What kind of car does Huawei make?

What is the supplier of intelligent networked automotive incremental parts? What exactly does it provide?

"Building a good car" and "building a good car" are the concepts put forward by Xu Zhijun, the rotating chairman of Huawei, in his speech at the World Intelligent Networked Automobile Conference jointly sponsored by the Ministry of Industry and Information Technology and the Beijing Municipal Government on 22 October, 65438. The next day, he specially led Wang Jun, President of Huawei Smart Car Solution BU, and He Liyang, Vice President, to meet with reporters to further elaborate these two concepts.

At first glance, I feel a little mysterious. What is the difference? However, if we understand Huawei's understanding of the coming future car, we will understand that he actually said two meanings. On the one hand, it clarifies Huawei's understanding of intelligent networked vehicles, on the other hand, it also shows Huawei's outstanding ability in this respect.

First, "building a good car" means that the connotation of a car has changed. The automobile industry and ICT (information and communication technology) industry are merging, and the future car is intelligent networked electric vehicle. In the future cars, ICT will become more and more important, and the mechanical composition has not changed much. In other words, to "build" a car, you can't just build a mechanical part. Such a car is incomplete and does not meet the needs of the future or even the present.

20 18 10 months ago, Huawei consumer BG had a very strong desire to build a car. 20 18, 10 This month, Huawei finally made clear its strategic choice: Huawei did not build cars, focused on its dominant ICT technology, and determined its position as an intelligent networked automotive incremental parts supplier. The picture shows Huawei's R&D base in Songshan Lake, Dongguan.

Xu Zhijun believes that the intelligent networked electric vehicle will become a new revolutionary development engine of human society, and its influence will go far beyond the automobile and ICT industry itself, bringing great changes to human society, automobile industry and people's travel.

Intelligent networked electric vehicles will become a platform for continuous value creation. After the traditional mobile phone moves towards smart phone, smart phone has become a platform for the whole industry to continuously create value. Once a car moves from a traditional car to an intelligent networked car, it will also become a platform for the car to continuously create value for the industry and continuously improve the experience for consumers throughout its life cycle.

Intelligent networked electric vehicles will change the whole society from owning cars to combining owning and traveling services in the future.

It will also make the entertainment system in the car change the decades-old tradition. Now the booming intelligent terminal industry chain and ecology can bring the hardware ecology and application ecology of mobile phones into the car through the intelligent cockpit system.

Traditional cars are designed based on EE architecture (that is, bus+distributed control architecture), and intelligent networked electric vehicles will adopt distributed network+domain controller architecture, which is what Huawei calls CC (Computing+Communication) architecture.

Because intelligent networking makes cars no longer isolated mechanical objects, at this time, the concept of car safety has moved from traditional safety to trustworthiness. The security problems faced by such cars are more important network security and information security issues. ?

Second, building a "good car" is actually a demonstration of Huawei's strength. Huawei can not only help car companies "build" cars, but also build "good cars". At the Shanghai Auto Show in April 20 19, Huawei's auto business made its first public appearance. At that time, Xu Zhijun mainly talked about Huawei's mission to help car companies "make" cars. In this forum, his specific interpretation is how Huawei builds a "good car".

Previously, Huawei has been exploring the integration trend of ICT industry and automobile industry in the past four to five years. In the process of the integration of the automobile industry and ICT, and in the process of self-driving and electrification to help the automobile industry, how to give full play to its technological accumulation advantages in the ICT industry to promote the development of the traditional automobile industry to the intelligent networked electric vehicle once sprouted the idea of building its own car.

Xu Zhijun said that before 20 18 10, Huawei consumer BG had a very strong desire to build his own car. "They say Apple is making cars. Why can't we make cars? " 20 18, 10 In June, Huawei finally made clear its strategic choice: Huawei did not build cars, focused on its own leading ICT technology, determined its position as an intelligent networked automotive incremental parts supplier, provided parts and solutions for the majority of car companies, and helped car companies "build" cars and "good cars".

Huawei believes that after years of development, the quality of China's automobile industry has basically improved, and its manufacturing capacity has also improved. There is no shortage of automobile manufacturers in China and even in the world. What is lacking is an enterprise that can continuously provide technology and parts for the future "four modernizations". Most of these parts and technologies happen to be the technologies that Huawei already has, which is in line with Huawei's genes, and this eventually prompted Huawei to make up its mind in the automobile business.

2065438+On May 29th, 2009, Ren issued an organizational change document, agreeing that Huawei should set up a smart car solution BU, which will be managed by the ICT Management Committee. At the same time, Wang Jun, the former president of Huawei's wireless network business department and Japanese operator business department, was appointed as the president of BU, a smart car solution.

Two days later, Huawei's smart car solution BU was formally established. In Huawei's organizational structure, BU and BG are parallel first-level departments. At present, Huawei has three BG departments and two BU departments, namely operator BG, enterprise BG and consumer BG. BU and today's smart car solution BU.

Then, as an incremental component supplier of intelligent networked electric vehicles, how will Huawei provide products and technical solutions for car companies? This problem was not clear at the Shanghai Auto Show in April, but now Xu Zhijun or Huawei has been able to spread it clearly.

Xu Zhijun said: "The traditional mechanical part is not Huawei's advantage. You asked me to make the chassis, but I can't make it. You asked me to make an engine, but I can't make it. We are very clear that Huawei is doing ICT technology, adapting to the needs of intelligent networked cars and matching what it can do and what it can do. "

He believes that in the future, after fully autonomous driving and electric vehicles, 60%-70% of the value of a car is related to electronics, computing, communication and software, while traditional parts only account for 30%-40%. Tesla essentially builds its value on hardware and software related to electronics. Hardware can be replaced and software can be upgraded.

"This may be a bit alarmist, but people in our industry must have seen this trend." Xu Zhijun said, "In the future, the value of automobiles will be built more within the scope of ICT technology." Xu Zhijun believes that this is a challenge faced by traditional automobile enterprises. Of course, as integrators, the original brand value exists, which is their barrier against suppliers.

Huawei's threshold for competitors is "one architecture as the foundation, three platforms as the development focus, and connectivity and cloud services simultaneously".

One architecture is the CC architecture of "computing+communication" mentioned above; The three platforms refer to MDC intelligent driving platform, CDC intelligent cockpit platform and VDC intelligent electric platform; Connection refers to Huawei's intelligent network connection solution, which solves the problem of high-speed connection between in-vehicle and out-of-vehicle networks; Cloud service is a service based on cloud computing, such as online car owner service, entertainment, OTA and so on.

"Now Huawei's smart car solution BU is facing the solutions and products of intelligent networked vehicles. From traditional cars to electric vehicles, incremental parts for intelligent networked electric vehicles are basically involved. How big is this volume? " Xu Zhijun sighed.

However, he said that Huawei has no profit expectation for BU, a smart car solution. "Our requirement is to make the products well, competitive and competitive. The market scale will gradually come out, and naturally you can make money. The same is true for doing 5 G. We were in the product stage and didn't know which year we could make money. I really finished the product and found that the time to make money was earlier than originally expected. "

What makes Huawei so confident? Xu Zhijun said: "We are compared with Bosch. In fact, our approach with Bosch is different. "

The bus must be the platform?

The new CC architecture gives the traditional automobile industry a shortcut to intelligent networked electric vehicles.

"If you have a vision, you must build a platform, not a car." Xu Zhijun believes that the traditional electronic and electrical architecture must be transformed from the traditional automobile to the intelligent networked automobile.

Looking at Huawei's smart car solution, the core is to innovate the traditional EE (Electronic and Electrical) architecture and computing architecture, and use CC (Computing and Communication) architecture to make the car have the possibility of software definition and continue to create value, which may really make China's intelligent networked cars usher in a golden period of development.

The traditional EE architecture in automobile industry adopts the architecture of bus+distributed functional units. Its design idea is hardware definition specification, characterized by function-driven vertical design and dedicated resources; At the same time, special sensors, special ECU and special algorithm; There is also a long development cycle and cannot be expanded.

Of course, this architecture is feasible for traditional cars, because the number of ecus is small-1993, and Audi A8 uses five ecus. However, with the improvement of vehicle electronization, or the traditional vehicles begin to move towards the stage of intelligent networking, the number of ECU in vehicles has exploded, and the number of luxury cars has exceeded 100. At this time, the problem appeared.

Due to EE architecture, different ECU is provided by different suppliers, which leads to uncoordinated and redundant computing power. The embedded OS and applications of different ECU are different, which leads to the OTA can not be maintained and upgraded uniformly, and the third-party application developers can not easily program with hardware. The rapid increase in the number of ECU increases the demand for internal communication, which leads to the increase of wiring harness cost and assembly cost.

"The CC architecture is essentially different from the traditional EE architecture, and the traditional EE architecture cannot move towards intelligent networked cars." Xu Zhijun said.

For example, intelligent driving involves the process of sensor environment perception, high-precision map /GPS precise positioning, V2X information communication, multi-data fusion, decision-making and planning algorithm operation, electronic control and execution of operation results. Do you need a strong "brain" in this process? (moving? Data? Center, MDC (Mobile Data Center), needs huge computing power to unify real-time analysis, process massive data and perform complex logical operations. When it comes to autonomous driving, the information of perception, processing and decision-making is huge. It is necessary to ensure the accuracy of decision-making, and must issue instructions to directly take charge of the four wheels. It takes a lot of network bandwidth to transmit such a large amount of information to MDC. In communication language, it refers to 10Gbps interface, even 25Gbps and 100Gbps interface.

Xu Zhijun said: "For autonomous driving, CC architecture is inevitable. 90% of traffic accidents in the world are caused by drivers' misoperation. After really moving towards autonomous driving, MDC will replace people, more powerful than people, and human accidents will be basically eliminated. " ?

The design idea of CC architecture is software-defined function, including distributed Ethernet and three domain controllers. It can decouple resources and functions, share resource pool, upgrade software, replace hardware and expand sensors.

"In the future, we hope to put the electric capabilities of hundreds of ECU and intelligent driving network in the car into three domain controllers, and then connect them through Ethernet to connect all the sensors." ? Xu Zhijun said, "Why a distributed network? Because in the future, cars need not only an Ethernet switch, but also one in every corner. Millimeter wave radar, lidar and camera in every corner are directly connected here, which reduces the wiring harness. "

It is natural for Huawei to propose CC architecture. It is an Ethernet in itself. Now we only need to study how Ethernet switch technology can meet the needs of vehicle scenes. In this architecture, in addition to one domain controller to control intelligent driving, there are two domain controllers responsible for intelligent cockpit and vehicle control, and each domain controller has an operating system.

In the future, intelligent networked vehicles need 360-degree real-time perception to realize real automatic driving, and the data of sensors such as lidar, millimeter-wave radar and camera should be transmitted to MDC in real time and efficiently, which requires high-speed Ethernet. In terms of communication language, it needs 10Gbps interface, even 25Gbps and 100Gbps interface, which can't be realized by traditional EE architecture, and it needs to develop towards computing+.

Although many people only know that Tesla is an electric car, it is a new force that takes the lead in building cars in intelligent networked cars. Most of the automobile industry's understanding of intelligent networked cars comes from Tesla model? S and subsequent models. Before Tesla appeared, the car needed to go back to the 4S store to update the navigation software. When Tesla continues to provide better entertainment services for its stock cars, enhance Autopilot's assisted driving ability, and even improve vehicle performance through OTA (such as faster acceleration and longer cruising range), it provides users with unprecedented technological experience. At the same time, it also shocked the entire automobile industry.

Again? Models? 3? Take Tesla's latest generation of electronic and electrical architecture as an example. Its central computing module (CCM) integrates the intelligent driving field (ADAS) and the infotainment field (IVI), and adds the left body control module (BCM-LH) and the right body control module (BCM-RH), completely replacing the previous hundreds. ECU？ Complex architecture composed of controller and wiring harness.

In addition to Tesla, Volkswagen Group of Germany is also carrying out a large-scale electrification reform. By acquiring software suppliers, it has set up a large-scale independent software team internally to achieve the goal of greatly simplifying the electronic and electrical architecture and software-defined cars.

Xu Zhijun told auto business review that all car companies are now studying new structures, but with different names. Huawei is called CC architecture, because computing and communication are its specialties. "Different car companies have different names. I don't know what the name will be in the future. Maybe the CC architecture is called more, so it is called CC architecture. "

On the whole, however, at present, China automobile enterprises, including the new forces of making cars, still find it difficult to find a product like Tesla, which is an overall solution from the underlying architecture to the driving experience, entertainment experience and body control. "There are very few car companies like Tesla now. A large number of car companies are mechanical and lack the ability of electronic hardware and computer software. They want to have a partner. " Xu Zhijun said.

The brand-new CC architecture developed by Huawei has given Chinese and foreign traditional car companies, including China's new car-making forces, a shortcut to intelligent networked electric vehicles.

Based on CC architecture, Huawei builds three platforms-MDC intelligent driving platform, CDC intelligent cockpit platform and VDC intelligent electric platform through three domain controllers responsible for driving, cockpit and vehicle control, and builds each platform into an ecosystem by providing chips and operating systems.

So, when will the CC Ethernet architecture really land?

Xu Zhijun told auto business review: "At least the process of our cooperation with a large number of partners is no longer the traditional EE architecture. The car of the future is a platform for constantly creating value. This platform is combined with the industry, the hardware can be replaced, the software can be upgraded, and consumers pay for the corresponding experience upgrade during the life cycle of the car and continue to create value for the industrial chain. "

"Platform+Ecology" Intelligent Driving Strategy

There are already car manufacturers doing smart driving applications on the development board of Huawei MDC smart driving platform.

201810 June 15, in Huawei? Connect? At the 20 18 conference, Huawei released an important carrier to support its driverless strategy-autonomous Mobile data center (MDC, mobile? Data? Center) platform.

At present, Huawei has launched the MDC intelligent driving platform? , support L2+ to L4 level autopilot. AI computing power demand and intelligent operating system based on Huawei's core.

Shengteng 3 10 chip adopts efficient and flexible CISC instruction set developed by Huawei. Each AI core can complete 4096 MAC calculations in 1 cycle, which integrates various arithmetic units such as tensor, vector and scalar, and supports various mixed precision calculations and data precision operations in training and reasoning scenarios.

As an embedded neural network processor (NPU), Shengteng 3 10 chip integrates the advantages of FPGA and ASIC, including the low power consumption of ASIC and the high programmability and flexibility of FPGA, so its unified architecture can adapt to various scenarios, with power consumption ranging from tens of milliwatts to hundreds of watts, and flexible multi-core stacking can provide the optimal energy consumption ratio in various scenarios.

Xavier for NVIDIA? Computing power 30? TOPS(TOPS: trillions of times per second), power consumption of 30W, energy efficiency 1? Coat/coat. It has high computing power, but it consumes a lot of energy. Huawei Shengteng 3 10? The computing power is 16? Top matching, power consumption is only 8W, and energy efficiency is 2? A coat/jacket.

Mobileye？ EyeQ4, computing power 2.5? Top matching, power consumption of 3W, energy efficiency of 0.83? TOPS/W, it is said that the computing power of EyeQ5 is 0 times that of the previous generation1,but it is expected to be distributed in 202 1. In contrast, Huawei is currently on the rise? The advantage of 3 10 is also particularly obvious. ?

Based on this, whether it is the optimal computing power and power consumption of Shengteng 3 10 or its unified architecture can adapt to various scenarios, it is determined that it will inevitably become the core of Huawei's self-driving MDC intelligent driving platform.

It is generally believed that L2 needs computing power: at present, the computing power required by 100TOPS and L5 has not been clearly defined (at least 100tops is predicted), and the current computing platform can only meet the needs of some L3 and L4 levels of automatic driving.

Huawei can support L4 intelligent driving platform based on 8 Ascendant 3 10? AI chip, the computing power is as high as 352TOPS, and the power consumption computing power ratio of the whole system is 1? TOPS/W is the best in the industry.

At the same time, according to the market demand of practical application scenarios of autonomous driving, Huawei has developed MDC intelligent driving platform products with matching computing capability by increasing or decreasing the number of 3 10 chips and lidar.

In other words, Huawei MDC intelligent driving platform can support the smooth evolution and upgrade of L2+~L4 autopilot algorithms with a set of software architecture and different hardware configurations for different levels of autopilot algorithms.

Xu Zhijun said that Huawei's biggest advantage is its AI and cloud capabilities. Based on Shengteng chip and intelligent operating system, MDC intelligent driving platform is built. Huawei also opens API (application? Programming? Interface (application program interface), I hope to build three ecosystems with the vast number of component providers, integrators, application developers and other partners-sensor ecosystem, intelligent driving application ecosystem and executive component ecosystem, and finally push the whole automobile industry towards intelligent driving. Is this what Huawei said? The "platform+ecology" strategy makes smart driving enter the fast lane.

One is the sensor ecology, including lidar, millimeter wave radar, camera and so on, which makes these sensors easy to connect with MDC. Xu Zhijun said that the MDC intelligent driving platform must build an ecology, because it is the brain, and all the sensors must be connected and know each other.

Of course, as one of the options, Huawei will also use its own? 5G？ Technology research and development of millimeter-wave radar to achieve all-weather imaging, while making full use of the world's leading photoelectric technology to develop laser radar, truly solve the cost and performance problems faced by laser radar.

At present, the global millimeter wave radar field has been roughly formed? ABCD？ (Autoliv? Autoliv, Bosch? Bosch, China? Mainland, Delphi) control field, but even the highest frequency? 77G？ Millimeter wave radar, the resolution is still too low, not only can not accurately model pedestrians and obstacles, but also in sensor fusion and synchronization, AI? In algorithm processing, the original data of millimeter wave radar is not friendly enough. The high cost of lidar also makes many players miserable. If Huawei makes a breakthrough in these two aspects, it can be said that it is a major breakthrough in the field of intelligent driving sensors.

The second is the application ecology of intelligent driving. Huawei's MDC intelligent driving platform includes hardware platform (self-developed CPU/AI chip) and self-developed vehicle control operating system. Huawei's autopilot operating system is an open system, just like Android for smart phones or similar to HarmonyOS system. Car companies, Tier 1 and application developers should be supported to develop various intelligent driving algorithms and applications based on the operating system, and the automobile industry should be supported to continuously provide innovative functions and services for intelligent driving.

The third is the ecology of execution components. The most important thing in intelligent driving is command. It is a brain that directs how the final executive part moves. There should also be an interface to access various executive components, such as electric drives and motors from any manufacturer. "We have formulated the interface standard, so that MDC can easily cooperate with all the executive components." Xu Zhijun said, however, Huawei still faces a series of problems and challenges such as laws, regulations, policies and standards, and needs to establish a broad ecological alliance to gather knowledge to promote the establishment of standards.

So, what proportion does Huawei's MDC intelligent driving platform account for in the whole automatic driving process? To what extent can the process of autonomous driving be promoted now?

Xu Zhijun answered the question of auto business review in this way: "Why is it called intelligent driving instead of automatic driving? Fully automatic driving and driverless driving are the ultimate pursuit. Autopilot is a gradual process, and the ultimate goal is to achieve complete unmanned driving, but it can create value in the process of moving towards this ultimate goal. Tesla has created value for everyone. "

For example, Tesla has made several smart driving functions, which consumers like very much. He put forward three functions that everyone would like in the case of road traffic in China, namely, automatic parking function, car parking function and car following function in the case of traffic congestion.

At present, Huawei has provided its partners with the development version of MDC intelligent driving platform, and partners have made intelligent driving applications on this platform.

Intelligent full-scene travel experience

The intelligent cockpit is not just a matter of fewer screens. How to do Huawei CDC intelligent cockpit platform?

About the intelligent cockpit, this is also a very fashionable topic in the automobile industry in recent two years, but it is not easy to really do it, or even talk about it, because it is also an evolving automobile future.

Now it seems that the screen inside the car is getting bigger and bigger, which is a sign of the intelligent cockpit, but obviously, most of the ecological loads are not much different from those of traditional cars. What do we expect from Huawei CDC intelligent cockpit?

Auto business review believes that the intelligent cockpit is a complete system composed of different cockpit electronic devices, rather than a simple LCD instrument, HUD, central control panel, central control vehicle information terminal and back seat. HMI？ Entertainment screen, inside and outside rearview mirror, etc. They are all carriers, but artificial intelligence, AR, ADAS, VR? And other technologies are integrated into the future cockpit layout to enhance the user's car experience and give them services that traditional cars do not have.

Based on the central control platform, the industrial chain of intelligent cockpit gradually extends to LCD instrument, head-up display and back seat entertainment, realizing multi-level information processing and unique human-vehicle interaction.

In-vehicle infotainment system (IVI) is an important carrier of information interaction in intelligent cockpit. It can realize three-dimensional navigation, real-time road conditions, IPTV, assisted driving, fault detection, vehicle information, body control, mobile office, wireless communication, online-based entertainment functions and? TSP？ A series of applications such as services have greatly improved the level of vehicle electronization, networking and intelligence. ?

The cockpit upgrade path can be compared to a smart phone. ADAS, cockpit electronic products are more abundant, the global competition pattern is more dispersed, everything is still evolving, and there is no real oligarchy.

Back to Huawei's CDC intelligent cockpit platform. The so-called CDC, that is? Cockpit? Master. Controller, cockpit domain controller. It can realize the seamless sharing of smart cars and smart phones in the whole industry chain such as hardware, software and application ecology, and establish a data center focusing on automobile scenes.

This kind of * * * enjoys three advantages: first, the IVI module is built on the Kirin chip of the smart phone, which can play the scale effect of industrial chain synergy and reduce the hardware cost; Second, based on the HarmonyOS system OS, * * * enjoys Huawei's "1+8" ecology and realizes cross-terminal full-inductive interconnection; Third, enjoy smart phones, enrich apps, improve car experience, open APIs, and realize the development of intelligent cockpit applications across terminal partners.

Among them, compared with the traditional multi-chip scheme, the single-chip scheme drives the intelligent cockpit, which is similar to the cockpit domain controller scheme, which can simplify the layout of cockpit processors, greatly reduce the system cost, and provide all-round intelligent interconnection experience such as multi-screen interaction.

The intelligent cockpit with one core and multiple screens has become a trend. Like 20 18? Year? 8? Month? On the 7th, Amber announced that it will be a new generation of Haval and Great Wall Motor. WEY？ The brand provides a single-chip intelligent cockpit solution, which can simultaneously drive all functions of vehicle electronic systems such as full-color LCD instrument, head-up display and central control entertainment. Another example is 20 19? At the beginning of the year, Huayang Group launched a new generation of automotive chips? I. MX8? And the latest car operating system? AndroidP？ Infotainment program.

At the same time, in the intelligent cockpit, the airborne hardware is also developing in the direction of modularization, and the proportion of software systems is increasing. Some automakers have begun to put IVI? Modular layout can reduce the complexity of different vehicle configurations and improve the reuse rate of a single module. ?

But what about Huawei's CDC? The intelligent cockpit platform looks better. According to Xu Zhijun, the hardware ecology, software ecology and application ecology accumulated by Huawei intelligent terminals should be brought into the intelligent cockpit. In the future, after autonomous driving, there will be more passenger services and safety services besides entertainment services.

He said: "We have a large number of smart phone users in China and around the world, and the whole industry has established a wide range of smart terminal ecology, which has truly achieved scale and low cost. The intelligent cockpit is in the car. Our biggest advantage is that the intelligent terminal and the intelligent cockpit platform * * * enjoy an ecology. "

Xu Zhijun said that Huawei communicated with car companies about the idea of CDC intelligent cockpit, and everyone most welcomed to move the ecology of Huawei intelligent terminals to cars and enjoy the ecology of smart phones. At the same time, develop API to realize the application development of intelligent cockpit across terminal partners.

For example, not only the entire entertainment system, but also the AR display on the dashboard in the future, and whether the driver is sleeping or distracted before autonomous driving, that is, the driver monitoring system (DMS). Can be solved through the intelligent cockpit platform.

Huawei hopes to empower the digital cockpit and build an intelligent full-scene travel experience through the chip +OS+ ecosystem. These experiences include intelligent escort, infotainment/car-home mutual control, life cycle service, smart office and home theater. The intelligent service engine it provides includes cockpit perception, decision-making and control, multimodal real-time interaction, insensitivity among people, cars and homes, and service discovery.

The ecology of the intelligent cockpit finally built by Huawei, the hardware can be replaced, the application is constantly updated, and the software can be continuously upgraded. Independent account system, cloud service, vehicle? Ota? Ability has become the general trend of smart car cockpit.

This article comes from car home, the author of the car manufacturer, and does not represent car home's position.