With the rise of China's domestic automobile manufacturing industry, the development of assisted driving ( ADAS ) is accelerating. At present, the process of intelligentization in the automobile industry is in the transition stage from L2 to L3. Although it has not yet reached L3, it has exceeded L2 and is in the L2+ stage.
In late April, Counterpointresearch released a statistical and forecast data. The agency believes that in 2024, the global sales of passenger cars with L3 ADAS level will exceed 25,000 units, and the Chinese market is an important driving force. It is expected that by 2026, the installed capacity of L3 passenger cars in China will exceed 1 million, accounting for 10% of the total shipments.
Counterpointresearch believes that the Chinese market has several advantages in developing ADAS, including government support, the issuance of multiple L3 test licenses, and the progress and technical accumulation of multiple suppliers in L2 testing. The above forecast is relatively optimistic, especially the level of 1 million L3 passenger cars in 2026, which is very desirable. In 2024, the number of 25,000 L3 passenger cars in the world is still relatively objective, which is a very small proportion among the tens of millions of passenger cars with ADAS systems in the world. It appears more in the field of commercial vehicles, such as Baidu's driverless taxis. In 2024, L3 passenger cars are still difficult to enter ordinary people's homes. So why is it so difficult to implement and promote L3 ADAS? This requires first looking at the specific levels from L1 to L5.
According to the definition of SAE in the United States, L1 vehicles can assist drivers in completing certain driving tasks in certain situations; L2 vehicles can complete certain driving tasks independently, but the driver needs to always observe the surrounding environment and take over when necessary; L3 vehicles can drive automatically , and the driver is almost not required to be ready to take over at any time; L4 means that there is no need for driver control under certain specific conditions; L5 vehicles can complete automatic driving under any conditions. It can be seen that the current ADAS system is in the transition stage from L2 to L3. In most cases, automatic driving cannot be achieved. In a few cases, it can be achieved, but the market size and proportion are very small.
01. Gains from the journey to L3
Four years ago, the automotive industry put forward slogans such as " software defines cars", " computing power arms race", and "centralization of E/E architecture". After that, a new track was born, represented by computing power chips , underlying software, algorithms, domain controllers , etc. At that time, companies represented by Pony.ai and WeRide advocated a "leapfrog" implementation approach, believing that the best implementation scenario for L4 autonomous driving is Robotaxi.
On the contrary, new car-making forces represented by Tesla believe that autonomous driving should be implemented in a "gradual" manner, starting from L2, gradually transitioning to L2+, L3, and finally achieving L4. From the perspective of technical routes, there are also disputes between multi- sensor fusion and pure vision routes. Chinese companies advocate enhancing the perception ability of single vehicles by installing laser radar , millimeter-wave radar , cameras , etc., while companies such as Tesla and Mobileye are taking the pure vision route, focusing on the development and iteration of visual algorithms.
After several years of iteration, some participants in the L4 track have begun to land in demonstration areas across China, but due to factors such as cost and laws and regulations, the large-scale commercialization of Robotaxi is still a long way off. On the contrary, new car-making forces represented by Wei, Xiao, and Li have begun to promote L2+ functions such as commuter NOA and city NOA on a large scale. Under this situation, some businesses that previously specialized in L4 have to participate in the L2+ solution business in order to survive.
In general, "gradual" has become the mainstream implementation method of autonomous driving. From the perspective of technical routes, Tesla's "BEV+Transformer", "Occupancy" and other visual algorithm technology architectures have been successful in North America through the FSD system. Chinese manufacturers have also begun to follow Tesla's example and rebuild their own perception architectures, and ADAS perception algorithms have become the focus of merchants.
02. ADAS hardware iteration
In the traditional distributed E/E architecture, the assisted driving system consists of several independent subsystems (such as forward ADAS, side and rear ADAS, parking assistance system, panoramic surround view system, etc.), and each subsystem has an ECU . The main structure of ECU is " single chip microcomputer + peripheral circuit ". In this architecture, Tier1 manufacturers package the software and hardware and provide them to the OEM in the form of "black box" delivery. Mobileye is a typical representative.
As the E/E architecture of the whole vehicle moves from distributed to centralized, the ECU corresponding to the ADAS subsystem is also integrated into the auxiliary driving domain controller, the main control chip evolves from MCU to a higher-performance SoC chip, and the software architecture is also upgraded to SOA architecture, including system software (virtual machine, middleware , etc.), algorithm module and application layer, realizing "software and hardware decoupling". The entire ADAS industry chain is also decomposed into several major links such as chip, hardware integration and production, software development, algorithm development, and application development.
In the early stage of industry transformation, a number of startups have emerged in the chip, middleware, algorithm development and other links. Their technical barriers lie in whether they have sufficient development capabilities and mass production experience in their respective links. For example, in the past three years, Desay SV has won many orders from car companies based on its mass production experience of ADAS domain controllers based on NVIDIA Orin chips. However, as some low- and medium-power intelligent driving domain controllers gradually move towards standardization, the requirements for the capabilities of excellent Tier1 companies (including chip suppliers, integration suppliers, algorithm suppliers, etc.) are no longer limited to a single link in the industrial chain, but need to rely on their leading advantages to integrate the upstream and downstream of the industrial chain, and have a comprehensive supply capability that integrates chips, algorithms, manufacturing, etc., in order to establish an ecosystem. Judging from the development since 2023 and 2024, Chinese manufacturers that have the ability to establish an ecosystem include Huawei (full stack self-developed from bottom-level chips to upper-level algorithms) and DJI (with self-developed algorithms and manufacturing capabilities, and can maximize chip performance).
At present, NVIDIA's Orin chip accounts for 75% of the market share of NOA main control chips. In April this year, the Chinese manufacturer Horizon released the new product Journey 6, which supports the city NOA function. Both NVIDIA and Horizon are trying their best to supplement the algorithm capabilities and gradually have the ability to provide a complete solution for intelligent driving .
In addition, Momenta, the leader in intelligent driving algorithms, has also formed a chip team to supplement the underlying hardware capabilities. To date, the hardware system of ADAS needs further iteration to prepare for L3 driving, such as domain controllers, cameras, various radars, etc. The functions of L3 ADAS are more intelligent, requiring the underlying chips (mainly domain controllers) to have higher computing power. At the same time, the requirements for low power consumption and compatibility will increase. To achieve L3 ADAS, vehicles are required to have strong perception capabilities, and the assembly volume and performance requirements for perception devices such as cameras, millimeter-wave radars, and lidars will increase. Among them, cameras will evolve to higher pixels, and millimeter-wave radars and lidars are expected to provide ADAS with stronger road information collection capabilities when pure vision solutions are not yet mature, and the penetration rate is expected to continue to increase. Compared with traditional mechanical hydraulic braking/steering, wire control braking/steering has the advantages of fast response speed, high compatibility with electrification architecture, energy recovery, and configurable multiple sets of redundant mechanisms, which are more suitable for L3 assisted driving vehicles. As the technology matures, wire control braking/steering is expected to become the standard for L3 and above intelligent driving. The driver monitoring system (DMS) is used to detect the driver's identity, fatigue driving and dangerous behavior. For L3 autonomous driving vehicles, the driver is required to take over the control of the car under special circumstances. Some national laws and regulations also stipulate whether the driver can make phone calls, watch entertainment systems, etc. under L3 autonomous driving conditions, which requires the configuration of DMS for monitoring to determine responsibility when an accident occurs. It is expected that DMS will become standard for L3 driving. China has issued the "Draft for Comments on Light Signal Devices and Systems for Motor Vehicles and Trailers", requiring L3 and above intelligent driving vehicles to be equipped with at least 4 blue-green autonomous driving sign lights, located in the front, rear, left and right of the vehicle body, to inform surrounding vehicles that they are in autonomous driving status. With the official implementation of the "Light Signal Devices and Systems for Motor Vehicles and Trailers" standard, the autonomous driving status indicator light will become a new market for the headlight industry.
Previous article:The timing of automotive semiconductors
Next article:Changan's L3 autonomous driving mass production experience is successful, and China's smart cars have reached a new milestone
- Popular Resources
- Popular amplifiers
- Next-generation automotive microcontrollers: STMicroelectronics technology analysis
- WPG World Peace Group launches automotive headlight solution based on easy-to-charge semiconductor products
- What is the car ZCU that we talk about every day?
- An article reviews the "no-map" intelligent driving solutions of various car companies
- Renesas takes the lead in launching multi-domain fusion SoC using automotive-grade 3nm process
- BYD and Huawei have made another big move!
- V2X technology accelerates, paving the way for advanced autonomous driving
- Rimac and Ceer to supply fully integrated electric drive systems for electric vehicles
- Huawei's all-solid-state battery has surfaced, achieving a major technological breakthrough!
- Red Hat announces definitive agreement to acquire Neural Magic
- 5G network speed is faster than 4G, but the perception is poor! Wu Hequan: 6G standard formulation should focus on user needs
- SEMI report: Global silicon wafer shipments increased by 6% in the third quarter of 2024
- OpenAI calls for a "North American Artificial Intelligence Alliance" to compete with China
- OpenAI is rumored to be launching a new intelligent body that can automatically perform tasks for users
- Nidec Intelligent Motion is the first to launch an electric clutch ECU for two-wheeled vehicles
- Nidec Intelligent Motion is the first to launch an electric clutch ECU for two-wheeled vehicles
- ASML provides update on market opportunities at 2024 Investor Day
- Arm: Focusing on efficient computing platforms, we work together to build a sustainable future
- AMD to cut 4% of its workforce to gain a stronger position in artificial intelligence chips
- Industrial Board Disassembly Appreciation DAQe-2206
- What is the function of each layer of PCB? (Detailed explanation)
- BlueNRG video tutorial: BlueNRG-1 GUI advanced
- Raspiberry Pico runs openmv successfully
- Maxim's open source portable precision calibrator
- About the story of Xiaobai writing bootloader
- Transient response
- 10 Rules for PCB Component Layout
- eetop.cn_100 Practical Tips for FPGA Design Experts (Original English Version)
- Some experience summary on AD conversion design