No longer rely on maps, Huawei’s smart driving experience is better.

Although intelligent driving with graph and intelligent driving without graph belong to intelligent driving technology, they adopt completely different methods. Traditionally, map-based intelligent driving relies on high-precision maps to provide centimeter-level road information to assist vehicles in planning their route. It’s like having a vehicle follow a precise path drawn on a map. However, the cost of this method is very high.

You may not know that making high-precision maps is a complex and arduous task. The daily results of surveying and mapping work usually do not exceed 100 kilometers, and the cost per kilometer even reaches several thousand yuan. And with the rapid infrastructure construction in China, the newly mapped road may be unrecognizable in a week. In addition, map update not only limits the display of intelligent driving ability of vehicles, but also means that vehicles must wait for the release of the new map.

Intelligent driving without map gives up the dependence on high-precision maps and realizes the arbitrary driving of vehicles throughout the country. This requires vehicles to have more keen perception, more advanced algorithms and more powerful computing power, which improves the standard of intelligent driving system for vehicles.

How does Huawei lead the way in intelligent driving without maps?

　　

1. Clear vision and foresight-lidar becomes the gold standard.

　　

At present, the intelligent driving technology without map can be roughly divided into two camps: lidar faction and pure vision faction. Among them, lidar has become the first choice for advanced intelligent driving because of its high performance of multi-threading, high-frequency scanning and long-distance detection.

　　

In the catastrophic traffic accidents and online celebrity dodge videos, lidar showed its outstanding ability. Under the condition of dark or backlight, the powerful perception ability of lidar shows its advantages, and at the same time, it puts forward higher requirements for computing power and algorithm. Therefore, the smart driving model equipped with laser radar not only costs more, but also can not be underestimated in the level of technology development.

　　

In other aspects, the intelligent driving scheme of visual system, led by DJI and Baidu, mainly relies on 2D video content, which requires much lower computational power and algorithm than that of lidar. However, in the real scene test, the visual system scheme did not show particularly outstanding performance, reflecting that there is still a gap in actual use. The limitation of camera affected by weather and illumination is also a problem that can not be ignored.

　　

Huawei uses multi-sensor fusion sensing scheme in system hardware, including lidar, millimeter wave radar, ultrasonic radar and visual sensing camera, which provides wider and more accurate three-dimensional environmental information. Huawei’s self-developed fusion sensing system deeply integrates various sensors to ensure the unimpeded data communication.

　　

2. See the spirit in detail, GOD network gets rid of the white list.

　　

In order to control the vehicle accurately, the intelligent driving system needs to identify and understand the external environment of the vehicle in detail. In the traditional technology, BEV+ whitelist method identifies the detected objects by matching with the preset whitelist in the system. However, this method has obvious defects, because it needs to constantly learn new obstacle types to update the white list. If you encounter rare obstacles outside the white list, the system may not be able to judge correctly, leading to hesitation or even wrong driving behavior.

　　

Huawei’s GOD2.0 network effectively meets this challenge. The network regards any detected object as a three-dimensional cube. This enables the system to easily identify obstacles on the driving route, even those unusual obstacles, such as scattered cartons, fallen trees, construction roadblocks, etc., can be effectively identified by the system. Therefore, without the restriction of white list, the system is no longer plagued by obstacles that are "unclear" or "incomprehensible".

　　

3. Being proficient in the identification of objects and roads, RCR technology has become a turning point.

　　

With GOD2.0 technology, Huawei Intelligent Driving can identify "things". It is also important for the system to recognize the "road". Intelligent driving without map requires accurate perception of the road, especially for complex intersections, country roads or sections lacking obvious signs. Huawei Intelligent Driving uses ordinary navigation maps combined with RCR network reasoning to create a broad road topology map, which enables the intelligent driving system to accurately interpret the real road conditions without relying on high-precision maps.

　　

4. Data processing is fast and efficient, and MDC chip enables L5 intelligent driving.

　　

After having the ability to perceive and understand the environment, the next step of intelligent driving system is to merge, analyze and calculate this information, and finally control the vehicle intelligently. The huge amount of data generated by sensors requires extremely fast processing speed, which brings great challenges to computing power.

　　

Huawei’s MDC series chips have excellent computing power, surpassing NVIDIA’s Orin, and can efficiently process the data of various sensors. At the same time, Huawei’s intelligent driving system is not simply piling up hardware, and the computing power of MDC hardware platform is enough to support the automatic driving assistance function from L3 to L5.

　　

5. The global road test takes only five days, and Huawei BU Computing Center is the only one in the world.

　　

After getting rid of the limitation of high-precision maps, accumulating rich driving experience becomes the key. The data collection of early intelligent driving system depends on the test of real vehicles on the road. The modern practice is to conduct simulation tests in the cloud through AI modeling, and Huawei once again leads the industry in this respect.

　　

Huawei Cloud Computing Center has the industry-leading 2.8EFLOPS computing power, which supports the rapid development of intelligent driving technology. Using the scene library built by Pangu model, the system completes more than 12 million kilometers of simulation test every day. Compared with the total road length of about 64 million kilometers in the world, Huawei’s system can achieve global coverage in only 5.3 days. A large number of simulation tests make the intelligent driving system have the same driving experience as an old driver.

　　

Huawei’s intelligent driving system not only allows users to enjoy a worry-free travel experience, but also realizes the leap-forward development of technology behind it. With its full-stack self-developed hardware, massive data resources, continuous pursuit of innovative technologies and forward-looking technical concepts, Huawei has launched a nationwide intelligent driving service without maps. For consumers, we no longer need to wait for the update of high-precision maps, nor do we need to worry about the degradation of intelligent driving systems caused by various situations.

　　

Huawei’s smart driving solution makes new energy vehicles truly practical and safe smart vehicles. In the field of intelligent driving, Huawei has demonstrated cutting-edge technology and innovation ability, which not only enhances the travel experience of users, but also sets a new milestone for the whole industry.