Predicting the future of robotics

Tom Wagner seems to have a robot for everything.

Need a set of eyes in the air or beneath the sea? There's a bot for that.

Need to vacuum but don't have the time? There's a bot for that.

Need information about a place no man can tread? There's a bot for that, too.

Started in 1990 by a trio of Massachusetts Institute of Technology roboticists, iRobot specializes in making robots for home, government and maritime use.

I spoke with Wagner, the company's CTO, about what's boiling in the company's Bedford, Mass.-based lab -- and what current problems future robots can really solve.

SmartPlanet: What can robots do for us?

TW: I would separate the robots of today from robots of where we would like them to be.

On the government side, we have robots called PackBot, SUGV, Warrior -- these are all track bots used in the field in a tele-operated sanction. As an operator -- the gentleman that's controlling the robot -- you'd use a laptop and a gaming controller and send the robot in the field to see something close without being near it.

It's made tremendous difference in the way wars operate. Today, we send out the robot so the soldier stays back.

But robots can do some of these functions in a more autonomous faction, without the human providing the small, detailed controls. I would call it "supervisory control." Instead of the operator gamesticking it out on the field, he can point on the map and the robot would autonomously drive its way to the path, avoiding obstacles along the way.

The same thing applies to building exploration. In the near-term, 2012 or sooner, we have a robot that we can send in the building and avoid walls and people and learn a map as it goes.

We also have a product called the Seaglider that can go out to sea for months at a time. It changes its buoyancy, rather than having a propellor on the back that drives it.

We have a variety of underwater robots in development. The Ranger is for shorter missions and goes faster. We have something called a Transphibian -- it's like a robotic turtle. It can hover in the water, settle on the bottom and stop motion, et cetera.

SmartPlanet: And, of course, the Roomba.

TW: On the home side, Roomba was released in 2002. It's actually a very special thing -- it's an autonomous robot that will clean your house. It's doing amazing things technically to do so, like follow walls, detect cords and tassels. It's actually a very sophisticated little robot.

In contrast to defense applications, the Roomba is autonomous.

And the problem it solves is harder than it looks. Let's say I come home at night -- I put my briefcase and shoes in a different place every night. The kids left their cords out. The dog is in its way. All of these things make the Roomba's job hard.

SmartPlanet: Where do we want robots to be in the future? What do we want them to accomplish? Or is it just for them to do more things better?

TW: That's actually a good way of expressing it: "more things better."

When you ask about robotics of the future...we want to give you robots that are more autonomous and do more functions for you on a daily basis. We focus very hard on solving problems that make a difference.

On the home side, it's problems that give you back time in the day.

On the Pentagon defense side, it's problems that put lives in danger.

SmartPlanet: What kind of problems do you envision robots solving?

TW: To me, a robot that would pick up after my kids would have infinite value.

What can they do by themselves? How can I make it more autonomous to do those things?

We can talk about Rosie from the Jetsons -- she puts things away and does the dishes and all kinds of things. I'd like to ship you a Rosie. Until then, you'll see more robots that are specific in function, like cleaning gutters.

You'll also see robots that work with other robots in the home. You can also imagine robots talking to your appliances. Projecting to the future, your home is highly networked and instrumented.

There's some technical stuff in there that we are doing today. One is natural interaction with people, understanding gestures and speech. We have active projects in that.

Approaching it in a limited fashion -- that's what makes these projects tractable, or solvable.

You can also teach a robot to identify people in the environment, follow them, recognize them, so they can differentiate people from chairs.

There are other things that we're doing to enable these capabilities, like our robot control software called Aware 2 that's on our advanced platforms.

SmartPlanet: What is your greatest challenge?

TW: It's hard to solve some of these problems in a general fashion. I might be able to make you a robot that can empty your dishwasher. But then I'd make you a different robot to pick up your kids' toys. And a different robot would mow your lawn. And another for folding your laundry.

That's a different problem than making a general-purpose robot.

In robotics, it's extremely hard to make a practical, reliable robot that you can ship as a product that solves a function and solves it well. I can make a robot that opens a door, but go to Home Depot -- there are a ton of different door handles.

SmartPlanet: Do robots fit into the greater vision of a smarter home?

TW: In your vision of the automated home -- the home is self-regulating its power consumption -- robots will absolutely be central to that. You'll probably be talking to a robot to control your home.

For the smart grid, they're controlling when people's air conditioning comes on. As your home becomes more automated, robots will absolutely have a role in that.

SmartPlanet: How about the healthcare industry? Much has been said about hospitals and robots.

TW: If you look at the demographic trending of an aging population, we're interested in helping people stay in their homes longer as age and disability occurs. Better support for people in their homes and in managed care facilities and hospitals.

On the home and consumer product side, they're more price-sensitive.

We've demonstrated automated vehicles on the industrial side. But to do that requires sensors that cost more than a home product can handle. The challenges and value proposition are similar. It's either a "reduce work" -- an efficiency improvement -- or "if I can take a person out of harm's way," or it's a combination.

You will see across the board, on all of these fronts, continued push upward on autonomy. We're still very far from what we see in movies. We're a long way from robots that problem solve the way they do in movies. But by carefully defining the application, the robot can understand a part of the world and accomplish a task.

SP: And what of human-robot interaction? Artificial intelligence?

TW: The robot in the near-term won't have the ability to interact with humans. If I bring one into my office, it won't understand the clutter on my desk, or what a telephone is. Robot general intelligence will lag behind humans for a long time. Robots can do certain tasks, but it will occasionally require help from a person.

The other piece of this is that you're going to see increasingly us work with robots as a social partner. It won't be my DVR, but my friend who's with me in the world. We see this already: people name their Roombas. As robots become more autonomous and able to interact with you, you'll see more and more of this.

The problems of understanding the world and moving around in it are very hard for a machine. A two-year-old can operate at a [intelligence and reasoning] level that surpasses automation.

This post was originally published on Smartplanet.com