The Softer Side of Robotics
Chris Atkeson sees his cuddly, inflatable machines as caregivers for the ill and elderly.
By Jennifer Pocock
When you search Google Images for the word “robot,” picture after picture appears on your screen of hard, shiny, metal-and-plastic humanoids, seemingly built to resemble the Terminator or Transformers. Such 200- to 400-pound human-size “metal monsters” won’t work for the people Chris Atkeson wants robots to serve. “There’s a very important class of people that aren’t being served, and that’s people with disabilities and older people who develop disabilities later in life,” says the professor of robotics at Carnegie Mellon University. “You can’t safely get your typical science fiction robot into a home to take care of people.” It will crush small children and pets.
His inflatable robots, by contrast, are safer, yet superstrong and adaptable. They can fall down and bounce back, and withstand extreme conditions. “The misconception we need to correct is that inflatable robots are somehow fragile,” Atkeson says. “If you need to lift a part of a building or the hood of a crushed car, you use an inflatable device.”
Besides being malleable, inflatables are also less expensive to build. “Traditional robots are like making cars, heavy and expensive,” he says. “Inflatables are more like selling clothes or making pool toys. Both are relatively cheap.”
How can these devices work in the home? Atkeson gives a common example: An elderly man comes home from the hospital and has a hard time getting around, especially climbing the stairs in his house. “Climbing the stairs is the most dangerous thing he can do, but he doesn’t want to leave his house” and move somewhere else during convalescence, Atkeson says. Caregivers aren’t allowed to help people up and down stairs. It’s dangerous for both the disabled person and the caregiver if either should fall. Also, most people want the ability to control their own lives without someone hovering over them. “Instead, we could build a responsive inflatable system into the stairs. Or we could design a wearable inflatable device that would activate if he fell.”
Other possible devices are adjustable, inflatable beds and wheelchairs. If someone with limited mobility wants to get out of bed and go to the bathroom, it might take a long time for a nurse or caregiver to come — or he or she might be at home without any help at all. But a blow-up bed could help position a person to move into a wheelchair that could expand and contract to catch him. Robots could also be built into a bathroom — the room where most accidents happen.
CMU’s inflatable robots have made it to the big screen. In 2011, Disney was looking to base a movie on the little-known Marvel comic Big Hero 6. “They didn’t want another Transformer or Terminator robot. They wanted a unique thing,” Atkeson recalls. At first, Disney’s researchers didn’t know what to do with inflatable robots. “I told them the reason we build this is to take care of people, and they said ‘OK! That’s a story!’” Baymax, the cuddly vinyl healthbot, is now among Disney’s most-beloved creations. It made a big impact in China, where Atkeson now finds himself a celebrity.
In the classroom, Atkeson combines robotics and human problem solving. His graduate course, Dynamic Optimization, gets students to think of how to make machines and humans smarter. “I believe a large component of intelligence is optimization, knowing good ways to do things and finding better ways to do things,” he says. “I promise my students that they will learn not only how to program robots, but also how to program themselves more effectively.”
His assignments are “poorly formulated problems” that have no formulas for how a system should work and are subject to argument. Such problems range from “What should I have for lunch?” to the life-altering “Where should I go to college?” “Whom should I marry?” and “Should I marry at all?” Programmers haven’t yet reached the point where robots can make such decisions. Most often, Atkeson says, the closest they get is to come up with a related problem that can be solved, rather than the original. “It is clear in robotics and artificial intelligence that we don’t know enough to ask the right questions, and much of our research activity is figuring out how to ask better research questions,” he says.
Students either love or hate his courses because they are so abstract, he says. “I want them to go from asking ‘What does the professor want?’ to ‘What do I want as a learner?’” he says. But posing and answering open-ended questions will ultimately lead to better technology and – like inflatable caregiver robots – better lives.
Jennifer Pocock is assistant editor of Prism.
Image courtesy of Chris Atkeson/Thinkstock