Huawei Wang Jun： Building a Smart Driving Technology System with Independent Innovation.

　　[Industry of car home] On December 16th-17th, the Global Smart Car Industry Summit (GIV2022) hosted by committee of 100, an electric vehicle from China, was held in Hefei. At the meeting, BU COO, Huawei’s smart car solution, and Wang Jun, president of the smart driving solution product line, said that we should build an independent and innovative smart driving technology system. After cutting off the supply of high-performance GPU chips in the United States, it has a huge impact on the field of autonomous driving. The rising architecture built by Huawei is self-developed from the bottom hardware, AI framework, training platform and tool chain, and does not rely on American technology.

　　Wang Jun introduced that after more than three years of development, Huawei’s smart car solutions have entered the stage of scale delivery. By the end of October this year, Huawei had released 1.72 million sets of components. Now it is the development stage of fuel vehicle to electric vehicle, and it is evolving to intelligence. Electrification is the change of energy system, and the essence is still the change of hardware. Hardware determines the upper limit of experience, and the performance and experience of electrification will gradually converge. The differentiation of electric vehicle experience mainly comes from intelligent systems, especially intelligent cockpit and intelligent driving. There are great differences between different models. The core of intelligence is computing power, software and data. Hardware capability is only the lower limit of experience, while software determines the upper limit of experience.

　　In practice, Huawei found that electrification is the best carrier of intelligence, and there are great differences between them. These two completely different technical systems, Huawei is trying to deeply integrate them, resulting in the effect of 1+1>2, becoming a bridge on the road of intelligent transformation of automobile enterprises and creating an intelligent electric vehicle with the best experience.

　　Intelligentization is a new track. Huawei believes that the system with the best experience and leading technology can be built by adopting the full-stack integration solution, such as the HarmonyOS cockpit system with optional mode and the high-order intelligent driving assistance system with HM mode. However, this method also faces the problems of large investment, difficult technology and long cycle. In the process of building a full-stack solution, Huawei introduced parts technology and products, and adopted diversified modes of cooperation to achieve a win-win situation according to the situation and needs of different car companies.

　　Wang Jun said that smart cars are highly personalized, customized and high-value consumer goods that can be produced in large quantities. To achieve a balance in contradictions, we need a strong infrastructure and development platform, which can be flexibly customized to form differentiation, and can be quickly developed and replicated. Huawei Smart Car Digital Base can provide basic network and hardware platform for car companies, and also provide basic software platform including underlying operating system, atomization service and development tools, so as to improve development efficiency, enhance experience through vehicle collaboration, and reduce development difficulty through open interface. Many models applying this platform are under development, and will be released and listed one after another from 2023. The new model has stronger expansibility and faster iterative development speed. On September 28th, 2022, SDV Working Group of China Automobile Industry Association released the official version of Software-Defined Automobile Service API Reference Specification 3.0, including 501 atomic service API interfaces and 343 equipment abstract API interfaces. As an important member of SDV working group, Huawei has completed the system domain integration with more than 30 devices from more than 20 manufacturers based on standardized interface specifications, and the development efficiency of new services has increased tenfold.

　　Wang Jun shared two integrated solutions. In terms of intelligent cockpit, Huawei deeply integrates the minimalist human-computer interaction experience of smart phones with the sophisticated space experience of automobiles, and builds a full-stack intelligent cockpit solution Harmony Space to provide full-scene smart travel services for drivers and passengers. Huawei is constantly improving its capabilities in voice, multimodal interaction, screen, Huawei Sound and ecology, which will be gradually released in the future.

　　Based on the Harmony OS car operating system, the partners jointly define the hardware interface, so that the hardware can be plug and play, replaceable and upgraded, and diversified hardware can be interconnected. Nine types of CASE are developed for car scenes, which open the language, vision, Huawei Sound and the underlying hardware capabilities of partners to applications through API interfaces, and rapidly develop a multi-device cooperative cockpit system. Our partner’s car canopy, electronic rearview mirror, holographic projection, steering system and intelligent health seat equipment are being applied to customers’ new models one after another after the integration test, helping partners to achieve commercial success and build innovative cockpit experience for consumers.

　　In terms of intelligent driving solutions, Wang Jun believes that intelligent driving can’t be achieved by piling up hardware, which can’t produce a good experience. Now the market is very "rolled", and we can often see how much computing power and sensors are deployed in publicity, but what kind of intelligent driving experience can be achieved is uncertain. Three years ago, we defined the highest hardware specifications in the industry at that time. Later, everyone said that Huawei’s intelligent driving system was good. Because the hardware configuration was very high and strong, the industry began to "roll up" in hardware competition, and fell into a strange circle of attaching importance to hardware, ignoring experience and being weak in security. This should not be advocated.

　　The most difficult thing for intelligent driving is the breakthrough of a large number of long-tail difficult scenes, which requires strong software algorithm ability and massive test and verification accumulation to achieve the most comfortable driving. Moreover, it is the best intelligent driving system to be safer than human driving. Huawei has been pursuing the ultimate experience similar to human driving. In September, Shenzhen officially released the high-precision map, and the Extreme Fox Alpha S HI version equipped with Huawei ADS system launched the NCA self-driving navigation assistance function for Shenzhen car owners. Huawei ADS is the first system to support advanced driving assistance in urban areas. I often drive this car when I go to work, and I use self-driving throughout the 20-30 km road, making it easy to realize point-to-point intelligent driving assistance. At present, the function of urban NCA depends on high-precision maps, and the cycle of composition and map review is relatively long. With the gradual release of high-precision maps in major cities in China, the function of urban NCA will be opened one after another, and the function of urban NCA in Shanghai will be officially launched in December.

　　Safety is the core of intelligent driving. Huawei proposes that driving should be safer and driving should be easier. Huawei has designed a 360 safety protection system with 16 active safety functions when driving, which gives full play to the integrated sensing ability and improves many active safety capabilities. For example, the traditional AEB, Huawei is faster, can find abnormal objects and realize active safety treatment. The sensing system configured for self-driving can be used in driving, and the overall self-driving hardware equipment is worthwhile. At present, the intelligent driving system is basically a fragmented experience. Huawei focuses on three driving scenarios: high-speed, urban area and parking, and strives to break through difficult scenes, open up the self-driving function of the whole scene, enhance the continuity of user experience and make intelligent driving easier.

　　Wang Jun called for building an intelligent driving technology system with independent innovation. Not long ago, the United States restricted the export of high-performance GPU chips to China, and AI sanctions began to be implemented. High-end chips mainly serve servers in high-performance data centers and are used for large-scale AI training in the cloud, which is very important for AI algorithm training of autonomous driving. Cutting off the supply of these high-end chips has a huge impact on the field of autonomous driving. Huawei has built an AI infrastructure solution for the whole scene of "end, edge and cloud" based on the AI processors of the Ascension series, covering the whole process of reasoning, training and deployment in the field of in-depth learning. Ascension architecture is self-developed from the bottom hardware, AI framework, training platform and tool chain, and does not rely on American technology.

　　The core of Huawei’s intelligent driving computing platform MDC is Ascension Series. Based on Ascension Computing, an Octopus cloud training platform is built, and an intelligent driving development, training and deployment platform with end-cloud collaboration is built, which takes root in technology and gives the world a second choice. Huawei has migrated all the training of intelligent driving to Ascension architecture, which means that many operators of Huawei’s own training have been built on Ascension and are open to other algorithm developers.

　　"We hope that everyone can use this system together, the crowdfunding operator library, and continue to develop in the black land of China, getting better and better. We will continue to work closely with universities and educational institutions to provide developers with training in Ascension System, and provide tools and technical support in the process of migrating to Ascension, and join hands with China’s industrial partners to build an independent and innovative intelligent driving technology system, share technological innovation achievements and win the future of intelligent driving. "