The hidden AI tech that’s packed into robot vacuums

Just how do they avoid tumbling down the stairs?

Early robot vacuum cleaners were quite simple: they randomly bumped around a room, changing course whenever they encountered an object so that eventually their random pattern of vacuuming would cover the whole floor. Robot vacuum cleaners today are packed full of sensors and computers and are considerably cleverer.

LiDAR, also used by autonomous vehicles, is often used to detect walls and obstacles by bouncing invisible beams of light off them. Some robots use VSLAM (visual simultaneous localisation and mapping) – a way of processing images from a camera to understand the surroundings of the robot and its location.

The robots might also use ultrasonic ToF (time of flight) sensors, which work in a similar way to the echolocation of a bat, to avoid obstacles or detect if the floor is hard, soft or missing. They have accelerometers to identify motion and sensors to detect if their wheels are tangled up. Finally, suction sensors detect if their airways are blocked and the dust bucket is full.

All this information is used by AI software to determine the location of the robot and help it build a map of your rooms and furniture as it cleans. The robots are designed to clean every square inch of the floor and so they attempt to plan the most efficient route while remembering the way back to the charging station.

However, they cannot always see everything (and we may move things as they clean) so the coverage is not always perfect. Navigation becomes more difficult if their sensors become partially blocked – a robot under your sofa may be largely blind and can become quite lost!

The cleverest robot vacuum cleaners show you their maps of your home and allow you to tell them where they should concentrate and where they should avoid. The stairs are still beyond them… but Dyson filed a patent for a stair-climbing cleaning robot in 2020, so watch this space!

Source