What is the level of Huawei’s “car building”?

The following photo, taken in the 31st year of Guangxu, shows a carriage and a car running side by side on the road. The violent collision of the old and the new era is frozen in the black and white photo. People's imagination of "a faster horse" is beginning to be broken.

From the steam age to the electrical age, and now to the information age, every great leap of human civilization is accompanied by technologies and products that subvert cognition, and quietly become the wheel that pushes the times forward in incomprehension and ridicule.

▲More than 100 years ago, Dobbin Company stated in an advertisement that their carriage is better than Ford. Image from: Twitter

However, after smartphones and mobile Internet have changed our way of life, people have found that the speed of technological gears has stagnated.

As if American economist Taylor Cowan predicted in "The Great Stagnation", "People have already picked up all the low-hanging fruits."

Humans have landed on the moon in 1969, but people are still working hard to land on the moon again today;

In 1970, the first Boeing 747 passenger plane took 8 hours to complete the first voyage from New York to London. The flight time has not been shortened so far. The supersonic passenger plane cannot overcome the cost and noise problems. Musk's rocket intercity transportation system is still in the drawing stage;

We have forgotten how long we have been at the technology product launch conference and have not chanted "Amazing!" like when Jobs took out the iPhone. What is the Next Big Thing that inherits the smartphone?

Brain-computer interface, AR/VR, Metaverse…There are many options for the answer, and the electrification and intelligence of automobiles is undoubtedly one of the answers that most manufacturers are betting on.

The seed players of smart phones such as Apple, Huawei, and Xiaomi are all trying to pick the car, the jewel in the crown of modern industry, as if they are about to become a new generation of car-making forces.

Volkswagen, Ford, BMW and other traditional automakers have also bet on their lives and put new energy vehicles on the agenda to replace traditional fuel vehicles.

The excitement of the new energy vehicle industry reminds people of the era when feature phones were replaced by smartphones. Although the view of comparing traditional car companies to Nokia is not rigorous, it is impossible to say that they have no such anxiety at all.

▲ Picture from: Upstream News

Just today, Huawei’s flagship store has officially started selling cars. Perhaps in the “xx family bucket” we will call in the future, electric cars will be an indispensable consumer electronics product.

What is the level of Huawei's "car building"?

On the eve of the Shanghai Auto Show in April 2021, a Huawei Inside's Polar Fox Alpha S HI version of the autonomous driving road test video "detonated" the social network, causing a wave of waves with one stone.

The performance of the polar fox Alpha S HI version in the video is very close to the autonomous driving picture in our mind. And more critically, this car will be delivered in 2021, and it won't be long before we may see it on the road.

Judging from the road test video, Huawei ADS (Autonomous Driving Solution) does not require the driver to take over during the entire process. You only need to set the starting point and the end point on the vehicle to complete point-to-point autonomous driving. Among them, scenes such as traffic lights, left turns, stoppages, electric vehicles "ghost probes" and other scenes on the road are like old drivers, and they can cope with ease.

The Polar Fox Alpha S HI version can be regarded as an L4 (fully autonomous driving capability) autonomous driving, and it can still reach the end safely even on complex road conditions. In order to achieve the L4 level of autonomous driving, Polar Fox Alpha S HI version is equipped with a set of hybrid sensing solutions, including three Huawei 96-line lidars, located in front of the vehicle, which can achieve 300° viewing angle detection, and the detection distance can reach 150m~220m .

Editor's note: At present, my country's laws and regulations only support L2 vehicles on the road, and L3 and above vehicles are not allowed to drive on the road. Even if they are equipped with related technologies, they cannot truly operate in accordance with L3 and above autonomous driving.

In addition, in order to make up for the inability of lidar to obtain color information, the Polar Fox Alpha S HI version is also equipped with 13 cameras, 6 millimeter wave radars, and 12 ultrasonic radars around the body. This set of hybrid solutions can basically meet all-weather 360° coverage. And this is also the hardware foundation that Polar Fox Alpha S can accurately detect in the previous road test, so as to easily deal with it.

The lidar in Huawei’s ADS uses 12 large antenna arrays (12T24R) that emit 24 and receive. It is the largest antenna array imaging radar that can be mass-produced in the industry. The resolution capability is 4 times higher than that of traditional radars. Object distance, and perceive the recognition perception of non-line-of-sight occlusion targets, so as to achieve an L4 level of autonomous driving experience.

The Polar Fox Alpha S HI version is equipped with three Huawei 4D high-precision lidars. The data collection is far more than the traditional L2 level automatic driving vision solution. It also requires higher computing power to quickly process a large amount of data. This central supercomputing intelligent driving chip specifically designed by Huawei for ADS has a computing power of 400TOPS, which can meet the L4~L5 level of autonomous driving scenarios, and also exceeds the 144TOPS calculation of the Tesla FSD (Full Self-Driving) computing platform. More than twice the force.

It is not enough for future smart cars to have high-profile hardware configurations, and the ability to learn and evolve in the state of the Internet of Vehicles is more important. Huawei uses the "triathlon" of super hardware, super algorithms, and super data lakes. The super hardware is reflected in the hybrid sensor and central supercomputing chip on the Polar Fox Alpha S HI version, while super algorithms and super data The lake relies on the collaboration of vehicles and clouds. This is the responsibility of Huawei's open platform for autonomous driving, which Huawei calls "Octopus".

In addition, Huawei's high-precision maps are more in line with China's road conditions, and have a more comprehensive restoration of elevated and high-speed road conditions, as well as complex scenes in the city, such as plugging and flashing of battery cars. However, at present, this high-precision map cannot cover the whole country. According to the plan, it will cover the national highway and first-tier city road conditions in 2021, expand to 8 cities in 2022, and increase to 20 cities in 2023.

Reflected in the Polar Fox Alpha S HI version, the L4 level of autonomous driving may only support the four major cities of Beijing, Shanghai, Guangzhou and Shenzhen at the end of the year, and subsequent updates will unlock more cities.

From the polar fox Alpha S HI version of the road test video, it does bring us surprises, and the L4 level of autonomous driving seems to be in front of us, not the future. However, the route of the road test is relatively fixed, and it is easier to simulate and recognize, and then it can complete the "automatic driving" with a better experience.

As for whether the same intelligent driving can be achieved in cities with high-precision maps without restricted areas after the delivery at the end of the year, there may be a question mark here. In the past six months, Huawei still has time for ADS to iterate algorithms and optimize decisions. As for the final results and conclusions, it may still depend on the road test performance of the mass-produced Polar Fox Alpha S HI version at the end of the year.

However, from the perspective of mass-produced hardware alone, the computing power of the 4D lidar and MDC vehicle-mounted computing platform equipped with Huawei ADS (the Polar Fox Alpha S HI version is a specially optimized central supercomputing) has already gone beyond Tesla’s front.

In addition to BAIC, Huawei will also launch HI version models with Changan and Guangzhou Automobile. Huawei helps traditional car companies build smart cars. Some are similar to the Nexus phones launched by Google and hardware manufacturers. They are all "pro sons."

Not only "making cars", Huawei has also begun to select this ecological brand through Huawei's smart selection, and the automaker Cyrus has opened the way to sell cars . Users can book vehicles in Huawei Mall and Huawei Experience Store. Huawei will start on April 21. Smart Selection SF5 will also be available in Huawei flagship stores across the country.

▲ Cyrus Huawei Smart Selection SF5 is launched on Huawei sales channels. Picture from: Huawei

Huawei's consumer business CEO Yu Chengdong said that Huawei has more than 5,000 high-end experience stores, which is a huge advantage for Huawei in selling cars.

Huawei's entry into the field of smart cars like this does not have the opportunity of only one or a few rounds of cards like other "new car-making forces". What Huawei wants is to stay on the card table.

A brief history of Huawei's “car making''

Except for the frame, seat, wheels and other conventional auto parts of the Polar Fox Alpha S HI version, the rest are made by Huawei, including "smart cockpit", "autopilot system", "intelligent perception hybrid sensor" and so on. Really building a car is just too close.

However, Huawei still stated that it will not build cars on its own, but will continue to focus on ICT (Information and Communication) technology and only wants to be a Tier 1 supplier of smart cars.

To make an inappropriate analogy, Huawei wants to make an Android system in the field of smart cars. It does not involve traditional hardware, does not make enemies with traditional car companies, but "makes friends" with traditional car companies to help them build competitiveness. , Even together to create "explosive models" to make the HI (Huawei Inside) plate bigger.

At present, the stunning performance of the Polar Fox Alpha S HI version of the fully automated driving road test video is a phased achievement of Huawei's "car building", and it has more room for improvement in the future.

For this finished product, Huawei began to set foot in "car building" as early as 2009, initiated the research and development of on-board modules, and released the on-board module ME909T based on 4G network in 2013. It is regarded as officially entering the Internet of Vehicles, and its attitude towards car building is also It has not changed, but it is "focusing on the ICT field and focusing on cooperation."

In 2014, Huawei set up an Internet of Vehicles laboratory in the “2012 Lab” where the entire HiSilicon system was born, and then began to concentrate its R&D efforts to begin deployment in the field of smart cars. Subsequently, Huawei started to cooperate with Dongfeng, Changan, and FAW in the field of Internet of Vehicles and smart cars.

In the following year, Huawei brought together eight manufacturers, including Ericsson, Intel, Qualcomm, and Audi, to form a 5G automotive communication technology alliance. Starting from the communication technology he is good at, integrate AI chips, computing platforms, cloud services and other service modules to provide a complete set of services for smart car data.

Not only has it made good friends with manufacturers on communication services, but also has made friends with traditional auto manufacturers in the field of smart cars for autonomous driving.

▲ Dongfeng Sharing-VAN 1.0 Plus L4 unmanned vehicle. Picture from: Dongfeng official website

In October 2014, Dongfeng and Huawei started to develop unmanned vehicles. In 2017, BAIC and Huawei launched a strategic cooperation to develop new technologies. In 2018, Dongfeng and Huawei, BAIC and Huawei began to cooperate more deeply, focusing on intelligent networked vehicles. In 2019, Changan and Huawei launched a strategic cooperation to create an "Innovation Center".

"Non-stop" can be used to describe Huawei's state of deploying smart electric vehicles in 2014 when it began to focus on research and development. Not only staying in research and development, but also beginning to cooperate with traditional car companies to "open the door and make friends with each other."

With the deepening of role playing in the industry, Huawei's automotive business has become more and more important internally. In June 2019, Huawei established the Intelligent Vehicle Solutions Business Unit (BU) to upgrade the automotive business that has been deployed for many years to a first-level department. The business direction is designated as smart driving, smart cockpit, smart electric, smart connected, and smart car cloud. Five major sectors.

After completing the layout of these five major sectors, at the Mate 40 series conference in November 2020, Huawei BU finally launched the smart car solution brand HI (Huawei Intelligent Automotive Solution). In the future, electric vehicles equipped with Huawei's full-stack smart car solutions will be marked with the HI logo, just like products equipped with Intel chips in the PC market have the Intel Inside logo.

Afterwards, Polar Fox Alpha S with Huawei HI certification appeared in front of us. Previously, consumers did not have much perception about the layout of Huawei's automotive business. In addition, Huawei's low-key "car building" was not known to many people. In fact, the strength of Huawei's automotive business should not be underestimated.

Judging from the road test results, the Polar Fox Alpha S HI version has an L4 level of automatic driving capability, which can complete point-to-point driving in a fixed area, just like a "driver" is hidden in the car. Huawei tried to enter the automotive field from producing in-vehicle modules, and then began to build "drivers" in five major sectors. It actually took less than ten years.

In the past ten years, except for the outer shell and the four wheels, Huawei has done almost all the parts involved in smart cars. The launch of the Polar Fox Alpha S HI version is not a small milestone. However, there are many challenges to face. How Huawei, which is dedicated to helping automakers build cars, meets the needs of different manufacturers, and how to break through the closed supply chain of traditional cars, these challenges are even as difficult as making their own cars.

Therefore, if these problems cannot be solved well in the next few years, Huawei may not be far from building its own cars.

Lidar “price butcher'' Huawei vs “visualist'' head Tesla

For new car manufacturers and suppliers of autonomous driving technology, the comparison with Tesla is always inevitable.

Huawei’s rotating chairman Xu Zhijun said in a recent interview that Huawei’s autonomous driving technology is no worse than Tesla. Su Qing, president of Huawei's intelligent driving, bluntly stated that Huawei's automatic driving is already in the first echelon in China.

The reason for this performance is largely due to the 3 lidars on the Polar Fox Alpha S. Although there are 1,200 algorithm engineers among the more than 2,000 R&D personnel in Huawei's autonomous driving business, Tesla may not have an advantage if it wants to fight on the algorithm.

At present, there are two main technical routes for autonomous driving, one is the lidar group represented by Weilai, Xiaopeng and Huawei, and the other is the vision group that Tesla respects. They are like the Jianzong and Qizong of the Huashan School. They belong to the same sect but each show their own branches.

Regardless of whether it is the autonomous driving business or the visual school, what they have to solve is the "perception" aspect of autonomous driving. You can think of them as the eyes of a car to replace the human eye to identify routes and obstacles, but the brain with 10 billion neurons is connected to the human eye. It is not easy for the machine to perceive the environment in the same way. .

In cars that support L2/L3 autonomous driving, millimeter wave radar + camera is the most common solution. The camera recognizes the object, and the millimeter-wave radar is responsible for distance measurement, speed measurement and azimuth measurement, which together constitute the vehicle's perception ability.

The advantage of this solution is lower cost, but it has many drawbacks. For example, it is difficult to identify stationary objects. Li Xiang, the founder of Ideal Auto, once compared the combination of a camera and millimeter wave radar to the eyes of a frog, which may judge objects such as street lights as obstacles and take braking actions.

In addition, optical cameras can only obtain 2D images, and 3D environment modeling can only be understood by algorithms. This puts a great test on the algorithms. A little calculation error may cause serious accidents.

As a "visionist", Tesla is able to lead the industry in L2 autonomous driving because its FSD (Full Self-Driving) software has an excellent algorithm for converting 2D images into 3D materials, which requires a lot of driving data. Train the neural network.

But even if Tesla's algorithm is excellent, it will be "unacceptable" in the face of China's complex road conditions. This is related to the fact that most of the algorithm training data comes from the United States.

And the algorithm faces some special scenarios that have not been learned, that is, "corner cases", which are often powerless. Even if the training data is rich, there will always be unknown extreme situations in reality. This has always been in the field of autonomous driving. problem.

Is there a simpler method that can achieve the same or even better results without complex algorithms and massive amounts of data? Yes, that is lidar.

Lidar can directly 3D model the details of the environment around the vehicle by emitting and receiving a large number of light beams instantaneously, and translate the world seen by the human eye into a machine understandable appearance and update it in real time.

Since there is no need to convert from 2D to 3D, Lidar can achieve more accurate environmental recognition under the same conditions. Since lidar is so good, why has there been almost no mass-produced model equipped with lidar before? The high price is the biggest obstacle to the commercial use of lidar.

In the early days, Waymo used the 64-line lidar provided by Velodyne, a sensor manufacturer (the more the number of lines, the higher the accuracy, and the more expensive it is). The single cost exceeds 70,000 US dollars, which is even more expensive than the price of a car.

So manufacturers such as Tesla chose a lower-cost millimeter-wave radar + camera, and used algorithms to make up for the lack of hardware configuration. Musk has opposed the lidar program on many occasions. Last year he even stated on the "Tesla Autonomy Day" (Tesla Autonomy Day) :

Using lidar is stupid… Anyone who relies on lidar is doomed to fail.

But if the mass production cost of lidar is reduced to a level that can be commercialized on a large scale, this statement may be questionable, and all this is happening.

Huawei released a 96-line mid- and long-range lidar product last year , and is advancing its production capacity according to an annual output of 100,000 sets/line. The current price is roughly in the hundreds of dollars. In addition to BAIC, high-end smart cars built by Changan Automobile, Huawei and CATL will also be equipped with this lidar.

Wang Jun, president of Huawei's Smart Car Solutions BU, revealed that Huawei plans to reduce the cost of lidar to US$200 or even US$100 in the future, which is a price butcher.

With the sharp drop in the cost of lidar, many car companies will launch mass-produced lidar vehicles this year. In addition to the Xiaopeng P5 released not long ago, car companies such as Weilai, Mercedes-Benz, Honda, etc. have all announced new models equipped with lidar.

The French consulting agency Yole Développement estimates that global lidar shipments will be about 340,000 in 2020, about 4.7 million lidar will be shipped in 2025, and about 23.9 million lidar will be shipped in 2030.

Although most car companies have decisively embraced lidar, the two solutions of lidar and vision are not all-or-nothing. On the contrary, many domestic car companies have chosen the "laser radar + camera + millimeter wave radar + high-precision map and other sensors" fusion solution. This is the case with the Polar Fox Alpha S HI version.

In addition, Huawei's autonomous driving solution also has a technology similar to Tesla's "shadow mode". The so-called "shadow mode" is a separate system independent of the main control system of the vehicle. It can obtain the same data and output driving instructions as the main control system, but these instructions will not be executed on the vehicle. It is used for training while ensuring safety. algorithm.

It is worth mentioning that some foreign netizens have photographed some Tesla models equipped with Luminar lidar for testing on the road. If one day Tesla releases a model equipped with lidar, it is not too surprising.

Cars will become super mobile terminals, what we need is not the next Tesla

Since automobile manufacturing involves almost all types of industries known to mankind, it is also called "the jewel in the crown of modern industry." This is much more complicated than consumer electronic products such as mobile phones, but more and more technology manufacturers are involved in the automotive industry, and even build their own cars, indicating that the barriers to entry are decreasing.

In an interview with Caijing Weekly , Wen Kai, senior director of strategy at Juwan Technology, a subsidiary of GAC Group , said that the degree of modularization of electric vehicles is higher than that of fuel vehicles. The high degree of modularity will reduce the difficulty for outsiders to enter. The advantages of software iteration will also be further highlighted.

In addition to the huge potential market for electric vehicles themselves, under the trend of the Internet of Everything, smart cars will also become an important mobile terminal. This is probably a more important reason for the influx of technology companies to "build cars."

As early as 2015, Apple's chief operating officer Jeff Williams once said that the car is the "ultimate mobile device." Coincidentally, Huawei has a similar attitude in the matter of "building cars." Su Qing, President of Huawei Smart Driving, said this :

The view of traditional car companies is to take the car as the basis and try to embed the computer. We believe that the computer is the basis and the car is a peripheral controlled by the computer. Traditional car companies will make a lot of small boxes, adding a function to a box, but we are a big computer, and we hang up the car. This is fundamentally different.

Last year, Wang Chenglu, president of Huawei's consumer business software department, said that "a car is a four-wheeled mobile phone" actually means the same thing. Huawei cares more about how to improve its IoT ecosystem through "building a car".

With the popularization of 5G and electric vehicles, 5G is very likely to have real killer applications in the field of autonomous driving and Internet of Vehicles, and automobiles will naturally become an important entry point for the Internet of Things.

The launch of HarmonyOS is obviously to adapt to the IoT era in the era of the rich Internet of Everything. Based on the distributed architecture, the HarmonyOS system can achieve seamless collaboration across terminals and all scenarios. The experience of mobile phones, computers, refrigerators, cars and other terminals are all Can get through.

Liu Xuewen, editor-in-chief of Aifaner, pointed out in the article " Huawei has finally become a super terminal company " that the super terminal company does not depend on the number of terminal types, but on the control of core components and operating systems, and the control of the ecosystem. Power and influence, as well as a forward-looking exploration of human-computer interaction.

Huawei and other manufacturers are vying to enter the "car-making" market, not because new energy vehicles will become the next smart phone, and Tesla may no longer be the most coveted template for these new car-making forces.

Just as people didn't need a faster carriage more than 100 years ago, do we just need a greener and smarter car today?

To create a seamless integration of all terminals, to integrate different hardware devices to form a super terminal, to create a new human-computer interaction model, may be the lever to leverage the changes of the times.

This article was co-authored by Du Yuanli and Li Chaofan

#Welcome to follow Aifaner's official WeChat account: Aifaner (WeChat ID: ifanr), more exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo