Make Room for Robots
August 01, 2013

If WPI computer science professor Sonia Chernova has her way, robots will soon be taking over.

Not in the 1950s sci-fi movie sense, but in terms of assuming a wide range of everyday tasks and applications. In fact, much of Chernova’s research is devoted to making robotics less of a niche technology and more accessible for general use.

She says that robotics is currently at a “tipping point,” poised to make the transition to daily life just as computers did in the ’70s and ’80s.

“At the moment, robots operate primarily in highly constrained and customized environments, such as factory production lines,” Chernova explains. “Or else they are remotely controlled by people, as with NASA’s Mars rovers.  What we are seeing now is the emergence of robotic technologies for less-constrained environments.”

One such example is telepresence, which uses robotics to allow people to participate in events and activities at a remote location, such as school or work. It is more powerful than standard videoconferencing, she said, because it facilitates greater interaction and movement from room to room.

Another area is medicine. More than 1.5 million surgical procedures have been performed with the aid of robots, she says, and medicine-delivery robots are being tested in multiple hospitals.  Robotics is also finding its way into automobiles with the use of sensors that enable automatic parallel parking and crash avoidance.

So if robots are already in our cars, will they soon be in our homes as personal assistants, similar to the Jetsons’  Rosie?

“One day, yes; but not anytime soon,” according to Chernova.  “The average household environment is too complex for today’s robots to operate in effectively.” She adds, however, that the technological advances that paved the way for robotics to enter medicine, automobiles, and education suggest that day is getting closer.

Still, Chernova concedes that it’s impossible to preprogram a robot with everything it needs to be ready for anything and everything in the real world.

“If you buy a robot and bring it home, it may be able to figure out the layout of your house, but just like for a visiting friend, you will need to show it where the silverware drawer is, where you keep your towels, and which cabinet holds the dishes,” she explains.

Beyond that, it’s helpful to be able to teach the robot the way you like things done in your home. Toward that end, Chernova’s research focuses on algorithmic techniques and interfaces that will one day allow people to teach robots such skills without programming.

Learning, however, is a two-way interaction. The teacher provides instruction and the learner asks questions. Enabling robots to ask such questions is necessary for an effective learning process.  But getting a robot to grasp the limit of its own understanding, then communicate that effectively to a person, is challenging.

“Some robots currently have the ability to ask simple questions, but additional work is needed to be able to structure more complex questions, as well as to interpret the answers received,” she says.

Many of those old sci-fi movies depict robots as physically indistinguishable from humans. It makes for a good plot device, but in the real world, according to Chernova, form should definitely follow function.

“Some researchers strive to imitate humans in both their physical form and in terms of the algorithms internalized in the programming.  However I find that our understanding of how humans work is often limited, so I usually take the functional approach.  If four wheels work better than two legs, that’s fine by me.”


― Mike D’Onofrio