cool hit counter What is the lifeblood of autopilot and do you really understand it? --up_Intefrankly

What is the lifeblood of autopilot and do you really understand it? --up

In a flash, half of 2018 has passed, and less than 2 years remain before the major OEMs claim to achieve L4 level autonomous driving in 2020, although the industry holds different attitudes towards this time node, but what should come sooner or later, the result is always clear. or so Before we put our lives in the hands of AI, do you have a deep understanding of these current "lifebloods" of autonomous driving?

One of the lifebloods: complete vehicle solutions with new energy features

If you take a closer look at the unmanned vehicle projects that are currently being tested intensively by various companies you will see that most of them are purely electric or hybrid vehicles, why? First of all, we know that the core of the current mainstream autonomous driving technology is a set of artificial intelligence with deep learning capabilities embedded system Using the new Audi A8L, which is already in mass production and capable of L3 level autonomous driving, as an example, the zFAS controller it uses is a highly customized system that integrates Nvidia Drive PX 2 and Mobileye EyeQ3 embedded solutions. This system is considered the brain of the entire car, responsible for both sensor data fusion and scene analysis, as well as the task of decision making and vehicle control, and therefore Ensuring the stability of this system is a top priority.

From the perspective of traditional fuel cars, the engine is the only energy supplier for the whole vehicle, once the engine strikes, then the blow to the control center of the self-driving vehicle is extremely fatal, after all, for the core components of power consumption of hundreds of kilowatts at every turn, the traditional battery in the long run can not perform the task of reliable power supply. Unlike new energy vehicles, which are designed on the basis of a high-capacity battery system, the reliability of their electrical supply is much higher than that of fuel vehicles, and they can also more consistently support the proper functioning of a large number of sensor components.

The 48V and 90V light-hybrid systems that are just starting to gain popularity are actually the equivalent of adding an insurance policy to the proper functioning of electronic systems in conventional internal combustion engine vehicles (with ample emergency power reserves) and is perfectly capable of serving as a vehicle for the transition of autonomous driving technologies like L2-L3.

On the other hand. Since the controller has to take over the work of the ECU to complete the control of the vehicle components at the same time, the complex control logic of the internal combustion engine can also become a liability compared to the electric motor, which is not only a waste of computing resources, but also a drain on internal resources such as CAN bus bandwidth, which is not pleasant to see when designing self-driving cars. And all of this is based on the existing L3 level of autonomous driving alone; as the level ramps up to L4 or even L5, with increased vehicle operating speeds and the need for high-precision recognition, a general-purpose like the Nvidia Drive PX 2 embedded system The demand for computing power will follow exponentially, and with it, a further increase in power consumption. To sum up, we can basically understand why self-driving cars are more associated with new energy vehicles, both are the trend, and can complement each other, with a mature pure electric vehicle products as a carrier, self-driving technology can better take off.

Lifeline 2: Laser sensor

With the carrier problem solved, we'll have to look to the more specific hardware needs of autonomous driving technology , which includes many categories, such as the previously mentioned controllers, CAN buses, sensors, and the relatively already popular global positioning systems (GPS) and inertial measurement units (IMUs). Let's pick up the more specific ones here, and put the controller on the back burner and talk about the sensors first.

There are many classifications of sensors, the more mature ones used in automobiles are of course the sonar-based radar sensors, and the optical image-based vision sensors. And for autonomous driving technology, which requires greater precision, its current lifeblood lies in laser sensors. Why do you say that? We know that vision sensors can detect objects well as well as distinguish colors, but are limited by lighting conditions; radar sensors can sense dense or large objects from great distances, but often with less accuracy. The conclusion is therefore that both have limitations in practical application.

However, combining them with LIDAR's six-dimensional (three axes + three axes of rotation) high-precision detection capabilities would theoretically allow everything to be observed from a 3D view, day or night, and ultimately give self-driving cars computer vision that is far superior to human eyesight.With the exception of Tesla, led by Elon Musk, every self-driving car company in the world relies on LIDAR technology for their projects.

However as we have inherently known the word laser for years, it is a high cost gadget and a set of laser sensors usually consists of one or more laser units, ranging from a single line to the current maximum of 128 lines, with the cost and recognition accuracy increasing exponentially as the number of units rises. Again using the A8L as an example, this L3 level autonomous driving capable car uses a 4-wire laser sensor, which is already expensive, and if it were to move forward like L4 level autonomous driving, then the number of laser sensor threads would inevitably rise further and the cost would soar.

The good news is that, based on the increasingly competitive market in the industry, the development of laser sensors has started to move towards cost reduction and capacity expansion, and the cost is in the process of gradually decreasing. For example, earlier this year, Velodyne LiDAR, the global LiDAR leader, announced that the price of its 16-line LiDAR product, VLP-16 Puck, has been reduced to $3,999 worldwide from the previous $7,999, which is undoubtedly a major positive news for the development and promotion of autonomous driving technology.

Well, that's all we'll talk about today, and in the next installment I'll discuss the other two lifebloods of autonomous driving technology -Artificial intelligence as the master control core of autonomous driving embedded system, and forL4 harmonyL5 High-precision electronic maps are particularly important for level of autonomous driving # Study hard # # Day Day Up #

1、Use Perls FilePath module to delete nonempty folders
2、How to learn more in open source projects
3、Ceph User Mailing List Vol45Issue3
4、Taobao June Top 10 Categories Sales Data Full Version
5、Python download swf of prettyloaded

    已推荐到看一看 和朋友分享想法
    最多200字,当前共 发送