Automation: How iRobot's Roomba vacuum cleaner became part of the family

CEO Colin Angle talks about Roomba's path from research thesis to vacuuming your house, and why nearly everyone names their robot.

ZDNet talked to the CEO and co-founder of iRobot, Colin Angle, to discover the story of the Roomba's development and why people get so attached to their robot vacuum cleaners

ZDNet: You originally started out while researching at MIT. How long were you there for?

Angle: I started out there as an undergraduate. I was looking for a university that would feed my passion to build. At MIT, I was an undergrad before getting into graduate school. While there I established a partnership with a professor in robotics. Now he and I and later another person [Helen Greiner] between us founded iRobot and since then it is the only job I've ever had.

How did the three of you come up with the idea for iRobot?

It was partly out of frustration because we could read about robots in books, we could see movies, but the robots weren't there. Now at MIT we had developed in my lab, led by Professor Brooks and Helen, a strategy for making machines intelligent.

My undergraduate thesis was a six-legged, walking robot called Genghis. Now Genghis was very important because before Genghis, if you wanted to build a walking robot you needed a supercomputer. But Genghis was made out of aeroplane servos and his brain was an 8-bit microprocessor with 256 bytes -- not Kilobytes -- of RAM.


Genghis: Small but clever enough to walk over and around objects much larger than itself.

Photo: iRobot

It was beautifully efficient code with a multi-threaded operating system that allowed us to run many small processes on an 8-bit microprocessor. And the way we controlled it was inspired by insects.

Insects are very successful. They have tiny, tiny brains yet they can move around complicated environments. They're amazing machines.

SEE: Special report: How to automate the enterprise (free ebook)

We thought, instead of trying to build a human, how do we build an insect? And so, we created this strategy for building insect intelligence which we called, at the time, Subsumption Architecture, where one process can subsume another and you have hundreds of small, little processes which, in isolation do nothing, but by coming together they have an emergent behaviour which is very rugged robocity.

Genghis, this little rocking robot, could climb over rough terrain, balance itself when it was taking steps over huge rocks and do things that no walking robot could ever do.

Does that mean it could see an environment and work out a plan?

It didn't so much have a model, it didn't imagine an environment. It thinks, I should step here and then step there. It had strategies -- how to move your leg, how to support weight and when you have enough of your weight supported, how to move forward. And it you fail? Retreat and find a different path.

What it didn't do was high-level planning. But what it did do was respond very reactively to the environment. This type of intelligence is now called Behaviour Control and is at the heart of most robotics. It would be difficult to find a robot today that didn't use this strategy.


Angle: "The first business model of iRobot doesn't sound practical at all."

Photo: Colin Barker

And Genghis became very famous in the robotics community. It helped me get into graduate school and it's in the Smithsonian in Washington.

We started the company with the idea that we don't know what we will do but we have a toolkit that suddenly makes robots possible and we want iRobot to create this future.

The first business model of iRobot doesn't sound practical at all. We decided to build a robot to explore the moon. We were going to do a private mission to the moon and we were going to fund this by selling the rights to making the movie. The crazy bit was that we got quite far. We had the producer of the movie The Blues Brothers on our board of directors -- and we had a relationship with NASA to buy the data that we were going to collect. We had designed this micro-robot so that a very small rocket could get it there.

What happened was that NASA ultimately pulled out but the technology that we had created led directly to the Sojourner robot that went to Mars. So my name's up on Mars and iRobot got started.

Soon we found ourselves working on over 10 different things for 10 different markets, trying to find the way that robots were going to create the future.

What was the next step?

It was the idea of the robot vacuum cleaner and it required a great inspiration because it's what everyone asked me for. It became quite odd. I would introduce myself to people as Colin Angle of iRobot and rather than say good to meet you they would say, "When are you going to develop a robot to clean my floor?"

They would think, I am talking to a robot person, what could he do for me? There was this passion: How could a robot give me a better life?

At the time there was a cartoon on television, The Jetsons, and they had a robot, Rosie, the vacuum. People would say to me all the time, you're into robots, when are you going to vacuum my floor?

I had to have a response to this, so I'd say, "How much are you willing to pay?" But at that time, building robots was very expensive so I asked them if they would pay $5,000. They always said no, for that money I'd hire somebody.

As time passed, iRobot did a number of interesting things. We worked with the company, Johnson Wax. They came to us and said that we would like you to build a product that would help us to clean floors in supermarkets and shopping centres. We said, OK and we worked on this huge robot that would move through a shopping centre and clean the floors.

It worked but was not successful. There is a long story why that is, but it was all about Johnson Wax suddenly competing with its own customers. People who clean shopping centres do not want to work on a product to replace them. But along the way, iRobot learned how to clean.

SEE: How to implement AI and machine learning (ZDNet special report) | Download the report as a PDF (TechRepublic)

Now we know robots, but we were not experts on cleaning but now Johnson Wax taught us the methodology of cleaning. I don't know if you remember the game, the Tamagotchi? A small LCD video that was an artificial pet and you had to go and push buttons to feed it. It became a tremendous fad for a while.

We looked at it, thought it was silly but then thought of it as a video game pet. We thought, why don't we create a real robot pet? First we built a robot doll. And we took that around different toy companies and they said, "It's very nice but what we really want is a robot baby". So we partnered with Hasbro. We did this for about three and a half years and we built some good toys and had some success. Most importantly, because of them we went to the best factories in China and learned how to build at very low cost.

The third interesting thing was that we were working with the US Department of Defense on robots to clear minefields. To be exact, we were working on how you build a robot that could go to every area of a field without missing any spots.

And so, we created an AI system that would allow robots to ensure coverage. After working on that we realised we had solved cleaning, low cost and coverage. So, there was a day when one of the engineers came to me and said, "Colin, it's time to build the consumer vacuum because we know how to do it now".

Now I gave two guys a $10,000 budget to go buy some parts and see what they could do.

They came back with something, and we looked at it, thought it looked interesting, invested more money and we got more and more involved in it. Twelve years after iRobot was founded, in 2002 we launched the Roomba.

We had no idea whether this was going to be successful. I had to fight with my board of directors because you needed to build inventory. After tremendous backwards and forwarding they said, OK, you can build 15,000 of these robots.

We built 15,000, plus a little extra, and in the first three months after launch we sold 70,000. That changed iRobot forever. We went from a company that was a tremendous academic success to a company that was actually finding a way to create a new industry.

It's hard to imagine that it's been 16 years since we sold that first Roomba and we've sold 20 million. Our vacuum is now 20 percent of all money spent on vacuuming. In this past year, in the US, for the first time ever, vacuuming declined in sales, while robot vacuuming increased by over 15 percent.

It's not at all what I predicted but after 28 years, it's real.

And it just took off?

Yes. We screwed up in every way imaginable. I remember, on the first day that we sold this product; we sold 500 units on our website. I called up my friend at Hasbro and said, "Steve, we just sold 500 of these robots, is that good?" He said, "That's crazy good! Do you know how you're going to ship all these robots?" So I said, well, we have a printer and I can print out the labels and fill them in by hand, we have the list. There was a pause and he said, "You should think about shutting down your website right now because you don't know what you're doing!"

We ended up doing that because filling out the shipping labels by hand took the better part of a week.

From there, we had a very unusual period and all these different stores would call us and ask for the product. We were told that this never happens. Normally, you have to go out and beg to be in the store. We were told that we should have a big operation to handle all these calls. I hired an intern.

It was all quite odd. And we had to learn to know what happens if the robot breaks. That we didn't understand and even till today, returns have a strange aspect because over 90 percent of the people who buy the robot name it. They give it a name.

We took turns manning the phones and we had some very strange conversations. People would ring us to tell us that their Roomba had broken down. We would tell them how to send it in and we would send them a working one. They would then say something like, "No, I'm not sending you Rosie!"

They wanted a doctor to come and perform Roomba surgery. So, it was just very bizarre but also very exciting for us.

SEE: Sensor'd enterprise: IoT, ML, and big data (ZDNet special report) | Download the report as a PDF (TechRepublic)

Now we're a big company but there were still bumps along the way. In 2002, we had 70,000 robots. We thought that that was fantastic and everyone was very excited about it. In 2003 we built 250,000 robots, so we thought, this is going to be great. But it wasn't great.

Now, after Thanksgiving, after Black Friday, we still had 210,000 robots in the warehouse. And we thought, "Oh no, we're done". You see we hadn't done any advertising for it. We didn't actually believe in advertising. Somehow we thought that advertising was unclean.

So, we got together and thought about what we could do and we looked back to Black Friday and the period around them and, at the time, one of the sales guys had asked , "Why did our sales triple yesterday?" and nobody knew the answer.

Part of the answer was advertising but advertising at specific events. There was a famous Pepsi ad shown during the Super Bowl in which a robot devoured a pair of boxer shorts. Because of the advert our sales tripled. There were others that had the same impact. Because of that we survived a difficult period and prospered.

Also, one other lesson we learned. To be successful it was important that the Roomba looked like a machine. Research showed us that if it looked cute people wouldn't take it seriously.

But despite our best efforts at making it look mechanical, useful and kind of industrial, people still gave it a name. To them it's like a pet. A very useful, mechanical pet. But definitely a pet.

More on robotics and automation

Meet Norman, the world's first 'psychopathic' AI

MIT's AI experiment is based on the 1921Rorschach test, which identifies traits in humans deemed to be psychopathic based on their perception of inkblots, alongside what is known as thought disorders.

A cheap, easy-to-use surgical robot for the developing world

Healthcare inequality comes down to cost and lack of access to the best treatments. This startup believes robots are one answer.

New NVIDIA chip could enable AI and robotics in many more industries (TechRepublic)

Nvidia says the heart of their project is Jetson Xavier, the world's first computer designed specifically for robotics.

Melbourne children's hospital emergency department gets real-time wait updates

The Royal Children's Hospital Melbourne (RCHM) has released an online tracker that gives real-time information about wait times in their emergency department (ED) 24 hours a day.

How to take care of your robot vacuum (CNET)

Regular wheel and brush care will keep your little vacuum sucking... in a good way.

AI investment rising significantly among early adopters

Early adopters are ramping up investment in artificial intelligence technologies in 2018, reveals an AI study conducted by Constellation Research.

Microsoft drops data center into the sea: 'It will keep working for five years'

Microsoft has dropped a 40-foot long data-center pod onto the seafloor off the coast from the European Marine Energy Centre in Orkney, north of Scotland.