How Does Self Driving Cars Work?

Self-driving or autonomous cars are the future of automobiles. These cars would relieve humans from driving work, making it safer and easier. 94 percent of fatal accidents worldwide are caused by human errors. It means, eliminating humans from the driving process will reduce the number of accidents to a much lower level. Making mistakes is human nature, and those mistakes by drivers create misadventures like accidents, traffic jams, noise pollution etc. The autonomous cars have the potential of wiping out all these misfortunes and making this world a better place to live. The car companies are spending millions on the development of these technologies. The scopes for autonomous bushes, trucks and taxies are quite high in future. Till now, these vehicles have limited autonomous capabilities. It would be very interesting to see a fully autonomous vehicle cruising on the streets. 

How Does Self Driving Cars Work? 

Self-driving cars are capable of performing their driving function like steering, braking, acceleration by sensing the surrounding conditions. Apart from this, they can also identify obstacles and interpret road signs. The 4 major types of functions performed by self-driving cars are:

1. Sensing and Identification

2. Localization

3. Path Planning

4. Control

Sensing and identification is done by using an integrated system of Cameras,  Lidars, Radars and Ultra Sonic sensors. Localization is done by advanced algorithms using GPS and the surroundings data collected by the car. Once the car knows about its surroundings and its exact position, the car's computer plans the trajectory for its motion. This is called path planning. The step after path planning is 'control' i.e. acceleration, braking and steering. The control functions help to manoeuvre the car on the trajectories decided by the car's computer. All these functions are made possible by the advanced components of the car i.e.

1. LIDAR
2. Radar
3. Cameras
4. Ultrasonic Sensors
5. High-Speed Processors

Now, let's look at all those components one by one. 

LIDAR

You might have seen a box like structures over some of the autonomous cars. They are called LIDARs. The meaning of LIDAR is Light Detection And Ranging. The LIDARS use LASER for creating a 3-dimensional mapping of the car's surroundings. They emit laser which is reflected back from the objects in its path. The LIDAR absorbs the reflected light and forms images of the objects in the surroundings. It also calculates the time taken by the laser light to travel back after reflection. This time is used to calculate the distance of the objects from the car. In this way, LIDAR forms a 360 degree dynamic 3D picture of the car's surroundings. This information is processed by the car's processor and is used to control the car accordingly. The problem with LIDARS is their high cost. Using LIDAR on cars may increase their cost excessively. This will make them out of reach of the common people. The other problem with them is, they can not function properly in bad weather i.e fog or snow. 

Cameras

The self-driving cars use a set of cameras to sense the objects around the car. Tesla's Self Driving Cars use a set of 8 cameras to create a 360-degree vision. Ford uses 7 camera for the same. Let's look at the camera arrangements in Tesla Cars for a better understanding. 

1. Front-facing cameras 

Pic Credit: Tesla

Tesla cars have 3 front-facing cameras, to provide a clear view of obstacles in front of the car. These cameras lie behind the windshield and each serves a different purpose. 

• Narrow front camera:  This camera has a range of 250 meters. It provides a long-range vision in front of the car. This is useful for high-speed driving.

• Wide front camera:  This camera has a short range of 60m. But it has a field of view of 120 degrees. This camera can monitor the obstacle which tries to cross the path of the car like pedestrians or cyclists. They also read the street signs and traffic signals. 

• Main front camera: This camera has a range of 150 meters and a field of view of 50 degrees. They are intermediate of the main camera and the wide camera. 

2. Side front cameras

These cameras are located on both sides of the car. They are two in number and have a range of 80 meters. They sense the cars which try to enter your lane from sides. They also provide safety when entering crossroads with reduced visibility. 

3. Rearview side cameras

They are two in number and are located on both sides of the car. They provide a view of the car's blind spots. They are useful while changing the lanes or when other cars try to overtake yours. The range of these cameras is 100 meters. 

4. Rearview camera

This camera is located on the back of the car. The range of this camera is 50 meters. It is useful when parking the car or when reversing it.

The Radar

The cameras and The LIDAR can not always provide a clear view. Their capabilities are limited by fog, rain, snow etc. To overcome these problems, self-driving cars are fitted with radars. The radars can sense obstacles within a range of 150 to 200 meters, around the car. They use radio waves to sense the objects in the vicinity of the car. They have a transmitter and a receiver.  The transmitter emits the radio waves. These waves travel all around and are reflected by the obstacles in their path. The reflected waves are absorbed by the receiver. The receiver estimates the distance of the object using time and speed formula. One major drawback of RADARS is their incapability of sensing small objects like pedestrian or cyclists. 

Ultrasonic Sensors

Ultrasonic Sensor Range Around The Car
Pic Credit: Tesla

Ultrasonic sensors are being used extensively in self-driving cars. They are used to measure the distance of obstacles from the car. These sensors emit ultrasonic vibrations which are reflected by the obstacles in their path. The reflected vibrations are again absorbed by the sensors. They, then calculate the distance of that object using the speed of sound and the time taken by the vibration to travel back to the sensor. Tesla Autopilot uses 12 such ultrasonic sensors. These sensors help the car during lane shifts and traffic jams. They sense other obstacles in the vicinity of the car and help it to avoid any collision. Some advantages of using ultrasonic sensors are:

1. They have a higher range as compared to cameras. 
2. Their performance is not affected by fog mist or other weather conditions. 
3. They can even sense transparent objects. 

Processor

The processors are the key elements in a self-driving car. These cars use very advanced processors having huge computing power. They collect the data from all the sensors, process it and decide the control functions accordingly. The huge computing speed of these processors helps the car in quick decision making. These decisions are a result of different kind of processing algorithms. These algorithms are generally classified into two types.

• Low-Level Algorithms:  These algorithms deal with data processing i.e. image processing, LIDAR processing etc.  The data processed by them are further used by other high-level algorithms.                                

• High-Level Algorithms: High-level algorithms perform functions like;

                 -Decision Making
                 -Path Planning
                 -Feature Identification
                 -Object Classification
                 -Mapping and Localization
                 -Traffic Light Detection

This was a brief explanation of self-driving cars. There is still a lot more information about autopilot technology. We will try to cover them in our upcoming posts. Please share your feedback in the comment section.