The Roomba Now Sees and Maps a Home
The world’s most successful home robot, the Roomba, is getting some new skills. The robotic vacuum cleaner is gaining a camera and image-processing software that will let it find its way around an area as it cleans.
Up until now, the device from iRobot has relied on a rather simple approach to navigation. It moved more or less randomly, using sensors to change direction whenever faced with an obstacle or the top of a set of steps. This gets the job done, but it’s hardly very efficient or intelligent.
The Roomba 980, announced by iRobot CEO Colin Angle at an event in New York today, follows a new approach. It comes with a camera that captures images of a room, and software that compares these images to gradually build up a map of the robot’s surroundings and determine its location.
The 980’s new vision and navigation capabilities allow the robot to cover a carpet more efficiently, but they also lay the groundwork for more advanced features in future models. Right now the Roomba 980 uses its map only to make sure it’s cleaned a room properly; it cannot recognize objects or classify different rooms, for example. But such abilities are likely in the pipeline.
“Being able to localize in the environment is a foundational capability,” says Chris Jones, director of strategic technology development at iRobot. “You can imagine the day when a robot in the home can perceive and understand salient objects in the environment—that’s a couch, that’s my oven—that type of thing.”
The navigation approach used by the 980 is known as simultaneous location and mapping, or SLAM, which means it builds a map as it goes along and refers back to it. SLAM has been studied in academia for decades, but it has recently found its way into commercial research efforts and products. Traditionally, however, SLAM has required a significant amount of computing power. Getting SLAM to work in a small embedded computer system required significant effort to refine and optimize the algorithms used.
John Leonard, a professor at MIT who has done pioneering work on SLAM, says the technique is rapidly becoming a valuable commercial tool. Many of the automated cars currently being developed by carmakers as well as by Google and perhaps Apple use approaches based on SLAM for navigation. The technique is also used to help drones navigate, he notes. “It’s an exciting time for SLAM researchers,” Leonard says. “iRobot bringing a SLAM-based product to market is just further evidence of that.”
The Roomba 980 also performs what’s known as “sensor fusion,” meaning it combines data from its various proximity sensors with imagery from its camera. This means that it can still navigate if the camera is obscured by, say, a cat riding on top of the robot—a surprisingly common occurrence.
Besides the new sensing and mapping abilities, the new Roomba features an improved cleaning system and roughly double the battery life. It can also be controlled via a smartphone app for the first time. The 980 costs $899, which is $200 more than the current top-of-the-line model.
iRobot is conscious of the security and privacy concerns that the new camera and connectivity may introduce. A representative explains that the maps are not transmitted from Roomba, and they are deleted after the robot finishes cleaning a room.
iRobot was founded in 1990 by researchers from MIT’s Computer Science and Artificial Intelligence Lab. The company’s first product was a remote-controlled machine for the military capable of bomb disposal and other tasks. The company got into home robots with the Roomba in 2002. Several generations of vacuum cleaner have followed, as well as robots for mopping and washing floors and cleaning gutters and pools.
Other companies have introduced competitive products within the robot vacuuming space. Last year the British appliance-maker Dyson demonstrated an automated vacuuming device that also uses cameras to navigate, but it hasn’t been released yet. And this year a startup called Neato Robotics began selling a robot vacuum that performs laser mapping as it moves around.
As competition heats up, home robots could become much more capable. Jones even hints that we might eventually see a Roomba capable of picking up your socks. “Another area on our roadmap is manipulation,” he says. “For that you need dense 3-D maps and models of the space around a robot, which is similar to the mapping you’re seeing here.”
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
ChatGPT is going to change education, not destroy it
The narrative around cheating students doesn’t tell the whole story. Meet the teachers who think generative AI could actually make learning better.
Deep learning pioneer Geoffrey Hinton has quit Google
Hinton will be speaking at EmTech Digital on Wednesday.
The future of generative AI is niche, not generalized
ChatGPT has sparked speculation about artificial general intelligence. But the next real phase of AI will be in specific domains and contexts.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.