Roomba’s Long, Bumpy Path From Prototype to Your Living Room

Photo-Illustration: Photos: Steve Gorton and Tim Ridley, Alexander Hafemann/Getty Images

The idea came to Joe Jones when he was working in the research department at the MIT Artificial Intelligence Lab in 1989, sparked by an innocuous challenge among colleagues to build something cool with Legos during a semester break.

“I decided I wanted to build a robot that would clean my apartment, because I’m basically a slob,” Jones, who was 36 at the time, recalls. “I thought it would be fun.”

That DIY Lego project eventually became the Roomba, perhaps the most novel, beloved, and successful household robot of all time, which has since bounced off the walls of more than 15 million homes worldwide on its way to becoming a pop-culture phenomenon, from Parks and Recreation to Saturday Night Live.

Roomba inventor Joe Jones, 2006. Photo: Courtesy of Joe Jones

But getting the “Dust Puppy,” as it was once known, to market, turned out to be a decade-long lesson in humility. One early boss, who was highly unimpressed, accused Jones of “playing with toys.” Some months, he couldn’t afford his mortgage. From conception to final product, the long journey of the Roomba is one of rejection, setbacks, failures, patience, fortuitous events, and, ultimately, triumph in the form of adorable autonomous tidiness.

Jones, now 63, the author of three books and the holder of more than two-dozen patents, walked us through the process of developing a product that would sweep a nation.

Where were you when you first conceived of a floor-cleaning robot?
I was living alone — 36 at the time, no spring chicken — and put together a little robot, made mostly of Lego with some microprocessors and some switches and a bumper to make it turn away from any obstacle it would hit. And I built a little sweeping mechanism that started off with a model brush that I chopped up in a certain way and glued gears to. It sort of worked! It didn’t fulfill my hopes that I could use it to actually clean my apartment because it wasn’t rugged enough, wasn’t robust enough. But it seemed to me that, yeah, with a little bit more engineering you could actually build something and it wouldn’t cost that much. I thought it was a great idea, but I was still working at the AI lab at MIT, so I kind of filed it away and didn’t do anything with it. A year or two later the AI lab ran out of money — laid off the research staff. I was in that batch.

“This is a photo of the earliest incarnation of Roomba. I built it for the AI Lab Robot Talent Show in 1989. It was called Rug Warrior because an earlier version looked like a contraption from the Road Warrior movies.” Photo: Courtesy of Joe Jones

Then what?
I found a job right away with another company called Denning Mobile Robotics. Denning made these big robots, the size of a dishwasher. So I figured, Why not try and get Denning to build this little mobile robot that will clean your floor and will cost way less?

Did you show them the Lego version?
I didn’t want to show that one. I wanted it to be a little more developed than that. I had met this other mechanical engineer who worked at Denning, Jack Shimek, and I got him interested in the project, and we spent a few weeks building up a new model. It actually incorporated one of those Bissell carpet sweepers. We added motors to it, a shell around it, added a microprocessor to program the thing to run around and bump into walls and sweep things up.

After a few weeks we thought this was great. Everyone’s going to love this. We showed it off to the whole company one day and it pretty much did what it was supposed to do. It ran across the floor, swept stuff up. We thought, This has legs, something’s going to happen with this.

Ten days later we were both laid off.

Ouch. So that was strike one.
Exactly. It wasn’t a great time for Denning. They were laying people off. It probably would have been a long shot for them to have adopted it either way, but we thought it would work. The company only went on for a year or so after that.

“This is Scamp, the first iRobot version of Roomba. It was the result of the ‘two weeks and $10,000’ project Paul Sandin and I did in 1999.” Photo: Courtesy of Joe Jones

So it’s February 1991, and you find another job at iRobot.
In those days it was called IS Robotics, before the name change. And, of course, this was a small robot and they did believe small robots could add value and turn into real things. They were supportive, but in those days iRobot just didn’t have the money to develop that sort of thing. We needed a partner. So we wrote a letter to Bissell.

The same company whose brush you rigged up in the early prototype.
We thought, This is similar to the carpet sweepers they already have, a similar price point. Bissell said nobody will pay more than $20 or $30 for a carpet sweeper, so you add a bunch of stuff to make it a robot and no one’s going to be interested. So they sent us packing.

So strike two — and nothing happens again for five years. You’ve said that the first blow was the worst, and that everything else just seemed a natural progression toward the project’s completion, but that still seems like a long stretch in the wilderness.
Right. So in 1996, iRobot did this project with SC Johnson. They’d come to iRobot and had wanted to develop a new robot for a new business, and wanted a robot that would go into a department store afterhours and vacuum and scrub the floor. So we worked on that for two or three years and we got to the point where we had the three things the robot needed to do, but on three different platforms. One would navigate, one would clean the floor, and one would move around autonomously and demonstrate the sensors.

It wasn’t too long after that that they decided to pull the plug.

They ultimately concluded that even if they had the robot that they wanted, they couldn’t break into the contract cleaning business for a number of business reasons. We were all pretty sad about that.

“Roomba needed to operate on all kinds of floors so we built a test fixture that tried to test as many different kinds as we could manage.” Photo: Courtesy of Joe Jones

Strike three … At what point were you tempted to just say forget it and move on to something else?
[Laughing] One thing I’ve learned about robots over the years is they always break your heart. One way or another. What you do is try again.

The fact that Roomba actually became a product is so incredible. Looking back on it, it seemed like the chances were really infinitesimal that it would ever happen. iRobot still didn’t quite have enough money in 1999 to support the development themselves. So what we did was to go out and look for financial backers.

We went out to the Proctor & Gamble headquarters, but they had an internal group that was also working on small robots, so the company wanted to have a bake-off. We went up to this house, the house of some executive, so I had to wait in the basement with my robot while the other guy from the internal group was upstairs and demonstrated his robot. Then at some point he came downstairs to wait in the basement and I went upstairs to demonstrate my robot. He wouldn’t even look at me.

The robot the P&G guy had demonstrated was actually based on a robot I designed for a book I wrote. He was using a robot I designed in competition against me. And they won.

“The first time the factory tests their molds, they just use whatever color of ABS plastic they have sitting around. Roomba was manufactured initially at Jetta, a company in China that usually made toys, ergo the colorful plastic.” Photo: Courtesy of Joe Jones

So now it’s been nearly a decade since you first conceived of Roomba, almost eight years since joining iRobot, and about five years since you first met with SC Johnson, and you decide to go back to them.
We did, and SCJ decided that they would support the development of Roomba. It was December, 1999, when we finally sta
rted working on it.

But there was a problem. SC Johnson was selling consumables. They insisted that Roomba had to have a consumable, so we spent a year developing a collapsible dust cup that you would buy at the store, then throw away. Now, that was silly. Roomba didn’t need anything like that.

Some upheaval happened at SCJ, and whoever made the decision to support Roomba got a new job or kicked out, and the new guy put the brakes on it. So they stopped supporting Roomba after they spent $1 to $2 million.

Plastic parts from the preproduction run. Photo: Courtesy of Joe Jones

At this point, did you go back to the drawing board, or did it feel like all was lost?
It’s actually really fortunate that both of those things happened, first that SCJ stepped in when they did, because if they hadn’t, iRobot would not have had the money left in the bank to support Roomba. And also because if they hadn’t gotten out when they did, Roomba would have had a component it didn’t need, and there’s a good chance it wouldn’t have been successful for that reason.

So that’s why it seems so incredible that Roomba ever saw the light of day. Both of those events were critical to getting it to market.

So you have this perfect storm, part luck, part perseverance, and iRobot decides to go all-in with their own money. They’re giving you the chance to put your idea into production. Were you apprehensive?
We were worried for a short period that iRobot would pull the plug, but they didn’t. They went ahead with it. Roomba launched in September, 2002. It would have been early 2002, or late 2001, when SCJ pulled out.

We could look ahead and see that we needed lots and lots of money for inventory. What iRobot did was they laid off a bunch of people so they would have enough money for the inventory. Unfortunately, they laid off the guy who had been the Roomba team leader and another one or two people who had worked on the project.

And then it sold north of 15 million units.
It was really surprising to us. We were thinking, 15,000, 50,000, or 100,000 units — then within a year or two they were selling a million units a year. The iRobot team had never designed a consumer product before. We were all novices at this. We felt like we’d done something.

“One of the first production Roombas in its initial packaging.” Photo: Courtesy of Joe Jones

Do you have your own Roomba at home?
I do. I have a Roomba that cleans that living room and a Scooba that cleans the kitchen.

Do your kids know that these are your robot brainchildren?
Yes, I made sure that they knew that. I think I get enough credit for what I did. When you read an article about Roomba, the approach that the reporter usually takes is whoever they’re talking to is the one who invented Roomba.

Is that frustrating?
No, not really. It wasn’t done by just one person. Having a good idea just isn’t enough. That good idea takes a lot of work from a lot of people to take a good idea and turn it into a good product. The fact that I had the idea would only be a small part of it.

What would your advice be to future innovators?
Don’t count on the strength of the idea to carry you all the way to the goal. That’s just one part of it. I didn’t expect instant success at any of these times. It would have been great if it had happened, but even then all you get is a shot. If things go well, you get a shot at success. There’s still no guarantee, even when it’s supported and gets all the resources and money that it needs.

Roomba has invaded the cultural consciousness, with cameos on SNL and Parks and Rec. It’s now part of the zeitgeist. Have you seen these Roomba cat videos?
[Laughs] Those things are what amazed us most. Nobody ever foresaw that people would put cats and babies and dogs and whatnot on top of Roomba and have them run around.

Do you approve?
Well, to me at least, the most horrifying application to which Roomba has been put … there were some folks who filmed Roombas being Froggers, like the video game.

Because your baby is dodging traffic, getting clipped by cars on the freeway?
I think that’s very ill-advised.

Roomba’s Long, Bumpy Path From Prototype to Your Living Room