This website uses cookies. By using this site, you consent to the use of cookies. For more information, please take a look at our Privacy Policy.

From ADAS to Autonomous Driving 2022

Dec 17, 2022      View: 269

Definition of ADAS

 

Advanced Driver Assistance System (ADAS) is a system that uses various sensors (millimeter wave radar, LIDAR, mono/dual vision camera and satellite navigation) installed in the car to sense the surrounding environment, collect data, identify, detect and track static and dynamic objects at any time while the car is in motion, and combine with navigation map data to perform system calculations and analysis, thus allowing the driver to This allows the driver to be aware of possible hazards in advance, effectively increasing the comfort and safety of car driving.

 

adas

The primary stage of ADAS technology is generally called Driving Assist (DAS), which is generally in the L1~L2 level of autonomous driving. Traditional DAS functions are generally based on simple information about the vehicle's own state for judgment and execution (e.g. ESC), without sensing the surrounding environment. Advanced Assisted Driving Systems (ADAS), on the other hand, have sensors that are applied to collect and analyze information in the vehicle's surroundings and can perform complex signal processing as needed to support the appropriate driving tasks.

 

The practical application scenarios of automotive advanced assisted driving systems usually include: navigation and real-time traffic system TMC, electronic police system (ISA), vehicle networking (VCS), adaptive cruise control (ACC), lane departure warning system (LDWS), lane keeping system (LKA), collision avoidance or pre-collision system (CAS or PS), night vision system (NVS), adaptive light control (ALC), Pedestrian Protection System (PPS), Automatic Parking System (AP), Traffic Sign Recognition (TSR), Blind Spot Detection (BD), Driver Fatigue Detection (DDD), Hill Descent Control (HDC), and Electric Vehicle Warning (EVWS) systems.

 

After the L3 level, ADAS will evolve to more advanced automated driving assistance systems, which can then perform driving tasks independently without driver involvement.

 

The Current Market Situation of ADAS

 

In the context of the "new four", the rapid development of autonomous driving technology is reshaping the traditional automotive industry. The four major automobile production and consumption markets in Europe, the United States, Japan and China are leading the development direction of global autonomous driving. In terms of technology, Chinese companies have already reached the first-tier echelon in the development and reserve of autonomous driving technology. However, due to the constraints of policies and regulations, the automation level is still generally at the L2 stage. According to data provided by Head Leopard Institute, it is expected that 63% of the world's cars will have L2 or quotient levels of autonomous driving by 2025. The global autonomous driving market is expected to reach $37.2 billion by 2023.

 

current market of adas

In November 2020, the country released the "Intelligent Vehicle Development Roadmap 2.0", which sets clear targets for the sales penetration rate and sales volume of new cars with different autonomous driving levels: sales of new cars equipped with L2 + and L3 autonomous driving functions are to reach more than 50% and 70% in 2025 and 2030 respectively; the percentage of sales of new cars equipped with L4-level autonomous driving functions is to reach 20%. According to IHS Markit forecast, the penetration rate of autonomous driving will increase rapidly in the future, and the penetration rate of L2 and above autonomous driving will reach about 34% in 2025, of which the penetration rate of L2 level is the fastest growing, and it is expected that the rate of L2 level autonomous driving system equipped will exceed 30% in 2025.

 

According to IDC forecast, global L1-L5 autonomous vehicle sales will reach 54.25 million units in 2024. IHS Markit data shows that by 2025, China's L2 level and above smart driving vehicle market penetration rate is expected to reach 34.2%, and the cost of autonomous driving systems will decrease as the price of sensors, controllers and actuators continues to dip, and the smart driving market space The increment is huge.

 

Technology Trends of ADAS

 

With the shift of autonomous driving from L2 to L3, in order to improve the safety and intelligence of autonomous driving, cars start to add sensors (LIDAR, cameras, millimeter wave radar, etc.) to achieve a more comprehensive collection of road conditions and environmental information, which requires increasing real-time, complexity and accuracy of data processing by in-vehicle chips, and the demand for in-vehicle computing power will also see exponential growth. In addition, the increase in the level of autonomous driving will also bring new business models.

 

Sensor

 

Increase autonomous driving capabilities by increasing the number of sensors and allowing multiple sensors to fuse. Multiple sensors of the same type or different types obtain information of different localities and categories, which may complement each other or be redundant and contradictory, and the control center can only give the only-right command in the end, which requires the control center to fuse the information obtained by multiple sensors and make comprehensive judgment. In the case of using multiple sensors, information fusion of sensors is necessary to ensure safety. Multi-sensor fusion can significantly improve the redundancy and fault tolerance of the system, thus ensuring fast and correct decision-making, which is an inevitable trend for autonomous driving.

 

Of course, there are prerequisites to achieve sensor fusion. Hardware level, the number should be sufficient, that is, different kinds of sensors should be equipped to ensure adequate information acquisition and redundancy; software level, the algorithm should be optimized enough, the data processing speed should be fast enough, and fault tolerance should be good, in order to ensure the speed and correctness of the final decision.

 

The number of cameras, millimeter wave radar, LIDAR, ultrasonic radar ...... various sensors - must be more and more, - on the one hand, there will be integrated sensors, the other - will be distributed sensors for the unified The number of sensors - will be more and more, - on the one hand, there will be integrated sensors, and on the other hand, there will be distributed sensors for unified computing, which can further - reduce costs and improve computing power. For example, the millimeter wave radar is now developed to 4D radar, using multiple radar cascade.

 

Multi-sensor fusion is not difficult to implement at the hardware level, the focus and difficulty are in the algorithm. Multi-sensor fusion hardware and software is difficult to separate, but the algorithm is the focus and difficulty, has a high technical barrier, so the algorithm will occupy a major part of the value chain. Algorithm is the core of multi-sensor fusion. Sensor fusion is the data and information acquired by multiple sensors are pooled together for integrated analysis to more accurately and reliably describe the external environment, thereby improving the correctness of system decisions.

 

Control Domain

 

Domain centralized EE architecture will be the dominant automotive EE architecture for a long time in the future, and the domain controller, as the core of the domain centralized EE architecture, will occupy an increasingly important position in the entire automotive industry chain. The corresponding chips and hardware solutions, operating systems and algorithms will become the focus of competition among upstream and downstream manufacturers in the whole industry chain.

 

Robotaxi Business Model

 

The development of intelligent and networked automobiles has become a major trend in the industry, and sharing, which will bring about innovation in the business model of automobiles, is the ultimate goal of the development of the automotive industry, and Robotaxi is an indispensable and important scenario and implementation method.

 

In terms of the development path of the Robotaxi model, the Robotaxi business model is currently charging customers by mileage.

 

The development of Robotaxi model is also facing a series of problems. First of all, the self-driving technology is not yet mature. To realize the concept of Robotaxi's fully driverless cab (eliminating the safety officer), Robotaxi's self-driving level must be above L4. At the same time, capital is consumed and technology iteration is slow. Robotaxi has huge annual technology development costs, but the overall technology iteration cycle is long due to the lack of data accumulation. The operation mode is not clear, and some of the Robotaxi vehicles in China have the problem that they can only run in fixed lines or small closed parks.

 

Robotaxi has great imagination and bright vision, but the stability and universality of self-driving technology are still insufficient, and the commercialization faces two major problems: technology and cost.

 

Vehicle-road cooperation or local driving?

 

The development path of autonomous driving is divided into two types in a broad sense: one is single vehicle intelligence and the other is vehicle-road cooperation.

 

The first option is to achieve autonomous driving by improving the car's own intelligence level, represented by Tesla and Waymo, the former advocates gradual progress (L2 slowly upward iteration), the latter chooses to directly cut into high-level autonomous driving (L4 and above).

 

The latter option is vehicle-road collaboration. The pursuit of autonomous driving through the interconnection of the vehicle and the surrounding things. Coupled with the hot 5G concept (faster peak network transmission rates for faster communication), autonomous driving through vehicle-road collaboration seems more promising. For the car, with the support of vehicle-road cooperation, it is equivalent to having [clairvoyant eyes and ears]. The sensing range can be extended to several hundred meters around the car, so that more accurate decisions can be made.

 

The current state is that the coverage of vehicle-road cooperation is limited and the integration of roadside devices is not high. Roadside monitoring devices mainly use cameras, but they may not be able to provide information when it is most needed, such as extreme weather (heavy rain, fog, etc.). If LIDAR is used instead, the reliability is improved, but the cost is high and the lifetime is limited. In the solution of vehicle-road cooperation, smart road needs smart car, and smart car also needs smart road, which is a complementary process. At present, the development of single vehicle intelligence itself is fast enough, while the development of smart roads is slower. How to achieve flexible vehicle flow control through sensor equipment, improve road efficiency, bring better car experience; provide users with more accurate road status information, provide more efficient travel planning, etc., is the first problem to be solved by the smart road.

 

Participants in the Autonomous Driving Industry Chain [HOT!]

 

Automated driving involves human-computer interaction, visual processing, intelligent decision-making and many other aspects. According to the core technology classification of the autonomous driving industry chain, it can be roughly divided into three categories: perception layer, decision layer, and execution layer. The perception layer involves industry chain links including camera, LIDAR, millimeter wave radar, ultrasonic radar, high-precision map, high-precision positioning, T-BOX, V2X; the decision layer involves autonomous driving chip, autonomous driving controller, operating system (OS), autonomous driving algorithm, simulation test; the execution layer mainly refers to the wire control chassis.

 

A. Autonomous driving chip SOC

 

As the core support for the realization of autonomous driving hardware, autonomous driving chip SoC naturally ushers in a broad development opportunity. In the field of autonomous driving, Mobileye started very early and launched the first-generation autonomous driving chip EyeQ1 as early as 2007. Although the chip is not high in arithmetic power, the software support is very good and can provide a good driving scenario experience.

 

At present, Audi, Tesla, Xiaopeng, Weimar, and a large number of mainstream new energy passenger cars are based on Nvidia's Xavier or Orin chips, the arithmetic power level is mainly 30TOPS, the high-end models launched in 2022, such as the Azera ET7, Xiaopeng G9 may be as high as 500-1000TOPS. The market share is more than 30%, mainly in L3 level autonomous driving.

 

Qualcomm launched the Ride platform at the end of 2020, which can provide different levels of arithmetic power, including 10TOPS of arithmetic power at the L1 level with less than 5 watts of power consumption, as well as more than 100 watts of power consumption and 700TOPS of arithmetic power in the configuration, the power consumption of the whole system will be almost more than 100 watts, but this is more aimed at the models after 2023.

 

Texas Instruments' chips have a larger market share in autonomous driving below L2.5, with a richer product line but the arithmetic power is concentrated in 8TOPS-48TOPS, which is not as good as Nvidia's chips, but the optimization, maturity and development of the chips are very good (TDA4 chips).

 

Horizon Journey Series 5, with an arithmetic power of 128TOPS, CoreChip V9 is also a similar competitor, in addition to the chip of Black Sesame, etc. Domestic chip manufacturers are growing fast, many car companies consider Horizon, Black Sesame, etc. as an alternative to avoid the problem of chip supply; at the same time, the international chip maker's R&D department is not in China, so it is difficult for car companies to learn chip-related technology from international chip manufacturers, which is the advantage of local chip manufacturers.

 

B. Domain controller

 

Autonomous driving domain is responsible for the safety and security of the underlying core data and networking data of the car in the state of autonomous driving, and is the core component to promote L3 and above higher level autonomous driving. Globally, the global Tier1 has basically laid out the self-driving domain controller products.

 

At present, there are four main types of players in the autonomous driving domain controller.

 

1) head new power companies, such as Tesla self-research autonomous driving chip, Wei from research domain controller and then find third-party OEM.

2) international Tier1, and chip vendors, do program integration after the development of domain controllers and sales to OEMs, such as Continental ADCU, ZF ProAI, Magna MAX4, etc..

3) Domain control software suppliers, for example, TTTech and SAIC have established a joint venture, Genesis Intelligent Driving, to support SAIC member companies with autonomous driving domain controller products.

4) Local Tier1, according to the information announced by Nvidia at the October 2021 Cloud Summit, car company customers currently using Nvidia's Orin series of solutions include Mercedes-Benz, Volvo, Azera, Xiaopeng, RISO, SAIC Zhiji and R Auto, and Desai Xiwei got most of these domain controller orders. There are also Huawei, Jingwei Hengrun, and Fritec laying out this-field.

 

C. Camera

 

Under the trend of car intelligence, in-car camera is an important sensing component of ADAS system, benefiting from the increase of ADAS penetration rate and the improvement of intelligent driving level, the number of single car camera is rapidly increasing. From the current distribution and number of cameras of intelligent driving models of new energy and traditional car companies, L1 level mainly carries 1 single or multiple current view camera; L2 level will increase 4 surround view/test cameras; L3-L4 level will climb to more than 13. According to the data of Foresight Industry Research Institute, the average number of CIS equipped in L4 autonomous driving vehicles will be increased from 6 in L1/L2 level to 29, and L5 level will be increased to 32.

 

In-vehicle camera mainly consists of CIS image sensor, module package, optical lens, infrared filter and voice coil motor, the cost share is 50%, 25%, 14%, 6% and 5% respectively.

 

According to China Industry Information Network data, the top five manufacturers of camera modules are Panasonic, Valeo, Fu Tu Tong, Continental, Magna.

 

In the field of car camera lens, the domestic has taken the leading position. Sunwoo Optical tops the list of global car lens suppliers, with a leading market share, and Ovation also quickly cut into the car lens market by acquiring patents related to Fu Tu film lens and Fu Tu Tianjin, becoming one of the main suppliers. The main competitors are Japanese and Korean companies, including McSell, Denshi Sankyo, Seko Light, etc.

 

CIS image sensor field oligopoly pattern is more obvious. ON Semiconductor is deeply involved in the field of automotive electronics and is the global leader in automotive CIS sensors. Howell Technology, a subsidiary of Weir Corporation, has made a breakthrough in the field of automotive CIS sensors, ranking second in global market share after ON Semiconductor. Sony, which has absolute dominance in the cell phone field, has started to increase its layout in the field of in-vehicle CIS sensors. Samsung and Sony - like Sony, as the leading cell phone CIS, entered the automotive market late, but are rapidly cutting into it.

 

D. LIDAR

 

The industry chain of LIDAR is relatively clear. The upstream components of LIDAR mainly include lasers and detectors, master control chips, analog chips and optical components, with obvious overseas advantages. Due to its early start, it has a certain first-mover advantage. Currently, the upstream core components are mainly dominated by overseas manufacturers, with high product reliability, mature technology and wide customer base.

 

1) laser industry: ams OSRAM (AMS OSRAM semiconductor), Lumentum (Lumentum), etc., with years of cultivation to occupy a leading position; domestic manufacturers such as RuiBo Optoelectronics, Changzhou Zong Hui core light semiconductor open breakout.

 

2) detector representative enterprises: mainly First Sensor, Hamamatsu (Hamamatsu), ON Semiconductor, Sony, etc.; domestic emergence of the Chengdu amount of core, Lingming photonics, core vision and other outstanding enterprises, product performance is basically close to foreign levels.

 

3) FPGA chip: usually used as the main control chip of LIDAR, the mainstream foreign suppliers are AMD (acquisition of Ceres), Intel (acquisition of Altera), etc.

 

4) Analog chip: overseas suppliers are the industry leaders in this field with advanced technology, sufficient production capacity and high maturity, the representative suppliers are mainly Texas Instruments, ADI, domestic manufacturers such as Shengbang Microelectronics (analog chip), Sand LJ (analog chip) are starting to make efforts to layout, and the performance is developing in the direction of meeting LIDAR demand.

 

5) Optical components: the technical level of the domestic supply chain has completely reached or surpassed the level of foreign supply chain, and has obvious cost advantages, has been able to completely replace the foreign supply chain and meet the demand for product processing.

 

E. Millimeter wave radar

 

Millimeter wave radar has the characteristics of small size, high cost performance, and can work around the clock, which is the core sensor of autonomous driving. In the global millimeter wave radar market, the top five market players are: Bosch, Continental, Hella, Fujitsu and Denso, among which Bosch, Continental, Hella and Denso are traditional Tier1 suppliers, and Fujitsu is a leading global information and communication technology enterprise, which also has strong strength in the field of millimeter wave radar. Bosch's main millimeter wave radar products are concentrated in 76-77GHz, mainly focusing on medium and long distance detection, and the maximum detection distance of LLR products can reach 250 meters. Continental's 77GHz and 24GHz products are leading, mainly focusing on 77GHz products, and the maximum detection distance of fifth generation LRR products can reach 300 meters. Delphi (acquired by BorgWarner in 2020) is also one of the international companies that monopolize 77GHz technology; while HELLA focuses on 24GHz frequency millimeter wave radar, and leads in short-range millimeter wave radar technology.

 

In the face of foreign companies close to monopoly of the market, in recent years, domestic companies increase the research and development of millimeter wave radar, looking for market breakthroughs to accelerate the pace of catching up with foreign investors. Dickson Westway, Hangzhou Smartwave Technology, Wuhu Senstech, Nanjing I.E. Eye Technology, Suzhou Anzhi Automotive, Beijing Xing Yi to, Shenzhen An Zhijie, etc.. With the growth of local enterprises such as Senstech, Gatland, Kishida Technology, Qingneng Huabo, Micro Degree Core Creation, Sand Jie Microelectronics, Shengde Micro IC, etc., the speed of 4D radar popularity may be further accelerated in the end.

 

F. Algorithm

 

Algorithm is crucial for autonomous driving. As an important application scenario of artificial intelligence technology, autonomous driving cannot be achieved without the large-scale deployment of algorithms, and their effectiveness affects every aspect of autonomous driving.

 

Algorithm providers can be divided into three categories, covering algorithm module providers, algorithm solution providers and scene solution providers.

 

Among them, the

 

1) Algorithm module vendors: provide single - function module algorithms, suppliers include traditional Tier1, such as Bosch, Continental, and Dexiawei, as well as - some software algorithm vendors, such as Minieye, etc.

2) Algorithm solution providers: can provide complete ADAS or autonomous driving solutions, such as Momenta, Minieye, Harness Technology, Zongmu Technology, etc.

3) Scene solution providers: autonomous driving algorithm providers specific to a particular scene, such as Baidu's Robotaxi, Tucson Future's Robotruck, etc.

G. Wire-controlled chassis

 

Wire-controlled chassis technology for self-driving cars, just like human hands and feet - like, decide whether the car can drive properly, as the implementation of the hardware technology, the development of wire-controlled chassis will determine the development of self-driving cars.

 

At present, the new - generation of wire-controlled actuation products are technically perfect, mature process, targeted towards the autonomous driving application scenario, initially entering the mass production stage, the main suppliers are Bosch, Continental, ZF-Trinity, Hitachi, IDEX and Brembo, etc. At present, the world's main line control system suppliers are Bosch, Continental and ZF-Trinity, of which Bosch takes the lead in self-research layout line control, occupying a leading market position.

 

Previous: Electrification Solutions for the Future of Mobility

Next: Electric Power Steering (EPS) Solutions