Self-driving is not as distant future as we thought. By 2025, cars on autopilot will no longer be something out of the ordinary on city streets, and in 2030 their mass production is planned. But we still have little idea of how an unmanned vehicle works. In this article, we will answer this question in detail.
What can an self-driving car do?
He can do a lot of things that are inaccessible to classic cars.
Firstly, he moves completely independently from point A to point B, and chooses the best route for this, taking into account not only the map data, but also information from the Internet about traffic jams.
Secondly, it independently adjusts the speed, slows down when cornering, and accelerates on straight sections of the path. It also finds a free parking space and parks on its own.
Thirdly, the unmanned car recognizes other vehicles, clearly “sees” through fog, snow, and rain notices road signs and traffic lights.
While the range of functions can be considered limited because the developers plan to improve the system in such a way that the unmanned vehicle can react with lightning speed to changes on the freeways and thereby avoid accidents.
WHAT LEVELS OF AUTONOMY ARE AVAILABLE?
There are 6 levels of vehicle automation, from 0 to 5. Level zero means a car that is fully driven by a driver, level 5 means 100% unmanned vehicle.
How an self driving cars working?
You can see how the drone works on the example of the Toyota Prius, which was tested by Google engineers and programmers.
Continuous scanning of the terrain using sensors: lidars (laser radars), cameras, radars, and high-precision maps are mandatory conditions for autonomous vehicle movement.
The self-driving car system interacts with the Street View service, which gives a panoramic view of the city streets from a height of 2.5 m.
BASIC SYSTEMS THAT PROVIDE AUTONOMOUS MOVEMENT:
Lidar is the heart of autopilot. This is a laser rangefinder that is installed on the roof of a car and generates a 3D map of the space within a radius of up to 100 meters. The control computer combines the obtained data with Google maps, which allows it to avoid accidents and comply with traffic rules.
Radar – there are 4 of them on an unmanned vehicle (sometimes more): two in front and two on the rear bumper. This system uses radio waves to determine the range, trajectory and speed of objects. The radar emits pulses that are reflected from obstacles and transmitted to the receiving antenna. Thus, the radars become the “eyes” of the car and allow you to instantly react to any changes in the situation.
The position sensor is a special device that determines the coordinates of the car on the map. GPS receiver allows you to track the location of the vehicle and the route it will follow.
Camcorder – located near the rear-view mirror. It detects color signals from traffic lights, objects that are approaching a potentially dangerous distance. Modern drones usually have 1 to 3 video cameras.
ALGORITHM OF UNMANNED CAR OPERATION
1) With the help of lidar, a volumetric map of the area is generated, and the control computer connects it with the data contained in the memory.
2) Based on the information received from radars, cameras and sensors, a special algorithm evaluates the situation on the road and takes into account the behavior of other road users.
3) The computer determines the trajectory of the drone, and also reacts to the situation on the road: the movement of other cars, the gestures of the policeman, the school bus in front, pedestrians, ice on the highway and many other factors.
GOOGLE INNOVATION: LEARNING CONTINUOUSLY
Automated cars learn very quickly due to the fact that all the information and practical experience gained is transferred to the Google database and can be used by all cars.
The database contains a huge number of scenarios that are encountered in real life: an uncontrolled wheelchair on the road, a pedestrian suddenly jumped out onto the roadway, etc.
But there are also non-standard situations. For example, when testing a Google drone on the road, a girl in a wheelchair was chasing a bird. Naturally, there was no such scenario in the database, but the computer slowed down any way.
And not because there was a bird on the road – otherwise the car would have to brake at the sight of every pigeon. In order for the drone to respond correctly to such unusual situations, engineers have to constantly improve the control system.
“HUMANIZATION” OF UNMANNED AUTO WORK
According to most experts, self-driving cars behave too well on the roads. For example, the first cars stopped on the road simply “seeing” a person – the computer immediately decided that a pedestrian was going to cross the road.
But the person could just stop tying their shoelaces or wait for a friend. Therefore, the engineers decided that it would be more logical to slow down, and not stop completely, especially since a sharp braking creates an emergency situation on the track.
But Google developers went even further and gave the unmanned vehicle a “voice” – the ability to honk. The signal is triggered automatically when an increased danger arises both for road users and for the automated machine itself.