Home > News content

Huawei's auto driving power surpasses Tesla's? The advantages and disadvantages are obvious

via:博客园     time:2021/4/19 21:52:53     readed:103


writing / Li Xin

Source: lianxianchuxing

On April 17, BAIC's new energy vehicle brand Jihu officially released the pure electric car alpha s, which is equipped with Huawei hi basic model and supports automatic driving on high-speed roads, while the high-level model supports automatic driving on urban roads.

In October last year, Huawei officially released Hi, the smart car solution brand, namely Huawei inside. Huawei hopes to cooperate with auto companies as a first-class supplier (Tier1) to build smart cars.

This time, the polar fox alpha s equipped with Huawei hi version is not only the first time Huawei's automatic driving technology has been implemented, but also the first time Huawei has demonstrated its L4 level automatic driving ability.

In order to realize automatic driving on urban roads, Huawei has also made heavy efforts. The autopilot scheme cooperated with Jihu alpha s includes 33 intelligent components, including intelligent cockpit, intelligent driving, intelligent Internet connection, intelligent electric, intelligent car cloud service, as well as lidar, millimeter wave radar and ultrasonic radar.

This set of intelligent driving solutions is naturally expensive.

That is to say, the HUAWEI HI version of the extreme fox Alfa S is 137 thousand yuan higher than the ordinary version, while the Tesla City automatic driving system FSD sells for 64 thousand yuan at home, which means that HUAWEI's automatic driving solution is much more expensive than Tesla.


Jihu alpha s Huawei hi parameter table, photo source: Jihu official website

The reason is that Huawei's autopilot solution is more expensive than Tesla's at this stage, because both sides adopt different technology paths in sensor solutions.

Tesla uses the path of vision algorithm, mainly with the help of camera, millimeter wave radar and other low-cost components, to model the surrounding objects, and add the data to the neural network for pure vision calculation. The advantage of this path is that the cost of camera is low, but it has limitations in accuracy, stability and visual field.

Huawei uses lidar path, with camera, millimeter wave radar and other components, to scatter laser around, judge whether there are obstacles around and generate point cloud image based on feedback, which has the advantage of strong anti-jamming ability, but the disadvantage is high cost.

At the end of last year, Huawei launched the 96 line high performance lidar for the first time at the summit hosted by the China Automobile Industry Association, and announced that it would reduce the cost to $200 or even $100.

Although Huawei's autopilot experience has reached L4 level, it still belongs to L2 level according to the law, which means that the responsibility subject is still the driver. In this case, how many consumers will spend 100000 yuan more for automatic driving technology, which may be a question mark.

In any case, as a high-level automatic driving scheme with mass production, Huawei hi's strength can not be underestimated. In the future, with the mass production and listing of intelligent cars that Huawei cooperates with other auto companies, automatic driving technology may speed up its entry into cities.


Just entering a lane full of cars on both sides, the car on the opposite side has entered the intersection in seven turns and eight turns, and it happened that the takeout boy on the other side was riding an electric car and suddenly ran out horizontally.

In the exposed Huawei automatic driving technology evaluation video, the driver has turned on the automatic driving mode in the whole process, and even in the case of complex road conditions, the system can handle it by itself. This evaluation video also shows Huawei hi's urban road automatic driving ability almost perfectly.


Jihu alpha, Huawei hi version automatic driving evaluation, tuyuan 42 garage

According to Huawei, in the urban commuting scene, the polar fox alpha s equipped with Huawei hi has been able to realize the point-to-point traffic of the whole scenic spot covering the urban area, expressway and parking lot.

Specifically, Huawei's automatic driving solution can intelligently identify traffic lights and judge whether to stop or continue driving; When the vehicle is at the intersection without waiting area, it can interact with the opposite straight and left turn vehicles and complete the left turn behavior.

Compared with high-speed roads, the complexity of automatic driving on urban roads is terrible. Moreover, the domestic road conditions are complex, and motor vehicles and non motor vehicles intersect with each other, which requires vehicles to have high precision recognition ability and strong computing ability to control vehicles to make accurate actions in real time.

It can be seen that it is not easy to realize the automatic driving between cities, and the automatic driving function of most current intelligent cars can not recognize the traffic lights, and Tesla can only realize the following parking after the front car stops, and can not recognize the traffic lights.


Huawei autopilot demonstration, tuyuan Huawei hi new product launch

According to Huawei's plan, after the mass production of intelligent vehicles of cooperative auto companies at the end of this year, Beijing, Shanghai, Guangzhou and Shenzhen will first provide high-precision maps, and then provide a series of new cities every three months, covering more than 8 cities by 2022 and more than 20 cities by 2023.

In addition, Huawei has also invested five years in the field of autopilot algorithm. The central computing unit of its extreme fox alpha s Huawei hi version is a domain controller developed by Huawei ads (full stack self-developed vertical integration solution) department, which can support 400tops / 800tops computing power.

At the same time, at the new product launch of Huawei smart car solution 2021 held on April 18, Liao Zhenqin, general manager of Bu smart car cloud service product department of Huawei smart car solution, also said that Huawei high-precision map supports all mainstream high-precision map standards. At the same time, it will open a multi map business alliance to jointly promote data update of high-precision map with partners.

Public information shows that there are more than 2000 people in Huawei's automatic driving team. Huawei said that it will continue to increase investment in auto parts industry. In 2021, the investment in R & D is expected to reach 1 billion US dollars, and the investment in R & D in automatic driving will remain 30 billion US dollars in the future % About the growth rate.


For a long time, Huawei has repeatedly made it clear that it is not going to build a car. Instead, as a supplier of incremental parts for smart cars, Huawei has cooperated with auto companies to launch smart cars.

At present, Huawei cooperates with BAIC and other automobile manufacturers in the mode that the whole vehicle factory is responsible for manufacturing, power and other links, while Huawei takes advantage of ICT (information and communication technology) to provide intelligent parts such as intelligent cockpit and automatic driving.


Wang Xing's comments

According to wired insight, in the smart car products jointly launched by Huawei and auto companies, the cars that use Huawei's automatic driving scheme will bear the logo of Huawei inside.

It is worth noting that the new products released by Huawei have been in mass production, that is, they can directly provide a full range of solutions for automobile enterprises.

In terms of intelligent cockpit solutions, Huawei focuses on computing platform, Hongmeng OS operating system and core services. The computing platform is based on Kirin chip, and the display platform includes car smart screen and ar-hud.

It can be said that in the future, Huawei's core technology capabilities will not only be reflected in the field of mobile phones, but also more in the field of automobiles.

As a platform for automobile design, Hongmeng OS supports the capabilities of one core, multi screen, multi concurrent and multi component collaboration.


Huawei harmonic OS vehicle and machine operating system, tuyuan Huawei hi new product launch

Specifically, after listening to music and watching navigation on the mobile phone, users can directly synchronize to the in car system to realize seamless flow.

At present, Huawei has also reached cooperation with more than 50 application partners such as station B, Gaode map, Tencent smart travel, iqiyi, etc.

In this new product launch, Huawei also released intelligent driving computing platform, which can provide different levels of automatic driving solutions. The intelligent driving computing platform mdc810 is an intelligent driving computing platform with the largest computing power and mass production in the industry. The dense computing power reaches 400tops, and it can meet the needs of congestion following ( TJP )、 High speed cruise ( HWP )、 Automatic parking ( AVP ) And so on.

It is reported that mdc810 intelligent driving computing platform is equipped with 16 cameras, 12 can buses and 8 vehicle Ethernet terminals. The platform has completed all the testing and mass production.

In addition, Huawei also released a mass-produced 4D imaging radar, which uses 12t24r large antenna array (12 transmit channels, 24 receive channels), which is 24 times higher than the antenna configuration of conventional millimeter wave radar 3t4r, and 50% more receive channels than the industry's typical imaging radar. This is also the largest mass-produced array imaging radar at present.

4D imaging radar inherits the excellent range and velocity measurement ability of traditional millimeter wave radar, which is not affected by light, rain and snow, and has the unique detection ability of non line of sight occluded targets.


However, if Huawei wants to really gain a foothold in the automotive field, it needs to solve not only technical problems, but also practical problems such as market acceptance and openness of laws and regulations.


From the hi new products launched by Huawei, it provides a full range of intelligent vehicle solutions, and all of them have the leading level in the industry. However, this is also based on high R & D and high cost, and Huawei is ready for long-term investment and non profit.

Even if Huawei wants to reduce the cost of lidar to $200 or even $100, judging from the current price of Huawei hi version of Jihu alpha s, the price of this lidar led solution is still high.


Huawei lidar, tuyuan Huawei

The reason is that the cost of lidar is too high, especially the implementation of multi line lidar technology is complex. Generally speaking, the higher the laser radar harness is, the higher the accuracy will be, the more perfect the surface profile of the object will be, and the cost will increase.

Take velodyne, an overseas lidar company, as an example, its 16 line, 32 line and 64 line lidars are sold for as much as US $4000, US $40000 and US $80000 respectively. By contrast, the hardware cost of a camera is only a few hundred dollars.

To this end, musk insists on using the vision fusion scheme. In the Tesla hardware system autopilot hw2.0, it is equipped with eight cameras to provide 360 degree look around function, including the front three eye camera, which are long-distance narrow angle, medium distance medium angle and short-distance fisheye. There are also two cameras facing the front and back of the side on the left and right sides, plus one rear camera.

However, although the camera is cheaper for sensors, the system has higher requirements for computing power. In the process of driving a car, the data volume of each camera will be built at MB (1024KB) level. Eight cameras will work simultaneously, as can be imagined that the number of data will increase.


Comparison of automatic driving environment perception technology routes, tuyuan Deloitte China

Huawei's lidar based solution is on the contrary. Although the hardware cost of lidar is high, the data it collects is standardized data packets.

Public information shows that, taking the vlp-16 line lidar produced by velodyne as an example, the data length of each frame is fixed at 1248 bytes, that is, about 1KB, while the output of this lidar is 480 frames per second. The total data per second is only 500kb, half less than that of the pure vision scheme.

It can be seen that if Huawei wants more users to buy intelligent cars equipped with Huawei's automatic driving solution, it must reduce the price of lidar, otherwise it is difficult for users to accept it.

In addition to the cost problem, Huawei also needs to face the reality of laws and regulations, which is also a common problem faced by all auto companies with automatic driving.


Huawei intelligent driving computing platform, tuyuan Huawei hi new product launch

This also means that the main experience of consumers is still L2 level assisted driving, and most consumers may not spend more than 100000 yuan now, just for the function of automatic driving, which is rarely used in daily life.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments