Jiwei Network reported that from the internal combustion engine drive to pure electric drive, from distributed to centralized control evolution, automotive products from closed systems to open systems - cars are gradually transforming into the next large-scale mobile intelligent terminal, high-end automatic driving and even unmanned driving is the ultimate goal of this "unprecedented change in a hundred years".
In the first half of the electrification competition, China's automobile industry has achieved good results, China Automobile Association data show that the Chinese brand's passenger car share reached 44.4% last year, of which the growth mainly comes from new energy vehicle brands, China has become an important driving force for the transformation of pure electric drive in the global automotive industry. The rapid penetration of electrification also has a positive correlation with the intelligence of the car, in which the automatic driving function has gradually become the standard of new models on the market.
"2022 will be the mass production year of domestic vehicle-grade large-scale computing chip." Shan Jizhang, founder and CEO of Black Sesame Intelligence, pointed this out at the China Electric Vehicle 100 Forum (2022) held at the end of March this year.
The industry generally believes that 2022 will be the first year of mass production of China's L2++ level autonomous vehicles, and the large-scale car specification chip, which is the core technology basis for realizing the automatic driving function, is undoubtedly the key to competition.
This "mass production long-distance run" with large computing power chips as the core is entering the sprint stage. After the accumulation and rapid development of local automotive chip companies in the past few years, they will be ready to meet the most critical mass production test.
In 2022, the domestic car regulation large computing chip enters the year of mass production
"Large computing power SOC chips, AI computing platforms, and good image processing capabilities are the basis for the evolution of autonomous driving." Shan Jizhang revealed that the Huashan No. 2 A1000 automatic driving computing chip developed by Black Sesame Intelligent has been fully mature in terms of functional safety, information security and reliability, and will achieve mass production this year to achieve L2-L3 level automatic driving.
Why is the first year of mass production of domestic car regulation large computing power chips in 2022?
In the view of Shan Jizhang, it is very critical to get stuck in the "time point when the main engine factory should define the technical route of the next generation model". In a media interview two years ago, Shan Jizhang pointed out that the cycle of chips is very long, and the cycle of the car is longer, "You have to understand this cycle very well, and understand what is needed in the time window of the market." The batch of cars with mass production in 2022 and 2023 is now (referring to 2020) has been almost defined. ”
The reason why Black Sesame Intelligence was able to seize the first wave of large-scale automatic driving chips on the car is inseparable from its own technology accumulation, business model, forward-looking products and technical routes.
Founded in 2016, Black Sesame Intelligence has just gone through its first five years. At the beginning of its establishment, the company decided to develop large-scale automatic driving chips and platforms, and insisted on developing two core IPs - the vehicle-level image processor NeuralIQ ISP and the DynamAI NN vehicle-level low-power neural network acceleration engine.
According to Shan Jizhang, the company spent three years to develop two core IP, and then launched a number of high-performance intelligent driving computing chips in the Huashan series, defined the Shanhai artificial intelligence development platform, and became the first AI chip company in China to collect functional safety expert certification, functional safety process certification, and product compliance certification. "Now for the main engine factory, in the field of automatic driving to choose chips, at present, only black sesame intelligence in China has products that can be benchmarked with NVIDIA."
Jiwei Network learned that at present, Huashan No. 2 A1000 series chips have completed all vehicle regulation level certification, and it is expected to be mass-produced within the year, which is the largest and most powerful automatic driving chip at this stage, and will also be the first domestic chip platform that meets the car regulations and supports the integrated domain controller of the single chip.
Statistics show that the current domestic passenger car L2 automatic driving has achieved large-scale commercial applications, 22.2% of full-caliber passenger cars in 2021 are equipped with L2 level and below automatic driving system, adaptive cruise (ACC), automatic emergency braking (AEB), lane keeping system (LKS), automatic parking (APA) and other technologies have been widely used on the car. But the real test is high-level autonomous driving above L3 level.
At present, the United States, Germany and other countries have begun to open the door of policy for L3 self-driving cars to officially be on the road. For example, the U.S. Department of Transportation has made it clear that fully autonomous vehicles no longer need to be equipped with manual controls such as traditional steering wheels, brakes or accelerator pedals; Germany announced that L3 self-driving cars will be conditionally allowed to drive on public roads and designated areas starting in 2022. At the end of last year, Mercedes-Benz successfully passed the approval of the relevant departments in Germany, becoming the first automobile company in the world to obtain the L3 level of automatic driving certification of the United Nations "Automatic Lane Keeping System (ALKS)", and also obtained the qualification to be legally on the road on the German highway.
All indications are that the mass production of high-level autonomous driving is heading into the fast lane. However, the real breakthrough from L2 to L3 level is a long process, which involves the close cooperation and upgrading of software, hardware, data and other technologies. Among them, especially the large-scale car specification chip, will be the core key to the breakthrough of high-level automatic driving.
The demand for large-capacity rail chips will usher in an outbreak. However, it is worth noting that there are not many products to choose from at present, and to the more redundant 1000TOPS autopilot chips above the hash rate, there are only Nvidia Orin, Qualcomm, Black Sesame A1000 Pro, Huawei Ascend 610 and other single SoC 200TOPS chips on the market to achieve multiple cascades. This potential demand gap also gives local automotive chipmakers an opportunity. In fact, a number of car companies have begun to develop new driver assistance systems based on local automotive chip manufacturers such as Black Sesame Intelligence and Horizon. For example, Black Sesame Intelligence, as early as the end of 2020, has successively won the front-loading mass production points of many car companies, achieving an important breakthrough in China's local autonomous driving chips.
The sprint of "mass production long-distance running": "chip, software system, track system" trinity
From upstream chips and components to autonomous vehicle computing platforms, system solutions and downstream applications, the intelligent supply chain of automobiles has gradually generated and expanded, and has formed an unprecedented impact and subversion of the original traditional automobile supply system. After the completion of the 0 to 1 technology deployment, it is more important to explore the commercial value of 1 to N, and the "long run" of commercial mass production of autonomous driving has also begun to enter the sprint stage.
Huashan No. 2 A1000 automatic driving computing chip
According to Black Sesame Intelligence, it is a very difficult process for car rules chips to reach mass production, during which many thresholds need to be crossed, including creating reliable chip products, developing mature software systems and building a perfect car rules system.
Jiwei Network learned that Black Sesame Intelligent's Huashan series of autonomous driving chips has experienced more than three years from product design, tape-out, packaging and testing, vehicle regulation certification and algorithm tool chain, to functional safety certification, automatic driving software package development and then to the improvement of the support industry ecology. Since its official release in 2020, the Huashan No. 2 A1000 chip has achieved the optimization and upgrading of the chip system, and all aspects have reached the requirements of stable mass production. At present, huashan No. 2 A1000 chip has been put into large-scale production, began to continue to ship to industry customers, and will achieve mass production in 2022.
Among them, in terms of certification, Huashan No. 2 A1000 chip has passed AEC-Q100 Grade 2 certification and ISO26262 functional safety product ASIL B certification, and has been ready to provide PPAP related support documents to customers. In terms of operating system, Huashan series chips have completed the porting and verification of various mainstream operating systems in the industry, including Linux, QNX, RT-Thread and Windriver. In terms of tool chain, the Shanhai artificial intelligence development platform independently developed by Black Sesame Intelligent has excellent performance, giving full play to the advantages of the A1000 multi-dimensional neural network accelerator and achieving excellent performance of more than 4000fps on the industry's typical model network. Shanhai ARTIFICIAL intelligence development platform has more than 50 kinds of AI reference model library conversion use cases to reduce the threshold of algorithm development for customers, can realize the comprehensive optimization of QAT and post-training quantification, ensure the accuracy of algorithm models, support dynamic heterogeneous multi-core task assignment and customer custom operator development, provide friendly and easy-to-use toolchain development packages and application support, and help customers quickly transplant models and deploy integrated processes.
In addition, it is worth mentioning that the middleware that efficiently supports the operation of the system is a very important part of the automatic driving application, and the self-developed automatic driving middleware for distributed computing has been polished for two years, and can now be adapted to CyberRT, DDS, SOME/IP standard protocols, supporting sensor abstraction, time synchronization, execution management, and management of various system services and software life cycle. Based on this, customers can quickly build upper-layer applications to form a framework for production-level autonomous driving systems. At the same time, Black Sesame Intelligent is also the earliest domestic autonomous driving chip enterprise to lay out and support AUTOSAR. After nearly a year of close cooperation with ecological partners, Huashan No. 2 A1000 series chips have been fully adapted to support Elektrobit and other basic software. This will greatly help customers using AUTOSAR save development and verification time and shorten the series production cycle.
From the complete algorithm system to the data closed-loop capability, software system, and then to the flexible cooperation mode of the industrial chain, Black Sesame Intelligence has built a complete set of advanced and complete process system, is the first domestic set of functional safety expert certification ASILB product certification, ASILD functional safety process certification, AEC-Q100 reliability certification, ASPICE CL2 certification, with a complete vehicle regulation certification of the automatic driving chip company.
At present, Black Sesame Intelligence has formed a complete automatic driving solution from L1 to L3 level, and has carried out a series of commercial cooperation with China FAW, Bosch, SAIC, SAIC-GM-Wuling, Dongfeng Yuexiang, Zhongke Chuangda, Asia Pacific, Baolong Group, Jingwei Hengrun, Junlian Zhixing, Soto Ruian, Lianyou Technology, etc. in L2, L3 level ADAS and automatic driving perception system solutions; algorithms and image processing technologies have achieved layout and commercial landing in the field of intelligent vehicles. In the future, based on the automotive standard design, learning image processing, low-power accurate perception of the automatic driving perception computing chip and automatic driving computing platform, Black Sesame Intelligence will vigorously promote the rapid industrialization of product solutions related to the autonomous driving industry chain
Write at the end:
The rise of China's power under the great change of automobiles
The report "Research on the Ecological Development Strategy of China's Intelligent Transportation Industry" predicts that by 2035, the ecology of the intelligent transportation industry in mainland China will be at the leading level in the world. In the field of intelligent new energy vehicles, China's auto industry is truly at the forefront of the world for the first time, and behind this is the need for the support of the local automotive industry chain to meet the growth and customization needs of domestic automakers. Under the current wave of automotive industry changes, a number of Chinese automotive chip manufacturers can be cultivated. Just as the growth path of the consumer electronics industry represented by smart phones in those years was accompanied by a large number of great chip companies in the United States, Japan, and Europe, and their rapid growth stage was when the electronic products of these countries swept the world. Now that the automotive industry is about to go through the same stage, China has the potential to become the most dynamic market. (Proofreading/Stella)