Looking back at the process of intelligent transformation, in the early stage, car companies are relatively weak in intelligent driving, so they usually choose to cooperate with intelligent driving solution providers. Take Tesla's intelligent transformation as an example, and its partners have gone through the process of going from Mobileye to Nvidia and finally to independent R&D. Prior to 2020, the major ADAS (Advanced Driver Assistance Systems) solution providers in the market included Mobileye and NVIDIA, with Mobileye at one point accounting for more than 90% of the market share. However, Mobileye's solution is a "black box" closed model, which does not support the self-developed algorithms of OEMs; In contrast, NVIDIA's solution is more flexible and helps automakers develop their own software. Emerging car companies such as Tesla, Xpeng, and NIO have switched from Mobileye to cooperation with NVIDIA, and Tesla is particularly ahead of the curve and already has the ability to have a full stack of software and hardware. Also in 2020, domestic intelligent driving solution manufacturers such as Huawei and Horizon Robotics have also emerged in the domestic market, and have established close cooperative relations with domestic OEMs such as Changan, Chery, and BAIC.
Figure: Changes in ADAS solution manufacturers of smart car companies
Minsheng Securities pointed out that in the past five years, significant breakthroughs have been made in the exploration of the industrial chain in the field of intelligent driving. For example, Tesla has mastered the full chain of capabilities from hardware to software, while domestic companies such as Huawei and Horizon have also made key progress in the field of intelligent driving chips. Looking forward to the future, the decoupling of software and hardware for intelligent driving will become a major trend, which will promote the further refinement and acceleration of the division of labor in the perception layer industry chain.
As smart driving solutions move away from their reliance on Tier 1 suppliers, the industry is shifting to a new paradigm: providing the necessary hardware and software support directly to automakers. This transformation is expected to enhance the position of the relevant auto parts suppliers in the industry chain and bring them new development opportunities.
According to System Plus' research, the main difference between Tesla's Model 3's triple camera and ZF's triple camera is that Tesla's design only captures image information, integrates three CMOS sensors on the same PCB board, and does not require a SOC information processor, but instead transmits image information directly to the controller for processing. In contrast, ZF's CMOS sensors are embedded on three different PCBs, and each camera is equipped with a full SOC, with the final information processed via Mobileye chips.
Figure: Tesla Model 3's triple camera compared to ZF's triple camera
System Plus estimates that ZF's triple-camera cost $165, while Tesla's triple-camera costs $65. For mmWave radar, Tesla chose Continental's ARS4-B radar module, which contains a 77GHz radar chipset and 32-bit MCU provided by NXP, but has not yet achieved decoupling of data acquisition and processing functions.
Minsheng Securities believes that Tesla has realized the decoupling of the camera system, and the industrial chain of other perception layers such as millimeter-wave radar will continue to divide labor, which is the future trend. It is expected that more and more OEMs will gradually move closer to Tesla's model.
In the perception layer of intelligent driving, sensors, cameras, radars, lidars and other devices are the key to collecting environmental information. With the development of technology and the increase of market demand, the division of labor in the industrial chain of the perception layer is accelerating. Enterprises focus on their respective areas of expertise, such as sensor manufacturing, algorithm development, system integration, etc., to achieve more efficient technological innovation and product iteration.
For more info:
