidly changing, according to Knight.

“There is a major shift in robotics right now, with much of the innovation and funding focused on integrating robots into human environments, from delivery robots to collaborative manufacturing,” she said. “We need to start mastering what I like to call ‘artificial social intelligence’ to make these robots successful.”

Artificial social intelligence would help robots interpret and mimic human cues — like body language, gaze direction, movement patterns, and facial expressions — to make them more effective at collaborating with humans, which is the goal of social robotics.

Knight was introduced to social robotics as an undergraduate at MIT, where she worked with Cynthia Breazeal, associate professor of media arts and sciences, who is considered to be the pioneer of social robotics. Knight went on to receive a master’s degree, with Breazeal as her advisor. She completed a second master’s degree and her doctorate at Carnegie Mellon University.

In addition to an impressive academic pedigree, Knight has one foot in the entertainment world. She has performed internationally, including on the TED stage, with her joke-telling robots. She is the founder of Marilyn Monrobot, a theater company that features robotic actors, and she has worked with Syyn Labs to create technology-based artworks such as the Rube Goldberg machine featured in pop-rock band OK Go’s “This Too Shall Pass” video. She is the executive director of the Robot Film Festival, and she was the robotic artist in residence at X, the research lab at Google’s parent company, Alphabet.

Borrowing methods from dance, improvisation, and stand-up comedy, Knight is bringing together the worlds of entertainment and engineering to make technology more effective and charismatic. Because the field of social robotics is still young, Knight says she sometimes has to explain to colleagues why it is important to integrate social intelligence into robots. But not at Oregon State.

“We have one of the most powerful collections of human-robot interaction researchers in the country,” she said.

Knight is part of the Collaborative Robotics and Intelligent Systems Institute (CoRIS) at Oregon State, which includes more than 25 faculty and 180 students who take a holistic approach to robotics and intelligent systems. Rather than focusing exclusively on advancing technology, the group considers how robots will operate in coordination with people.

ROBOTS FOR THE REAL WORLD

The institute’s mission is to create robots that function outside the comfort of a lab. One of the best-known examples of this is Cassie, the bipedal robot developed by Jonathan Hurst, associate professor of mechanical engineering.

Cassie was designed to handle rough terrain, for situations like disaster recovery or package delivery. In order to be successful in those situations, robots will need to act in ways that are predictable to people, which is the perfect opportunity for social robotics to have an impact.

Knight’s research has shown that people interpret the motions of robots as  having intention. For example, varying the motion of a robot carrying candy — fast or slow, sudden or smooth — can influence whether people think it is offering the candy to them or is too busy to be interrupted.

Integrating this knowledge from social robotics research into technology like Cassie is one purpose of CoRIS at Oregon State. Knight and Hurst hope to work together in the future to integrate social communication into how Cassie moves.

But even simple robots can benefit from social intelligence. Knight is working with Clearpath Robotics, a company that makes autonomous, mobile-transport robots for Fortune 500 companies. (Imagine a sleek, industrial-sized Roomba that can transport parts between workstations at a jogging pace.)

“Our vehicles are fundamentally co- workers to hundreds or thousands of people working in factories and warehouses,” said Ryan Gariepy, chief technical officer at Clearpath. “Increasing our products’ social awareness will mean that there will be a much more positive reception to their introduction into facilities around the world.”

‘IT’S LIKE A WEIRDO WHO DOESN’T FOLLOW THE SOCIAL RULE’

Knight and graduate student Bohkyung Chun started their work with Clearpath by studying the ways in which the employees interact with their own robots. Specifically, they looked at whether the employees anthropomorphize the robots, attributing human characteristics like personalities, intentions, and emotions to them.

What they found is that, although the employees rarely admitted to viewing the robots as creatures or beings, they interpreted the robots’ behaviors as if they were human co-workers. For example, one employee said, “When it drives too close to people, it’s like a weirdo who doesn’t follow the social rule.”

This is where social robotics comes in.

“One of the most important things to know about social robotics is that people don’t like to admit they anthropomorphize machines. But they’re really upset if the robots do things that don’t follow anthropomorphic rules,” Knight said.

Understanding what those rules are will help Knight integrate social intelligence into the robots. Now that Knight and Chun have completed the ethnographic study of employee-robot interaction, they will move into exploring how the robots can be modified to include social intelligence in ways that will make them safer and more effective at their jobs.

EXCUSE ME. CAN I GET BY?

For simple robots that don’t talk, another effective way to communicate is through motion. In a recent study, Knight and colleagues at Stanford University looked at what she calls “bystander interaction” between humans and robot furniture. These robots are similar to the mobile factory robots in that they can orient and move across the floor, but they must negotiate with people in order to rearrange in new configurations.

Their initial study on bystander interaction was set up so that the human was engaged in solving a word search puzzle while standing in a narrow passage where a chair robot tries to get by. If there was enough room, the robot would go behind the person. But if there was not, it would use one of three motion gestures: forward and back, turning side-to-side, or staying still. They measured whether the person moved out of the way to see which gesture was the most effective.

The forward-and-back motion as the only gesture to succeed the first time the robot attempted to interact with the person, because it was the clearest. However, that same gesture was also perceived to be the rudest. o, if the forward-and-back ion was repeated a second or third time, the person was more likely to get annoyed and less likely to yield. Whereas the other two gestures — side-to-side and staying still — were more effective the second or third time.

“It’s crazy how complex human behavior is,” Knight said. “What I learned is that in bystander interaction you want to be informative initially, but then communicate as little as possible because you’re interrupting that person.”

Knight will apply research on expressive motion to her future work with Clearpath Robotics. The transport robots also have the capability of using light and sound for communication, which could help make their messages more explicit and understandable to their human co-workers — and just a bit more like R2-D2.

“I like my work to touch the real world,” Knight said. “Sometimes that’s through live demonstrations or robot theater, but I also like touching the world by integrating my work into a real company. Being able to collaborate with Clearpath is a fantastic instance of using research to improve real robots.”

Chat now!
2.34052