🚗 Self-Driving Cars

3rd Grade 🤖 Artificial Intelligence

A self-driving car (also called an autonomous car) is a car that can drive itself, without a person holding the steering wheel. It uses cameras, sensors and artificial intelligence to “see” the road, figure out what’s happening around it and decide when to go, turn and stop.

Free · US Letter · no sign-up
Google Classroom Save to Pinterest Email me

How does a car “see”?

A human driver has eyes and ears. A self-driving car has sensors. Cameras film the road in every direction, and AI recognizes traffic lights, road signs, lane lines, cars, bikes and people walking in the pictures.

Radar sends out radio waves and measures when they bounce back, so it knows how far away things are and how fast they’re moving — even in rain and fog. Many self-driving cars also have lidar, which sends out invisible laser beams and uses them to build a 3D map of the surroundings. On top of that, there are detailed maps and a navigation system that knows where the car is.

Why so many sensors? Because each one has a weakness: a camera struggles in the dark or in glare, and radar has a hard time making out small details. When you combine several sensors, each one covers for the others’ weaknesses.

From seeing to deciding

Once the car knows what’s around it, it has to predict what will happen: Is the kid on the sidewalk about to cross? Will the car in front slow down? Then it decides what to do — slow down, go around, stop or keep going.

You can think of it as three steps that repeat all the time: see, predict and decide. You do this too when you cross the street — you look left and right, judge whether the closest car could stop in time, and only then cross. The difference is that the car does it with calculations, without really understanding what a kid or a ball is.

It does all this many times every second. The AI learned from huge amounts of recorded drives, and on top of that there are clear rules the engineers set, like stopping at a red light and letting people in the crosswalk go first.

Levels of self-driving

Not every “smart” car is self-driving. Self-driving is usually divided into six levels, from 0 to 5. At level 0, the person does everything. At the lower levels, the car only helps — for example, keeping a safe distance from the car ahead, warning you when you drift out of your lane or braking in an emergency — and the driver must pay attention the whole time.

At the higher levels, the car drives by itself, at least in certain areas and conditions. At level 5, the highest, the car is supposed to drive itself anywhere and in any conditions a person could drive in, with no need for a driver. On most roads in the world, people are still driving, and cars without drivers operate only in limited areas and conditions, as part of tests and certain services.

Why is it so hard?

A road is full of surprises: a ball rolling into the street with a kid running after it, a police officer directing traffic with hand signals, road work, a sign covered in mud, or an animal suddenly jumping out. A human driver uses experience and common sense. A self-driving car has to handle situations it has almost never seen in its training data — and that’s the big challenge.

Bad weather — snow, heavy rain or fog — can also interfere with the sensors. That’s why self-driving cars go through an enormous amount of testing, in computer simulations and on real roads, before they’re allowed to drive around people.

Promises and questions

Most car crashes are caused by human mistakes: being tired, getting distracted by a phone, speeding. A self-driving car doesn’t get tired or look at a phone, so there’s hope that someday it will help reduce crashes. It could also help people who can’t drive, like some older people or people with disabilities.

But there are hard questions too: Who is responsible if a self-driving car causes a crash — the company that made it, the programmers or the owner? How do we make sure the system is safe? And what will happen to people whose job is driving? People need to decide these questions together.

Key facts

  • A self-driving car uses cameras, radar and sometimes lidar to see its surroundings.
  • Combining several sensors covers for the weaknesses of each one.
  • Self-driving is usually divided into six levels, from 0 to 5.
  • Rare, surprising situations on the road are the biggest challenge.
  • Most crashes are caused by human mistakes, so there’s hope self-driving will reduce crashes.

Quick quiz

  1. What is a self-driving car?

  2. What does radar do in a car?

  3. Why do self-driving cars use several kinds of sensors?

  4. How many levels of self-driving are usually counted?

  5. What is the biggest challenge for a self-driving car?

  6. What should the car do after it spots a kid on the sidewalk?

  7. Why is there hope that self-driving cars will reduce crashes?

  8. Which question are people still working on about self-driving cars?

💭 Think and talk about it

  1. Think of something surprising you once saw on the road. How would a self-driving car handle it?
  2. If a self-driving car caused a crash, who do you think should be responsible? Explain why.
  3. How could cars without drivers change life for your grandparents, or for people with disabilities?

Questions kids ask

Are there already cars that drive completely by themselves?

There are cars that drive with no driver in certain areas and conditions, as part of tests and early services. But on most roads in the world, people are still driving, and many “smart” cars only help a driver who must pay attention.

What is lidar?

Lidar is a sensor that sends laser beams in every direction and measures when they bounce back. That’s how it builds an accurate 3D map of what’s around the car.

Can a self-driving car make mistakes?

Yes. Like any AI system, it can be wrong, especially in new and rare situations. That’s why it’s tested so much, and in a regular car the driver is responsible and must watch the road.

🔎 Ask Thinkapedia another question