With the continuous upgrading of intelligent driving hardware and architecture, the industry is hotly discussing how to develop algorithms, software, and operating systems related to them in a coordinated manner to build a new ecosystem of intelligent hardware and software.
On the afternoon of December 6, 2023, at the theme forum of "Building an Automotive Intelligent 'Core' Ecology" at the 2023 Global Automotive Chip Innovation Conference and the 2nd China Automotive Chip Summit Forum, Mei Jinren, Director of Business Operations of Zerobeam Technology, Gao Song, Director of the Algorithm Platform Department of the National Intelligent Connected Vehicle Innovation Center, Wang Zuopeng, Marketing Director of the Embedded Storage Division of Shenzhen Longsys Electronics Co., Ltd., Chief Technology Officer of Wuxi Beiwei Sensing Technology, Shi Guangyi, Professor of the School of Software and Microelectronics of Peking University, Li Shuaijun, Chief Architect of Foretek Intelligent System, Zhang Xiaoxian, Deputy General Manager of Puhua Basic Software Co., Ltd. and Dean of the Strategic Research Institute, Gong Chen, Senior Engineer of Information Security of China Merchants Testing Vehicle Technology Research Institute, Jiangsu Yuntu Semiconductor*** General Manager Geng Xiaoxiang and other industry elites focused on how to quickly improve the technical level of AI intelligent chips in this wave of transformation and upgrading, accelerate the construction of a new ecology of intelligent hardware and software, and improve the level of automotive intelligence. Chen Xinglin, deputy secretary-general of the China National Association, presided over the theme forum. 
Collaborative development will become the main mode of development
Chips have become the industrial chain and value chain center for intelligent driving of intelligent networked vehicles, and play a vital role in the transformation and upgrading of automobiles from electrification to intelligence and networking. However, how to promote its better integration and development in the entire industrial chain still needs to be solved through basic collaboration and innovative practices. For complex cyber-physical systems such as intelligent networked vehicles, the open source and openness of software and hardware, as well as the overall ecological cooperation form, are the directions that need to be focused on. In this context, Gao Song put forward several problems in the development of autonomous driving systems at this stage. "First of all, there is a large number of repetitive 'wheel-making' in the development of autonomous driving systems, which is mainly manifested in the difficulty of reusing massive scene data. Although chip companies, sensor manufacturers, and autonomous driving solution providers have accumulated a large number of data resources, they lack unified standards and sharing mechanisms. "Secondly, the development of basic algorithms between cars and chips lacks an ecology of joint construction and joint application, resulting in a waste of resources. The third is that the investment in assessment capacity is also repetitive, and the utilization efficiency is relatively low. In addition, when it comes to operating systems, Gao Song said that although many companies are engaged in kernel, middleware, virtualization and functional software development, they have not yet formed a major baseline version, nor have they formed a collaborative development ecology based on the baseline version. 
Gao Song, Director of the Algorithm Platform Department of the National Intelligent Connected Vehicle Innovation Center.
Gao Song believes that from the successful experience of open source and openness in the software industry, for the intelligent networked vehicle industry, it is expected to cultivate useful talents through practical engineering traction and practice based on actual positive development and testing processes. The National Intelligent Connected Vehicle Innovation Center extends its work to the entire automotive industry chain, promotes industrial collaboration, and jointly innovates a platform for the integration and collaborative development of industry and education. It is hoped that through pragmatic cooperation with SoC chip companies, sensor companies, OEMs and Tier1 and other resources, the gap between talent demand and supply will be solved, and the common development of colleges and universities and vocational colleges will be realized. In the field of smart cars, AI technologies support the digitalization of R&D, such as quality control, iteration, and knowledge base search. Mei Jinren shared the R&D practice of AI and smart car solutions based on ZeroBeam Technology. In the field of intelligent cockpit, ZeroBeam Technology explores the construction of large models in the cloud, and is also committed to the research of automatic annotation, data synthesis, efficiency improvement, and collaboration between small models on the vehicle side and large models in the cloud. It also focuses on AI basic capacity building and ecological expansion, continues to tackle key problems and independently develops full-stack solutions and products for the cloud, pipe, and end of ZeroBeam Galaxy intelligent vehicles, and is committed to building a more open new ecology of the automotive industry and leading the new trend of "AI + software-defined vehicles". 
Mei Jinren, Director of Business Operations of Zerobeam Technology.
In the field of intelligent driving, a key sensor technology is indispensable for precise positioning and orientation - inertial sensors. In terms of inertial attitude sensors, inertial chip design, algorithms and intelligent vehicle applications, Wuxi Beiwei has made remarkable research achievements. Shi Guangyi made an in-depth analysis of the industrialization and research results of inertial attitude sensors, and said that the research and development of multi-sensor spatiotemporal error calibration, multi-node dynamic tilt algorithm and integrated navigation reliability enhancement algorithm will become the core algorithm of inertial chips involving precise positioning and orientation of intelligent driving. "Therefore, in many intelligent driving applications, we place special emphasis on the precise control of dynamic attitude, and believe that positioning and orientation technology will usher in more applications and promote the industry to continuously achieve new technological achievements. This requires the joint efforts of inertial manufacturers, satellite navigation manufacturers, and high-precision sensor manufacturers to achieve the goal of intelligent driving integrated navigation, including the flexible use of edge computing. ”
Chief Technology Officer of Wuxi Beiwei Sensing Technology
Guangyi Shi, a professor at the School of Software and Microelectronics, Peking University.
Software and hardware determine the future upper limit of intelligent "core".
Today, automotive chips and operating systems play a key role in building the ecosystem of the smart car industry. According to the McKinsey report, by 2023, the global marketing revenue of automotive basic software will reach 8 billion US dollars, the proportion of software in automobiles continues to increase, and software and chips will become the core elements of the competition in the second half of the future automotive revolution. A more efficient, smoother, and more impactful infrastructure built by software and chips will be the highland for future innovation. In this context, the industrial chain relationship between software companies and chip companies is undergoing changes, and many chip and software companies directly provide solutions for automobile manufacturers, breaking the traditional chain relationship and gradually evolving into a complex network ecology. Zhang Xiaoxian believes that this change has brought some coupling risks in terms of convergence security, especially in the environment of multi-network access, multi-entity interconnection and multi-functional integration, which poses new challenges to the combination of functional safety and network information security, and needs to be handled by new methods. Of course, the final solution still needs to be supported by the underlying operating system and chip technology. 
Deputy General Manager of Puhua Basic Software Co., Ltd
Zhang Xiaoxian, President of the Institute of Strategic Studies.
In Li Shuaijun's view, there is a long-term and short-term iterative relationship between algorithms and chips. In the long run, the algorithm defines the direction of the chip, while in the short term, the algorithm needs to adapt to the chip. "Although there are differences in the speed of computing power rise and the level of computing power of each chip, the overall trend is on the rise, which confirms the general trend of continuous development of chip computing power. However, he also said that this technological development may start to diverge: on the one hand, the industry may adopt leaner computing power to build affordable and practical intelligent driving products;On the other hand, chips may continue to develop towards large computing power and try to apply more powerful AI algorithms and models to solve the problem of intelligent driving in a higher dimension. 
Li Shuaijun, Chief Architect of Foretek Intelligent System.
Li Shuaijun also emphasized that the domain controller composed of heterogeneous chips has more landing value at present, and there are still significant differences between the current simple perception chip and the complete intelligent driving solution. "In the development trend of intelligent driving around heterogeneous systems, it has become crucial to take into account more computing cores and storage bandwidth, such as NPUs and AICOREs. This puts forward more comprehensive requirements for chips, including more CPU computing power, higher speed cross-core storage and communication bandwidth, and more secure fault monitoring and system management. Therefore, he called for investment in large computing power chips, and allocated resources for storage bandwidth, positioning chips, etc., to improve the shortcomings of the entire domain control and intelligent driving system, rather than simply improving the long board, in order to achieve better and faster value embodiment. The chip security architecture needs to be comprehensively improved
Automotive intelligence is a field that is highly bound to safety. When talking about the security of automotive data storage, Wang Zuopeng briefly analyzed the current situation and challenges of the automotive data storage market. With the rapid development of new energy vehicles, the application market of on-board memory chips is expanding rapidly. According to research agency data, by 2025, the global demand for DRAM will reach 4.6 billion US dollars, and the Chinese market will account for one-thirdThe demand for NAND Flash will reach $9.8 billion, and the Chinese market will account for $3 billion. This shows that the automotive data storage market is a huge and expanding field. However, the current share of Chinese brands in this market is only in the single digits. "Fortunately, with the requirements of the Data Security Law for vehicle storage security, in the past two years, more and more domestic vehicle specification memory chip companies have entered the automobile market. 
Shenzhen Longsys Electronics Co., Ltd. ***
Wang Zuopeng, Marketing Director of Embedded Storage Division.
Wang Zuopeng said that the standards of the core technology of domestic memory chips mainly include five aspects: storage particles, main control chips, firmware, packaging and testing of control memory particles in the main control chip. However, at present, there is no domestic enterprise that fully covers these five aspects, and most enterprises only meet the autonomy and control of one or two of them. He believes that the localization standards should be phased in, from meeting some of the standards to meeting all the standards, which requires a process. Localization should not only conform to the significance of marketization, but also conform to the significance of national strategy. In Geng Xiaoxiang's view, the safe and controllable development of automotive-grade chips is bound to be a long track. The demands placed on chips are constantly increasing, and chips must strike a balance between reliability and low failure rates. In addition, there are implicit requirements in terms of functional safety and information security. 
Geng Xiaoxiang, general manager of Jiangsu Yuntu Semiconductor.
Focusing on chips and automotive information security, Gong Chen said that building a security system and strategy for the integration of vehicle functions built by the underlying network security chip is very important for the overall construction of intelligent networked vehicles, network security and data security in the future. "In the current context of intelligent networked vehicles, network security and data security, the storage and ECU nodes involved in the vehicle are flooded with a large number of computing networks, especially in the case of unmanned driving and AI algorithm decision-making data in the future, early consideration is particularly important. ”
China Merchants Testing Vehicle Technology Research Institute***
Gong Chen, Senior Engineer of Information Security.
In terms of network security, due to the coverage of CAN network, in-vehicle Ethernet, and the interaction between vehicles and cloud platforms, in Gong Chen's view, reducing the potential attack surface and building a closed-loop network security ecological chain are issues that should be prioritized by the underlying chips. The second is the protection of data security, which requires a comprehensive and complete system architecture at the chip level to ensure the network security conditions in the vehicle-end and Internet of Vehicles environment. In order to promote the development of the entire chip industry, especially in areas closely related to information security such as algorithms and computing power storage, Gong Chen called on participants in various industries to jointly formulate norms and standards to lead the standardization transition of Internet of Vehicles communication chips and chip modules through a series of specifications, and finally establish a certification system for automotive chip security certification. A detailed explanation of the blockbuster report of intelligent driving domain control
Entrusted by the research group of the China National Association, Chen Xinglin interpreted the "China Intelligent Driving Domain Controller Development Report (2023)" (hereinafter referred to as the "Report"). The report is jointly compiled by the China Automobile Association and industry organizations, vehicles, parts and semiconductor-related enterprises, and is an authoritative report that systematically and completely introduces the development status and trend of intelligent driving domain controllers and chips, which will help the competent authorities, automobiles and chains, and service organizations to fully understand the development of the industry and promote the high-quality development of the automobile industry. 
Chen Xinglin, deputy secretary-general of the China Association of the People's Republic of China.
According to the report, with the rapid development of intelligent driving systems, the intelligent driving domain controller has become a combination of software and hardware systems, including supporting chips such as main control intelligent driving SOC chips and power storage, and software including operating systems and middleware, application algorithm software, etc. It is expected that by 2025, the shipment of intelligent driving domain controllers will increase from 114 in 202270,000 sets grew rapidly to 8 million sets. For the core of the intelligent driving domain controller industry chain - intelligent driving SoC, the report suggests that the SoC function should be upgraded from standard to customized to support the development of ultra-large computing power of cabin and driving integration. At present, international chip manufacturers have fully launched product matrices, leading computing power, and some manufacturers have released intelligent driving SOC chips with computing power of up to 1000TOPS. The domestic intelligent driving SoC chip has been rapidly expanding in the L2 and L2+ markets, and is equipped with some models. In comparison, there is a big gap between domestic core and large computing power SoC chips and international leading enterprises in terms of maturity and brand premium. In addition, the report also sorted out the technical route and core manufacturers of the intelligent driving domain controller, and believed that the software and hardware of the driving and parking system are gradually intersecting, the integrated driving and parking solution will be implemented on a large scale, and the intelligent driving integrated domain controller will become an important direction in the future through the integration of SOC. Foreign domain controller manufacturers include traditional Tier1 manufacturers and self-developed domain controller OEMs, while domestic intelligent driving domain controller participants include local Tier1 enterprises, technology companies and vehicle companies, and vehicle companies are more inclined to full-stack self-development. Domestic domain controller manufacturers have a first-mover advantage in L2+ high-level assisted driving application scenarios, but the stability of the first chain is weaker than that of international manufacturers. Conditions and suggestions for the development of intelligent driving domain controllers. With the rapid development of China's intelligent networked vehicle industry, the demand for automotive chips and intelligent driving domain controllers has increased, and the report recommends the establishment of intelligent driving SOC and intelligent driving domain controller research projects to accelerate key technological breakthroughs and generalization, standardization and large-scale development. The industry should strengthen the research of the standard system and form an independent and controllable test, evaluation, certification and accreditation system. At the enterprise level, it is recommended that domestic new energy vehicle companies strengthen their own capabilities, develop large computing power and medium and low computing power chips, intelligent driving domain controllers, and pay attention to the design of functional safety and information security.