How did the Roomba become the first robot most people let into their homes?

For most of the twentieth century, the household robot was a fixture of science fiction: a metal butler that cooks, cleans and chats. What finally arrived in American living rooms in the autumn of 2002 looked nothing like that. It was a round, flat machine about the size of a large dinner plate that bumped its way around the floor, sucking up dust. It was called Roomba, and it may be the most important robot most people have actually owned.

iRobot, an American robotics company, introduced Roomba in September 2002 (Wikipedia gives 17 September). By the end of that year, iRobot's website described it as a brand-new product, called it the first automatic vacuum in the United States and listed it at $199.95.

Where the idea came from

The inventor usually credited is Joe Jones, an engineer who first thought of a floor-cleaning robot in 1989 while working at MIT's artificial intelligence laboratory. His early models were built partly from LEGO bricks. In the 1990s he showed prototypes to a robot company and to the vacuum maker Bissell; neither wanted to make it.

The idea came back to life in 1999, when the consumer products company S.C. Johnson agreed to pay iRobot to develop it. The project was nicknamed "DustPuppy" before it became Roomba. S.C. Johnson eventually withdrew after spending between one and two million dollars, but iRobot kept going and brought the product to market itself.

How the first Roomba worked

The surprising thing about the first Roomba is how little it knew. It had no map of the house, no camera and no idea where it had already been. Before starting, the owner pressed one of three buttons to say whether the room was small, medium or large, which set how long it would run.

Its sensors were simple:

• A bumper across the front half that registers when the robot hits something.
• Cliff sensors on the underside that shine infrared light at the floor; if the floor suddenly "disappears," the robot knows it has reached a stair edge and backs away.
• An optional Virtual Wall, a separate box that projects an invisible infrared beam the robot treats as a solid barrier.

With those inputs, the robot switched between a handful of behaviours: spiralling outward from a spot, following along a wall, crossing the room in a straight line, and turning to a random new angle after a bump. No single behaviour cleans a room. Mixed together and repeated for long enough, they cover almost all of it.

This design follows the philosophy of Rodney Brooks, an MIT researcher who was also iRobot's chief technology officer. Brooks argued that robots should work more like insects than like chess players: instead of building a detailed model of the world and planning every move, they should react quickly to what their sensors report, using simple rules tuned to their surroundings.

A worked example: cleaning by rules

Picture a square room with a sofa against one wall. Here is a plausible sequence for a rule-driven robot like the early Roomba:

The robot never "knows" the room is clean. It simply keeps applying its rules until time is up. Because each pass is partly random, it crosses some spots several times and others only once, which is why it takes longer than a person with a vacuum. In exchange, it needs only cheap sensors and a small processor.

Why the Roomba mattered

Compared with competitors of the early 2000s that tried to map rooms, Roomba's random approach cleaned about as well and cost much less to develop and build, so it could be sold at a far lower price. That price was the breakthrough. A robot people could buy as a gift brought robotics out of the laboratory and the factory and into ordinary homes.

The Smithsonian's National Museum of American History holds an original Roomba in its collection and describes it as the first successful domestic robot. iRobot kept improving the line: models from July 2004 replaced the original, later ones added self-docking and dirt sensors, and the Roomba 980 of September 2015 introduced a camera-based system that builds a map of the floor. In September 2018 iRobot added a base station that empties the robot's bin automatically.

The Roomba also became a favourite teaching example in computer science. It shows that randomised strategies can succeed reliably in practice even though they offer no guarantee on any single run.

Is a Roomba intelligent?

It depends on what you mean by intelligent. The original Roomba did not learn, plan or understand anything. It followed fixed rules that engineers wrote by hand. In the vocabulary of 2020, that is closer to clever automation than to modern machine learning.

Yet it does meet a classic definition used in AI textbooks: an agent that senses its environment and acts to achieve a goal. Brooks's insight was that a lot of useful, lifelike behaviour can come from simple reactions rather than deep reasoning, and the Roomba is the best-known proof of that idea.

Newer Roombas move further toward AI. Camera-based models build maps, recognise which room they are in and let owners send them to clean a particular area. Each generation adds perception and planning, but the original lesson still stands: a robot does not have to be smart about everything to be genuinely useful at one thing.