Julie A. Adams (left) and Heather Knight (right) in Graf Hall at Oregon State University.
When you imagine the technologically advanced world of the future, you are likely to think of robots. Robots are already sorting mail and manufacturing products, but they are not yet fully integrated into our lives in the way that science fiction depicts.
The potential for robots to improve our lives is tremendous. Already, researchers are working toward applications such as using robots to treat Ebola victims, assist first responders in evacuating a building, and help care for an aging population.
But all of those applications come with a host of questions. Can patients learn to trust robots? Will humans understand what robots intend to do? How will robots and humans communicate with each other? Can we create robots that we enjoy having in our lives?
Integrating robots into our lives is central to the research of two new computer science faculty members in the College of Engineering at Oregon State University. And yet their areas of research are very different.
"As an institute and one of the larger organized robotics programs in the country, we have to have a focus on policy." - Julie A. Adams
Advancing robotics safely
Julie A. Adams, professor of computer science, and Heather Knight, assistant professor of computer science, are part of the expanding focus on robotics and intelligent systems at Oregon State that was formalized by the establishment of the Collaborative Robotics and Intelligent Systems (CoRIS) Institute. Adams is the associate director of deployed systems and policy for CoRIS, and Knight is part of the core faculty.
Adams will be working to make sure that CoRIS is a leader in developing rules and regulations relating to policy and safety issues with robots and artificial intelligence. Additionally, as Oregon State’s technical point of contact for the FAA’s Center of Excellence for Unmanned Aerial Systems, she will engage Oregon State faculty in support of efforts to integrate drones into the U.S. commercial airspace.
“As an institute and one of the larger organized robotics programs in the country, we have to have a focus on policy,” Adams said. “We have to understand the ethical and policy questions. And we have to make sure that our students, many of whom will have jobs in industry, understand these questions and their implications.”
She makes the point that policy and ethics are particularly relevant for robotics graduate students at Oregon State, who are required to work with real robots, preferably in a deployed environment, which means outside of the laboratory.
Policy and trust are very relevant to Adams’ own research, which focuses on deployed systems where humans and robots need to work closely. For example, Adams is part of a team developing a system to control a drone swarm in urban combat situations for the Defense Advanced Research Projects Agency (DARPA). Her research could also enable first responders to work in teams with drones to find survivors of disasters such as the flooding from Hurricane Harvey, or to track the progression of wildfires.
Julie A. Adams talks more about robot-human interaction on the Engineering Out Loud podcast.
In those instances, you can imagine teams of drones and humans working together to collect data and passing it to a team leader. A portion of Adams’ work has focused on how to best form the teams and develop a plan to conduct the mission within constraints such as time or footprint (how many individuals can be part of the operation).
“This research is intended to support the human decision maker — to create intelligence in the system and reduce the workload for the human. The reality is that in all these situations, be they military or First Response, the human is not doing just one task, they are doing multiple tasks. And every time they are distracted that will impact the decision-making process,” Adams said.
Reducing the workload for the human is just one way of improving the likelihood of human-robot teams succeeding. Another aspect that Adams works on is the communication system between team members. For emergency situations it is critical that information between humans and robots is conveyed accurately and quickly in environments that are likely to be noisy, distracting, and stressful. One direction that Adams is exploring is creating a system to interpret the physiological response of a human so the robot knows when it is appropriate to interrupt and how to best interact with a person.
Designing friendly robots
Knight also works on human-robot communication, but in very different area than Adams. Knight’s research aims to develop socially intelligent robots that can interact in ways that are more acceptable to humans.
“There is a major shift in robotics right now, and a lot of the current funding and innovation in robotics is focused on integrating robots into the human environment,” Knight said. “If we want innovation in robotics to be able to continue we need to start mastering artificial social intelligence to make them more functional and to keep them from being rejected by people altogether.”
People expect robots to act in human-like ways. Body language, for example, is an essential part of human communication, but few robotic systems take that into account. Knight looks at ways in which robot behavior affects how people react to them.
She has found that how a robot moves — fast or slow, sudden or smooth — can be interpreted differently by humans. In one study where a robot delivered candy to people in an office building, the slow-moving robot delivered twice as much candy in the same amount of time as the fast-moving robot. People reported that the fast-moving robot “looked busy,” and so they may have been more reluctant to interrupt the robot.
For real-world applications, such as a bipedal robot delivering packages to people’s homes, it might be better to program a system that prioritized connecting with humans so as to not be threatening. But there could be other robot tasks for which looking busy would be more effective.
Knight has looked to the world of entertainment for inspiration in developing more expressive robots. Her joke-telling robot, Data, has traveled the world with her, performing for audiences over the past seven years. Data started out as a robot that would choose jokes based on audience response, but has developed into something more complicated, as Knight incorporates lessons from stand-up comedians who have to be spontaneous.
“Stand-up comedians have interesting, rather algorithmic, ways of dealing with things like hecklers that I find really fascinating, and that I think it could apply to a lot of robot service situations,” Knight said.
Knight’s passion for bringing humans and robots together goes beyond research. She is the founder of Marilyn Monrobot, a theater company that features robotic actors like Data, and creates electronic art such as the Rube Goldberg machine featured in OK Go’s video “This Too Shall Pass.” She is the executive director of the Robot Film Festival. And was the robotic artist in residence at X, the moonshot factory.
“I don't think you should design systems in isolation from the people they affect,” she said. “I explicitly incorporate people into my research methodology, but I think it's also important to engage with the public more broadly to expose the world to the concepts we’re thinking about.”