How a robot decides what to do at a closed door

Bump sensors notice when something is touched

A program alone would leave Pip blind. To find out what's happening around it, a robot uses sensors, which are parts that notice one thing about the world, like a touch, a sound, or how bright it is. The fact a sensor hands to the robot is called an input, which simply means information coming in.

The simplest sensor is a bump sensor. Pip's front bumper is a curved strip of plastic on a tiny spring. Behind it sits a little switch, like the button on a doorbell. When the bumper hits something, it slides back a few millimeters and presses the switch. The switch sends one tiny message to Pip's brain: "I touched something."

You may have seen this in action at home. A robot vacuum cleaning the kitchen rolls under the table until its bumper taps a chair leg. Tock. The switch clicks, the vacuum stops, turns a little, and tries a new direction. It might tap the same chair three times before it finds a way around, but it keeps going without ever getting stuck pushing against the leg.

A bump sensor only tells the robot about something after it has already touched it. That works for a soft chair leg. It isn't so good when the thing ahead is a person's foot or a glass vase, and that's why many robots also carry a sensor that can feel ahead without touching.

Distance sensors bounce sound and light to see ahead

A distance sensor tells a robot how much empty space is in front of it. Pip's two round "eyes" are really one of these, an ultrasonic sensor. One eye sends out a quick squeak of sound that is too high for people to hear, and the other eye listens for the echo.

The trick works like shouting in a big empty gym. When the wall is far away, your echo takes a moment to come back. When the wall is close, it comes back almost right away. Pip's sensor times the echo, and that tells Pip how far away the wall is. Bats find bugs in the dark in much the same way.

Some robots use light instead of sound. An infrared sensor shines an invisible beam of light and measures how much of it bounces back. A close object reflects a lot of light, and a far one reflects only a little.

Plenty of toy robots use this. Hold your hand in front of one, and as your palm gets closer, the sensor sees less and less empty space. When the space gets too small, the toy rolls backward, as if it's shy. It never touched you, because it "saw" your hand coming.

This is how Pip stopped a few centimeters from the closed door instead of crashing into it. Its sound sensor noticed the space shrinking and gave the input "something is very close."

Light sensors follow paths on the floor

Some sensors point down instead of forward. A light sensor measures how bright the surface under it is. A white floor bounces back lots of light, and a thick black line bounces back very little.

A line-following car uses this to drive along a road drawn on paper. Picture a sheet of white poster board with a curvy black marker line on it, and a little car with a light sensor peeking down at the front. While the sensor sees dark, the car is on the line and keeps rolling. When the line curves and the sensor suddenly sees bright white, the car knows it has drifted off and turns until it finds the dark line again.

It drives in tiny wiggles, left, then right, then left again, but it follows the line all the way around. Robots in some big warehouses steer the same way, following stripes or markers on the floor to bring boxes to the right shelf.

Pip has a light sensor too. Under its belly, it watches for the blue tape that marks the path to the art room. Pip also has a small microphone, which is a sound sensor, so it can hear a sound like a knock. Each sensor notices just one kind of thing, so a robot that needs to know a lot carries several.

If this happens, then do that

Sensors hand the robot facts, and the program turns those facts into actions. The part of the program that chooses is a decision, a moment where the robot picks between two actions depending on what is true. Programmers write decisions as if-then rules.

An if-then rule works like a rule you follow on the way to school. If the crossing light is red, then stop. Otherwise, walk across. You check one fact and the fact picks your action.

Delivery robots that roll along sidewalks in some cities use rules just like this. When one of them reaches a street crossing, its program checks the signal and the cars around it. If the way isn't clear, then it waits at the curb. Otherwise, it rolls across. Nobody pushes a button for it each time, because the rule was written ahead of time and it is ready whenever the robot reaches a crossing.

Pip's program has rules too. Here are three of them.

• If the bumper is pressed, then back up a little.
• If something is closer than 10 centimeters, then stop. Otherwise, keep rolling.
• If the floor under me is blue, then keep following the tape.

Try predicting Pip's next move with these rules. Pip is rolling down the hallway and a cat walks right in front of it, very close. Pip stops, because the distance rule says so. Now the cat walks away and the space ahead is open again. Pip rolls forward, because the "otherwise" part of the rule is now the one that is true.

Robots check their rules over and over, many times every second. That's how they can react so quickly when something changes.

Putting it all together to clear the hallway

Now we can watch Pip at the closed door, one moment at a time. Every robot repeats the same three-part loop, called sense, think, act. It senses the world, thinks by checking its rules, and then acts. Then it starts the loop over again.

Sense. Pip's sound sensor sends its squeak and the echo comes back right away. The input is "something is 3 centimeters ahead."

Think. Pip checks its rules. Something closer than 10 centimeters means stop. A second rule written for doors says that if the path stays blocked, Pip should beep twice and wait.

Act. Pip stops and plays beep beep.

Before reading on, guess what happens next. A girl in the art room hears the beeps and opens the door.

The loop starts again right away. Pip senses, and this time the echo takes much longer to come back, so the way is open. Pip thinks, and the "otherwise, keep rolling" rule is now true. Pip acts and rolls through the doorway. Its light sensor finds the blue tape on the other side, so it follows the tape to the table and sets down the crayon box, which is step 6 of its program.

Pip never had a single clever idea of its own. It had a program with steps in the right order, sensors that noticed the door and the tape, and if-then rules that turned each new fact into the right move. Every robot you meet, from the vacuum under your table to a delivery robot at the curb, runs on those same three things.