1
The skeleton of autonomous vehicles – drive-by-wire chassis technology
As autonomous vehicles, their chassis differs significantly from traditional cars. As the primary platform for intelligent driving, autonomous vehicles utilize drive-by-wire chassis technology, and future advanced autonomous driving will be based on such chassis. Indeed, the traditional and cumbersome "iron machines" will be replaced by sensors, control units, and electromagnetic actuators driven by electrical signals.
Drive-by-wire technology refers to a technology that uses "wires" or electrical signals to transmit control, replacing the "hard" connections of traditional mechanical linkages to achieve operation. A drive-by-wire chassis consists of five major systems: steering, braking, shifting, throttle, and suspension. Drive-by-wire systems eliminate some bulky and less precise pneumatic, hydraulic, and mechanical connections, replacing them with electrically driven sensors, control units, and electromagnetic actuators. Therefore, they offer advantages such as compact structure, good controllability, and fast response speed.
In the future, as the three core technologies (electric drive, battery, and motor) gradually mature, charging convenience will be greatly improved, and safety and reliability issues will be largely resolved. Drive-by-wire chassis technology will also gradually mature, leading a revolution in the automotive industry.
2
The eyes of autonomous vehicles—sensors
When it comes to the hardware system of autonomous driving, the most important components are the various sensors that perceive external information. Thanks to the collection of various information from the world by these sensors, algorithms can better understand the external scene and situation, which directly determines the final stability and safety of the autonomous driving system.
Currently, the sensors equipped on autonomous vehicles mainly fall into several categories, including LiDAR, millimeter-wave radar, cameras, and ultrasonic radar, each with its own distinct characteristics and applicable scenarios. Cameras, with their mature technology and low cost, have become the first and most widely used perception hardware installed in vehicles. In-vehicle cameras are the primary visual sensors in ADAS systems and are among the most mature in-vehicle sensors. However, because cameras, like the human eye, passively receive visible light, their visual performance is poor in backlighting or complex lighting conditions, and they are also susceptible to adverse weather conditions.
Millimeter-wave radar is least affected by weather conditions and offers the best all-weather performance. Similar in principle to lidar, millimeter-wave radar commonly uses 24GHz, 77GHz, and 79GHz frequency bands in automotive applications, corresponding to short-range, long-range, and medium-range radar respectively. While millimeter-wave radar's long wavelength provides excellent object-orbiting capability and minimizes weather-related impacts, its detection accuracy is significantly reduced due to the excessively long wavelength.
LiDAR offers the highest accuracy, meeting the requirements of L3-L5 autonomous driving. Using laser light as a carrier wave, LiDAR has a shorter wavelength than millimeter waves, resulting in high detection accuracy and long range. Furthermore, LiDAR can generate 3D "point cloud" images of obstacles by collecting information from laser beams in different directions. Currently, due to its high technical difficulty and cost, it has not yet achieved large-scale vehicle installation. However, as the industry chain matures and costs decrease, the LiDAR industry may experience explosive growth.
Due to the distinct characteristics and cost considerations of various sensors, most mass-produced vehicles on the market currently only equip themselves with onboard cameras. The final layout and selection of sensors for autonomous vehicles still needs to be tested over time. However, in any case, the demand for sensors will definitely increase in the future, and as the accuracy of sensors gradually improves and the cost gradually decreases, it will bring about a qualitative change to autonomous driving technology.
3
The brain of autonomous driving – computing platforms and chips
In the era of autonomous driving, both automotive architecture and main control chips will exhibit a significant trend towards centralization. Specifically, automotive architecture will evolve from a distributed E/E architecture to a centralized computing architecture, and main control chips will shift from a single CPU to a System-on-a-Chip (SoC) chip incorporating AI modules. In particular, controllers need to receive, analyze, and process a large volume of complex signals, and intelligent vehicle systems need to handle massive amounts of unstructured data such as images and videos. The existing distributed computing architecture corresponding to a single ECU or a single-module domain controller cannot meet future demands, making a centralized computing architecture the primary development trend. A centralized computing architecture allows for the control of the entire vehicle using a single computer, facilitating more convenient over-the-air (OTA) software upgrades. Therefore, autonomous driving SoC chips with high computing power and low latency characteristics have significant growth potential in the future.
Automotive data processing chips mainly include two types: MCUs (Microcontroller Modules) and SoCs (System-on-a-Chip). MCUs have a simple structure, containing only a single processor unit (CPU + memory + interface unit). They appropriately reduce the CPU frequency and specifications, integrating memory, interfaces, and other structures into a single chip, primarily used for control command calculations in the ECU. SoCs, on the other hand, include multiple processor units (CPU + GPU + DSP + NPU + memory + interface unit), exhibiting a higher degree of integration. The future trend of automotive intelligence demands a significant increase in the intelligent architecture and algorithm computing power of vehicles, driving a rapid shift in automotive chips towards equipping them with more powerful SoC chips.
Currently, the autonomous driving chip market is mainly monopolized by foreign giants. Intel's Mobileye was the first company to mass-produce and install autonomous driving chips in vehicles. Subsequently, Nvidia launched autonomous driving chips with better performance, and Tesla, as an OEM, quickly launched autonomous driving chips for its electric vehicles. Horizon Robotics' chips are currently the only domestically mass-produced and installed products. Huawei and Black Sesame Technologies are also among the top domestic players due to the excellent performance of their chips. In addition, companies such as DeePhi Tech, Cambricon, and Westwell Technology have also joined the competition in the domestic autonomous driving chip industry. In the future, the development of the autonomous driving chip field will directly determine the market size and the success or failure of companies.
4
Summarize
The above analysis covers the main hardware components currently found in autonomous vehicles. These components will play a crucial role in the future development of autonomous driving technology, and their functionality, safety, and stability will directly impact the size of the autonomous driving market and public acceptance. As hardware technology matures and costs decrease, it will enable autonomous driving software algorithms to perform more complex and safer operations, ultimately determining the upper limit of autonomous driving's future development.
