Tech Industry

The leader of the robot pack

Could a Roomba be a "physical avatar"? iRobot CEO Colin Angle talks about this and other possibilities in the growing robotics market.

Call it the summer job of fate.

While a student at MIT, Colin Angle passed by a lab managed by Professor Rodney Brooks. And Brooks picked Angle to help him on a summer robotics project. The end result, unfurled in 1990, was a crab-like walking robot called Genghis. The project helped Brooks get tenure and Angle into grad school.

But rather than go to grad school, Angle started a company with Brooks. They later linked up with Helen Grenier, and their efforts became iRobot, one of the early standouts in what appears to be a growing market for autonomous robots. The company has shipped 1.2 million Roombas, a robotic vacuum cleaner that sells for around $300. iRobot has also created a military robot called the PackBot and is crafting a reconnaissance vehicle with mower king John Deere. Later this year, Scooba, a robotic mopper, will hit the market.

Angle, the company's CEO, met recently with to demonstrate the next version of the Roomba and talk about the future of the robotics market.

Q: When you talked to a year ago, robotics were still something of an oddity. Now there seems to be a lot more enthusiasm. Is the market taking off?
Angle: I think it's still young as an industry, certainly. If you view the practical robot market--that is, robots that are not humanoid or robots created for the sake of it--there may be five or six companies making lawn mowers. There's probably a dozen companies making vacuums, but no one anywhere near the scale that Roomba has been able to achieve at this point. There are a lot of industry analysts that are all very bullish on where it's going.

I know a bunch of venture funds that have money set aside for robots and are looking for good companies.

The U.N. report on robotics sees a sevenfold increase in robots from 2004 to 2007, and there are some Japanese people talking about 39 million household robots by 2010. iRobot has shipped 1.2 million Roombas.

Who is in second place?
Angle: Well, Electrolux, with maybe 40,000. Second place is not well defined yet, but it is true that there's a lot of really high-power interest in this today. I know a bunch of venture funds that have money set aside for robots and are looking for good companies. You're going to see some other good management teams being formed around...robotics.

Are your competitors going to come from university labs or established companies?
Angle: It's going to be hard to do exactly what we did and create an integrated system and establish a national brand and a national distribution platform. I think that there's an opportunity for companies to piggyback on top of iRobot platforms. The Roomba has a serial port, and we'll be announcing an API of sorts for it, so that you could create accessories for the Roomba if you wanted. If you did that, you'd have a solid base of over a million users.

There are a variety of different add-ons for the PackBot. The (reticulating) arm is a payload--it just bolts on. The fiber-optic spooler is a payload that bolts on, and there (are) various sensor heads. We did a payload where we took a company's chemical/biological sensors and then interfaced that to the PackBot. There was an issue where they were finding what might have been mass gravesites, and there were real concerns these sites were protected by toxic chemical or biohazardous stuff or perhaps even nuclear-reactive things to (keep) these sites from being excavated. The robots were sent over to help characterize the threat.

What sort of add-ons could you have for the Roomba?
Angle: For the Roomba, there is a group that's working very seriously and looking at the idea of using the Roomba as a physical avatar. I might log into a Web page and see what the robot sees, hear what the robot hears and be able to drive the robot from my Web page.

Over time, there'll be probably more of this type of interest. The rate at which it becomes economically interesting? I've no idea. We're putting up the APIs and doing what we can to encourage people.

How much of your revenue comes from consumers, and how much comes from military/corporate contracts?
Angle: Both divisions are comparable in size, by head count or whatever method you want. Both are very, very substantial, and both have high growth as it is. I mean, the PackBot is a product that's got a price tag of about $120,000 with the arm on it, and we've got 200 of those things over in Iraq.

Special coverage

Recent CNET
photo galleries feature
bots meant to be lovable
pets and others de-
signed to tote guns.

Are most of the (nonconsumer) contracts with the military?
Angle: Our focus right now is trying to meet the needs of the U.S. Army and address the challenges that are happening over in the Middle East. We've been told that we've saved over a hundred lives in Iraq, and we have substantial orders to yet fulfill in getting equipment over there, and there is feedback coming about how we can add functionalities to the robots.

We are also part of a very substantial program called Future Combat Systems. It's a $22 billion program (iRobot has a $50 million contract). The idea of Future Combat System is basically to make a unit that is extremely mobile, so it can be deployed anywhere in the world in 72 hours and have all of the information gathered by any element, whether it be a soldier or a robot. All of the information would be fed into a network, so the right people can look at the relevant data. So in an asymmetric situation like we're facing in Iraq, it's all about picking out the enemy (in) a sea of noncombatants and then getting results of that interpretation back down to give better instruction to the people on the ground.

How would a system like that work? Would it take a video feed and then scan it through an artificial intelligence system?
Angle: Could be. You could have large numbers of nodes gathering data and feeding that into the network.

The (military) robot's true value is...decreasing the battlefield fog.

What sort of research are you doing on swarming robots?
Angle: Once you make it cheap to have nodes, then the question becomes how many can you control with how many people. Swarming technology is all about unattended control of large numbers of things. We did a demo where we started with 128 robots in the corner, hit a button, and they rapidly diffused throughout the entire floor, creating (something) almost like a grid, so that you're able to download the physical plan of the place based on them talking to each other.

If you had a robot that needed to go and recharge, it would go ask its neighbor, "Hey, where's the recharger?" and the message would propagate the network until it got to a robot near a recharger who'd yell, "Over here." The robot needing the charge would then proceed toward where its neighbor thought that the rechargers were, and another robot (would) just similarly fill the empty space.

We did some other silly things. All the robots had really good, sound systems, so...we did self-organizing orchestration. The tenors and the sopranos would randomly give themselves a part.

What is the military's goal for robots?
Angle: The robots are about giving our soldiers new and important ways of doing dangerous missions. In Afghanistan, many of the caves had booby traps. Many of the caves contained unstable weapons caches, so (investigating those areas) was a lousy job. The idea of having a robot able to go in first made all the sense in the world. It was a better way.

We've evolved beyond Vietnam, where they would tie a rope around someone and lower them in a cave so if they got shot they (could) pull them out. Clearing buildings is almost worse than caves. Every time you go in a door you have to make a decision. Do you jump in? You could get shot. Do you throw a grenade in? It could blow up an innocent person in the room. The robot's true value is...decreasing the battlefield fog.

The other thing that robots actually allow for is the deployment (of) nonlethal technology: flash grenades, goop guns, things (like) that. The reason why today they are not in use widely is because an M16 works better (and seems the wiser choice for a soldier who's life is on the line). There is a different list profile associated with the application of robot technology.

Switching back to consumers again, in Korea and Japan, you see a lot of humanoid robots. Is there much of an opportunity there?
Angle: I think humanoid robots are interesting and economically irrelevant for the foreseeable future. The PackBot will climb stairs wildly faster and wildly more robustly than a humanoid robot, and you can take PackBot, you can throw it off of the second-story building and it will keep going. What do you think would happen if we did the same thing with a humanoid robot? It would be mangled, twisted steel.

On the entertainment side, I think there are some situations where humanoid robots have shown to be economically viable. The Robosapien was very cool, and it came at a great price point. You got more than $100 worth of fun out of the Robosapien.

Will it take time for people to acclimate to robots in the home?
Angle: Anecdotally, once you buy a Roomba, you're pretty happy. It's a better way of...handling the routine floor maintenance. If you have shag carpets all over your floor, you probably don't like Roomba because Roomba doesn't work well on shags, but if you have medium pile carpets and hardwood floors, and area rugs, you'll love it.