Basic Knowledge Basics of autonomous driving - Part 3
The third part of the series "The Basics of Autonomous Driving," focuses on the essential technologies for autonomous vehicles. First, we take a look at the sensor technologies.
If autonomous vehicles genuinely are intended to help us in our everyday lives - especially concerning reducing the number of traffic accidents - they must have a close eye on the traffic situation at all times. To do this, they need to have complete knowledge of their surroundings - a 360° view of all the factors influencing their driving patterns.
This goal can only be achieved by integrating a large number of sensor mechanisms. Some of these are already used in today's driver assistance systems, while others are still in development and are designed specifically for autonomous driving. Without these technologies, it is hardly possible to imagine vehicle autonomy taking off in the fast lane, let alone becoming suitable for the masses. The different stages of autonomy - from cars in which humans have complete control to the utterly autonomous vehicle - were presented in the second article of the series. Now the focus is on the fundamental sensor technologies.
The sensor mechanisms
Three key technologies give autonomous vehicles "vision": LiDAR, cameras, and radar. Currently, each of these technologies is at a different stage of development.
The least complicated is probably radar, which is already part of the vehicle equipment and supports individual systems such as adaptive cruise control. It has a significant role in the development of autonomous vehicles, especially in low-speed scenarios such as parking or traffic jams. But the technology also performs effectively at higher speeds, for example, when changing lanes on motorways.
The latest mmWave radar systems for vehicle applications use the microwave spectrum to determine the range, speed, and relative angle of detected objects. They typically operate in the 77 GHz frequency band and can detect even the smallest movements.
Radar has many advantages: It is a secure technology that works reliably regardless of fluctuating environmental conditions. The required hardware is compact and comparatively inexpensive, as it is already widely established on the market. However, it also has some inherent limitations, especially concerning the amount of data it can generate. Autonomous vehicles must, therefore, rely on a large number of different sensors and not just on a single, isolated mechanism.
LiDAR is a technology that is used in almost all development programs of car manufacturers. As a supplement to vehicle radar systems, it is of vital importance. LiDAR uses pulsed light waves emitted by a laser source and reflected by objects in the surrounding area. Based on the time it takes a pulse to return to the sensor at the starting point, the distance covered by the signal can be calculated. The process is repeated millions of times per second to create a 3D map of the environment in real-time. This map provides information about the shape and dimensions of vehicles, the traffic infrastructure as well as cyclists and pedestrians, making it easier to identify and locate obstacles. A key advantage of LiDAR is that the technology offers a "bird's eye view" compared to other sensor options. Ford, for example, has already invested heavily in the technology. They are using Velodyne's HDL-64E LiDAR technology in the development and testing of their autonomous vehicles. The first model in which the technology is implemented is planned for 2021.
The HDL-64E is a 64-channel system with a horizontal field of view of 360° and a vertical field of view of 26.9°, with a range of up to 120 meters. The number of supported channels is crucial for possible vehicle speed. According to Velodyne, a car equipped with a 32-channel system could only drive autonomously up to a speed of 57 km/h, but by doubling the number of channels, much higher speeds can be achieved.
One of the biggest hurdles for LiDAR (and one reason why visionaries like Elon Musk consider the technology unnecessary) is the high cost involved. Current systems are priced in the mid-five-digit range. Even with falling unit prices, their use would still be costly. And the costs are not the only problem. Although LiDAR can map the environment of a vehicle, it does not work in detail needed to read, for example, road signs.
Autonomous vehicles additionally require high-resolution camera systems for image recognition and classification tasks.
By equipping a vehicle with front, side and rear cameras, a 360° real-time view of the environment can be generated. This makes it possible to eliminate blind spots, detect speed limits in time, and reliably maintain the lane. The number of cameras required depends on the system's field of view (which can be up to 120°) and whether "fish-eye" cameras are used (with a super wide-angle lens for panoramic views).
As with any sensor technology, its advantages must be weighed against its limitations. Although camera systems provide a high-resolution image of the environment, they have problems measuring depth and distance, and the distance between objects would have to be calculated to determine the exact position of a detected object. Cameras also find it more difficult to identify objects in poor lighting conditions, such as fog or at night.
System developers are increasingly focusing on the influence of sensor systems on other autonomous vehicles. There has been a particularly intense debate in the recent past about whether LiDAR could interfere with the operation of digital cameras - this would pose a severe and potentially life-threatening problem when autonomous vehicles collide.
In conclusion, we can state the following: It is not yet clear exactly which combination of sensors will be used in future autonomous vehicles. It seems clear that it will be a mixture of different mechanisms (such as radar, image sensors, and LiDAR) so that their respective advantages can be bundled and the respective disadvantages of the systems compensated. The combination of different technologies will result in a wide range of functionality and redundancy, which are urgently needed to ensure that self-propelled cars in road traffic do not pose a potential danger but rather optimize our mobility.
This article was first published in German by next-mobility.news