How do Self-Driving Cars Work?

Self-driving Cars are autonomous car’s that runs on their own without the need of a driver. The car has inbuilt sensors, actuators, cameras, and a bunch of hardware and programs controlling the vehicle. Today in this post we will simply look at the steps to get an overview of how self-driving cars work?

Self-driving cars have five core components:

1. Computer vision

2. Sensor fusion

3. Localization

4. Path planning.

5. Control.

We look at all these things separately-

1. Computer Vision:

Humans use their eyes and brain to visually sense the world around them. We identify different objects around us. Computer Vision is the science that provides the machines or computers, the ability to sense the world around them.

Self-driving cars have lots of cameras installed on them. They see the world around them using those camera images like lane lines or other vehicles or peoples on the road.

2. Sensor fusion:

Now, that we (Self-driving cars) know the world around us. We should also know how far the people or other vehicles are on the road from us. We should also know the speed at which different objects are moving in the environment. For this, we use sensor fusion to Integrate data from sensors like radars and lasers together with the camera images or data to build a comprehensive understanding of the world around us. We also use temperature and humidity sensors to detect the weather. It is a useful measure to take in a bad weather condition.

Now, we(self-driving car’s) have the whole picture of the dynamic world around us.

3. Localization:

After a rich understanding of the world around us, we localize our-self in the world. You must be thinking that this is quite an easy task with all your cell-phones having GPS. So it might seem like we already know where we are in the world. But in fact, GPS is only accurate to within about 1-2 meters. 1-2 meters error is quite big as if the car gets wrong, it could be off on the sideways hitting things. So, we use much more sophisticated mathematical algorithms that help the vehicle to localize itself with-in 1-2 centimeters.

4. Path Planning:

Once the world is clear, we know where we are in the world. We should plan our path to reach the destination. For this, we chart a trajectory through the world to get where we want to go. First, we predict what the other vehicles around us will do. Then we decide which maneuver we want to take in response to those vehicles. Finally, we built a trajectory, or plan to execute the maneuver safely and comfortably.

5. Control:

 The data collected by sensors are sent to the actuators, to set the vehicle in motion. We also have planned our trajectory, now it’s the time to follow it. The vehicle needs to take turns, hit the throttle and brakes. Therefore, we need a control system to be on the exact trajectory planned.

Here we track our efficiency by comparing the two trajectories. First, the ideal or planned trajectory and second trajectory is what we actually follow while moving. Many vehicles do better on straight roads i.e. both trajectories overlap. Just on turns, there is a slight difference between the two trajectories. If you drive, you know this happens to us too. And machines too are doing a good job.

Many companies like Google, Tesla, Uber, Nissan are working hard to develop a fully automated car by using the above components. I hope this information helps. Finally, thank you.