Electronics

Autonomous driving transforms the brains of electric vehicles, and the market expects that SoCs with artificial intelligence computing functions will become the mainstream of autonomous driving chips above L4; including Tesla (Tesla), Nvidia (Nvidia) and other international manufacturers actively deploy .

According to He Xinyu, a senior industry analyst at the Industrial Intelligence Research Institute (MIC) of the Taiwan Association for Information Technology, analyzes that electric vehicles (EV) and autonomous driving (AD) accelerate the development of automotive controllers (MCUs) towards regional and centralized control applications, requiring higher Computing power and highly integrated design architecture have driven the transformation of automotive control chips towards system-on-chip (SoC).

Especially for the automatic driving function of electric vehicles, He Xinyu expects that the system-on-a-chip will compete with the car control chip, especially the automatic driving function above the L4 level, which will reduce the number of microcontrollers and increase the number of artificial intelligence computing capabilities. System-on-a-chip, integrates sensing components through a regional design framework.

Observing the current development trend of autopilot system single-chip manufacturers, He Xinyu pointed out that, for example, the electric car manufacturer Tesla has switched to self-designed and developed autopilot system single-chips. Traditional car manufacturers include autopilot chip architectures from manufacturers such as BMW and Audi. It adopts L2 level stack L3 level design, integrating controller and system single chip.

The legal person estimates that in the future, the single chip of the autonomous driving system above the L4 level will enhance the higher-efficiency artificial intelligence computing capabilities by adding neural network units.

Observing the single-chip design of the autonomous driving system, the legal person pointed out that there are three main architectures. Among them, GPU chip manufacturers NVIDIA and Tesla use processors to integrate special application chips and graphics chips (CPU+ASIC+GPU) design architecture; Intel (Intel) ) Mobileye and China's Horizon Robotics (horizon robotics), which are reinvested, use CPU+ASIC architecture; Alphabet's subsidiary Waymo and Baidu's Apollo use CPU+Field Programmable Logic Gate Array (FPGA) architecture.

Global chip design and software manufacturers are actively deploying single chips for autonomous driving systems. The legal person said that the United States includes Nvidia, Mobileye, Qualcomm, Xilinx, etc., France has VSORA, and Sweden has Veoneer, which Qualcomm has recently finalized acquisition; Mainland China includes Horizon, Hisilicon under Huawei, Black Sesame Technologies invested by Xiaomi, Xinchi Technology, Cambrian Technology, etc.

It is worth noting that in the layout of electric vehicle semiconductors, Hon Hai Group also locks in the three major areas of ZCU of smart gateway, smart driving system single chip and vehicle infotainment system (IVI) smart cockpit system single chip.

According to legal analysis, Mobileye currently accounts for more than 70% of the market for L1 and L2 autonomous driving chips, and Nvidia has locked up a single chip for autonomous driving systems above L3. Most Chinese manufacturers are still unable to mass produce autonomous driving single chips, and only Horizon mass production is small. Part of the chips cut into China's domestic ideal electric car.

Observing the competition and cooperation between vehicle controllers and system-on-chips in electric vehicles and autonomous driving, He Xinyu believes that system-on-a-chips cannot completely replace vehicle controllers in the short term because automobiles are complex systems with electric vehicle chassis, body and power. The system is still based on vehicle controllers.

She took the Tesla Model 3's chassis body control platform as an example. Compared with the Model S's distributed architecture, the Model 3 chassis body platform adopts a regional control architecture. The number of controllers is limited and the controller integrates related functions.

He Xinyu expects that the number of microcontrollers used in electric vehicle chassis and body applications will remain stable, but the functions will tend to weaken, and the braking function will be the main function; the trend of electric vehicles and autonomous driving will not significantly change the vehicle use required by the power system Controller.

In addition, she believes that the newly-added battery management system (BMS), vehicle controller (VCU) and gearboxes of electric vehicles still require high-end vehicle control components. Therefore, electric vehicles require high-end main On the contrary, the price of the controller will be higher, and the controller in the power system area will show an increase in price and volume.