How does the robot vacuum know how to navigate and dodge objects so skillfully?

Compact robot vacuum cleaners are becoming more and more popular in households, gradually replacing bulky vacuum cleaners. These robots move around the floor, vacuuming up dirt and then returning to the socket to charge. How are these robots able to move and clean so intelligently? It turns out that inside each small robot vacuum cleaner, there are many types of sensors along with pre-programmed preeminent features, which help them work automatically or follow the control of the user.

The robot vacuum cleaners do not use the camera system to observe the world around them. Instead, they contain a variety of sensors that detect and measure objects, thereby avoiding obstacles. The basic sensors in a robot vacuum include: cliff sensors, bump sensors, wall sensors and optical encoders. Depending on the version, the robot vacuum also has additional sensors, such as a vacuum cleaner, to measure the amount of dust being sucked.

Sensors in robot vacuum cleaners help them move smart

'Cliff sensors' are sensors that measure the distance between the robot base and the floor, by reflecting infrared light from the floor. If this distance increases suddenly, it means the robot is approaching the edge of the stairs or something like that, so it will back away to avoid crashing (hence the name cliff sensors- "cliff sensors". ").

The name 'bump sensors' also reflects the use of this type of sensor: if the robot vacuum cleaner hits something (like a wall or the foot of a chair), the impact will activate the sensor. Wall sensors are similar to cliff sensors, they tell the robot when it is near a wall, and then adjust itself to follow the wall.

Optical encoders (optical encoders) are considered the most important part, they are located on the wheel of the robot, helping to calculate the distance traveled. Called optical encoders, they use light sensors to detect how many times the wheel has turned, thereby figuring out how far the robot has traveled.

Thanks to a clever combination of sensors, the robot knows how far it has traveled, what it has hit and which areas are unsafe.

So what allows smart robot vacuum cleaners to distinguish between cleaned and un-cleaned places? The answer may surprise many: 'Thanks to ants!'

The path of the iRobot Roomba 980 robot vacuum is marked with purple light

Indeed, most modern robot vacuums were born based on the research work of Rodney Brooks, a roboticist from the Massachusetts Institute of Technology (USA) and an entomologist. He is an active member of the new artificial intelligence (AI) research movement that aims to break away from complex problems to focus on the basics of intelligence. This movement explains that ants are inherently not intelligent, but they still have complex behaviors and precise ways of handling situations.

How? By following a simple set of rules. Specifically, an individual ant does not have a high level of intelligence, but it does have a simple set of rules that allow it to find food, return to the nest, and guide the rest of the colony. Likewise, robot vacuums don't need to know the exact size of a room to clean. Instead, it just needs to know how to react in a few different situations to be able to clean up the whole room. Robotologists call these rules 'behaviour' - outlined in iRobot patent filings from 2002.

The idea of ​​​​programming a robot vacuum cleaner is based on the intelligence of ants

These behaviors are extremely simple: the "go straight" behavior tells the robot to keep going straight until it hits something. The "turn around" behavior tells it that when it hits something, it has to stop, turn in another direction, and move straight back. The "spiral" behavior indicates it moves outward in a spiral, cleaning the floor in extended circles. The behavior "follows the wall" indicates it follows the wall by "backward" and "straight".

When you apply them to the robot vacuum, these rules allow it to navigate the room. The patent outlines a typical sequence of behavior:

1. When the user places it on top of a dirty spot, the robot will begin to follow a "spiral" until it hits something or has covered a sizable area.
2. If it encounters a collision, it will switch to a "follow the wall" behavior for a short time.
3. After a certain distance, the robot enters the "straight" state at a random angle. According to the patent, iRobot calculates this distance as (3/4 × mean distance between "collisions") + (1/4 × most recent distance between "collisions").
4. The vacuum cleaner goes "straight" until it collides with something, it backs up and then goes straight again.

"I don't have the pinnacle of wisdom, but I'm smart enough to be what you need"

A similar set of behaviors tells the robot what to do if it gets stuck. It will first try to free itself in a number of ways (slowly moving, rotating, backward, etc.), before giving up and starting to beep miserably to be rescued.

We often think of robot vacuums as intelligent machines that use a laser to map a room, a complex computer to map the way. But really, like ants, robot vacuums are just simple machines, obeying a few pre-programmed instincts.